geographically weighted regression as a predictive tool ...1351572/fulltext01.pdf · this thesis...
TRANSCRIPT
IN THE FIELD OF TECHNOLOGYDEGREE PROJECT CIVIL ENGINEERING AND URBAN MANAGEMENTAND THE MAIN FIELD OF STUDYTHE BUILT ENVIRONMENT,SECOND CYCLE, 30 CREDITS
, STOCKHOLM SWEDEN 2019
Geographically Weighted Regression as a Predictive Tool for Station-Level RidershipThe Case of Stockholm
KARIM OUNSI
KTH ROYAL INSTITUTE OF TECHNOLOGYSCHOOL OF ARCHITECTURE AND THE BUILT ENVIRONMENT
1
Abstract
English/ Engelska/ Anglais
This thesis studies a new regression method, Geographically Weighted Regression (GWR) to predict ridership at the station level for future stations. The case study of Stockholm’s blue line is used as new stations will be built by 2030. This paper is written in English.
Historically, linear regression methods, independent of the geographical location of the
observations, was and is still used as the Ordinary Least Square regression method. With the rise of GIS-softwares these last decades, geographically dependent regression can be used and previous preliminary studies have shown a dependency between ridership and location of the station within the network.
GWR equations for new stations are determined and used to predict their respective
ridership. GIS-data was collected using Geodata and Traffikverket (Traffic Authority) and ridership as well as socio-economic related material for the base year of 2016 was used in order to determine, first, significant variables from a group of candidate ones (Workers, number of bus lines and type of change were chosen) and, second the OLS and GWR equations. Significances of both models were compared and the OLS equation was used in order to determine the hypothetical ridership of the new stations if they were present in 2016. GWR equations were then determined using these calculated ridership of these new stations. Having all GWR equations (each station having its own equation), ridership was thus predicted for the new stations for 2030 using assumptions and planned, programmed development around the stations (population, apartment to be built…) and compared with the official predictions.
The results show that the GWR method, generally, overpredicts ridership when compared
to the official predictions. Many reasons can explain this overprediction like the assumptions made with regards to the number of buses as well as the method followed to calculate the number of workers around each station.
Three main conclusions were drawn for this case study. One main conclusion, specific for
this study and two other, more general, conclusions were deduced from this study. First, GWR is a good predicting tool for future stations that are close to most currently available stations. Second, GWR is a good predicting method for stations where limited changes in the future environment will occur.
2
Sammanfattning
Swedish/ Svenska/ Suédois
Denna avhandling studerar en ny regressionsmetod, Geografically Weighted Regression (GWR) för att förutsäga antal resenärer på stationsnivå för framtida stationer. Fallstudien av Stockholms blå linje används eftersom nya stationer kommer att byggas år 2030. Denna rapport skrivs på engelska.
Historiskt används linjära regressionsmetoder oberoende av observationens geografiska
placering som den ordinarie Least Square-regressionsmetoden. Med ökningen av GIS-programvaror de senaste decennierna kan geografiskt beroende regression användas och tidigare preliminära studier har visat ett beroende mellan antal resenärer och plats för stationen i nätverket.
GWR-ekvationer för nya stationer bestäms och används för att förutsäga deras respektive
antal resenärer. GIS-data samlades in med hjälp av Geodata och Traffikverket och antal resenärer samt socioekonomiskt relaterat material för basåret 2016 användes för att först fastställa betydande variabler från en grupp kandidater (Arbetare, antal busslinjer och typ av förändring valdes) och för det andra OLS- och GWR-ekvationerna. Betydelsen av båda modellerna jämfördes och OLS-ekvationen användes för att bestämma det hypotetiska antal resenärer för de nya stationerna om de var närvarande 2016. GWR-ekvationerna bestämdes sedan med hjälp av dessa beräknade antal resenärer för dessa nya stationer. Med alla GWR-ekvationer (varje station har sin egen ekvation) förutsades således antal resenärer för de nya stationerna för 2030 med antaganden och planerad, programmerad utveckling runt stationerna (befolkning, lägenhet som ska byggas ...) och jämförs med de officiella förutsägelserna.
Resultaten visar att GWR-metoden generellt sett förutsäger antalet resenärer jämfört med
de officiella antalet resenärer. Många orsaker kan förklara denna överförutsägelse som antaganden om antalet bussar och metoden som följdes för att beräkna antalet arbetare runt varje station.
Tre huvudsakliga slutsatser drogs för denna fallstudie. En huvudsaklig slutsats, specifik för
denna studie och två andra, mer generella, slutsatser härleddes från denna studie. För det första är GWR ett bra förutsägningsverktyg för framtida stationer som ligger nära de flesta tillgängliga stationer. För det andra är GWR en bra förutsägningsmetod för stationer där begränsade förändringar i den framtida miljön kommer att inträffa.
3
Résumé
French/ Franska/ Français
Cette thèse étudie une nouvelle méthode de régression, la régression géographiquement pondérée (GWR), pour prédire le nombre de voyageurs au niveau des stations pour de futures stations. L’étude de cas de la ligne bleue de Stockholm est prise vu que de nouvelles stations seront construites d’ici 2030. Cette thèse est rédigée en anglais.
Historiquement, les méthodes de régression linéaire, indépendantes de la localisation géographique de des observations, étaient et sont toujours utilisées comme méthode de régression des moindres carrés ordinaires (OLS). Avec le développement des logiciels SIG au cours des dernières décennies, l’utilisation de régression géographiquement dépendante devient plus accessible et des études préliminaires antérieures ont montré une dépendance entre le nombre de voyageurs et l'emplacement de la station dans le réseau.
Les équations GWR pour les nouvelles stations sont déterminées et utilisées pour prédire leurs nombres de voyageurs respectives. Les données SIG ont été collectées à l’aide de Geodata et de Traffikverket (Autorité des transports). Le nombre de passagers ainsi que les données socio-économiques pour l’année de référence de 2016 ont été utilisés afin de déterminer, en premier lieu, les variables significatives d’un groupe de candidats (travailleurs, nombre de lignes de bus type de changement ont été choisis) et, deuxièmement, les équations de OLS et de GWR. Les valeurs significatives des deux modèles ont été comparées et l'équation OLS a été utilisée afin de déterminer le nombre de voyageurs hypothétique des nouvelles stations si elles étaient présentes en 2016. Les équations GWR ont ensuite été déterminées à l'aide de ce nombre de voyageurs calculé de ces nouvelles stations. Disposant de toutes les équations GWR (chaque station ayant sa propre équation), le nombre de voyageurs des nouvelles stations pour 2030 a donc été prédite à l'aide d'hypothèses et de développements planifiés et programmés autour des stations (population, appartement à construire…) et comparés aux prévisions officielles.
Les résultats montrent que la méthode GWR surestime d’une façon générale le nombre de voyageurs par rapport aux prévisions officielles. Plusieurs raisons peuvent expliquer cette surestimation, telles que les hypothèses émises concernant le nombre d'autobus et la méthode suivie pour calculer le nombre de travailleurs autour de chaque station.
Trois principales conclusions ont été tirées pour cette étude de cas. Une conclusion principale, spécifique à cette étude et deux autres conclusions, plus générales, ont été déduites de cette étude. Premièrement, le GWR est un bon outil de prévision pour les futures stations proches de la plupart des stations actuellement présentes. Deuxièmement, le GWR est une bonne méthode de prévision pour les stations où des changements limités dans l’environnement futur auront lieu.
4
Introduction
Background
According to the World Bank in 2018, more and more people are moving to cities leaving behind rural areas. In fact, since 2007, more people live in these cities than in rural areas for the first time in history. Even if the rate of urbanization is decreasing, it has constantly been positive with a value always superior than 1,9% since 1960. More people in cities means a higher number of people moving around during the day leading to an increase to the number of passengers within the public transportation network. SL, the region of Stockholm authority, quantifies this increase to more than 2 million passengers per day in 2013 from 1,6 million in 2003, a 20% increase in merely 10 years.
Figure 1: Diagram illustrates travelers per day by cars and public transportation in thousands (Trafikförvaltningen, n.d.)
Investing for a more efficient and demand-satisfying public transportation network is thus more important than ever. According to the American Public Transportation Association, “every $1 invested in public transportation generates $4 in economic returns” (American Public Transportation Association, 2019). Moreover, funding for public transportation is increasing from a bit over $40 billion to over $70 billion in 20 years in the United States of America.
Figure 2: Total Funding in Public Transportation in the USA (in billions of 2017 dollars)
This continuous increase is met with a growing concern within the transportation authorities, policymakers and even private firms demanding a more precise, detailed and accurate prediction of ridership in order to explain this never-reaching level of investment. Today’s models, mainly lineal regression models, are used in order to predicted ridership or the transit share for a specific region. However, the assumptions previously made in order to ease their use and limit their cost are leading to errors in predictions, creating uncertainties and
– 6 –
Stockholm is growing – and so is public transport Stockholm County is growing rapidly, in recent years by about 40,000 inhabitants every year. By 2030 the (county’s) population is expected to have increased to about 2.6 million (from just under 2.1 million in 2010). This will increase pressure on public transport services. Roads and railways are already congested, particularly in the central parts of the city and during peak traffic. In-commuting from other counties will also increase and accessibility to public transport will need to be adapted to the changing needs.
PUBLIC TRANSPORT should be perceived as the most attractive form of travel for every-one, including the elderly and travellers with disabilities. It is therefore crucial to the Stockholm region of the future that public transport develop at the same pace, at least, as the population increases and that the entire transport system be planned so as to facilitate public transport’s long-term expansion.
THE COUNTY COUNCIL invests billions in public transport every year. Over the coming years, the county council will be investing more than ever to meet the needs of a growing population. The biggest investments will be made in upgrading the infra-structure but to an increasing degree also new construction and expansion.
MAJOR INVESTMENTS over the next ten-year period include:
• extension of the metro
• the metro’s Red Line
• the Commuter Train programme
• extension of the Roslagen Line.
Public transport’s positive development from 2003 to 2013
The diagram illustrates travel per day by car ( ) and public transport ( ) in thousands.
0500
1000150020002500300035004000
1500
2000
2500
2,000
1,900
1,800
1,700
1,600
Sou
rce: Facts abo
ut SL an
d th
e cou
nty 2014
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013Year
Thousands
Public transport
Car 1,742
2,017
1,785
1,8481,907
1,663
5
risks that are increasingly higher to bear for investors. One of these assumptions is space independency between the observations. In 2000, Fotheringham et al. proved that global regression models estimate a limited number of parameters between the dependent and independent variables with estimated parameters independently determined with regards to spatial characteristics. This lead to a huge disadvantage of this models with regards to observations that are geographically dependent. In fact, multiple studies like the one by Cordazo et al. in 2012, have proven a high correlation between transit use or ridership and geographic locations of the observation, with high spatial autocorrelation with closer observation having a higher influence than farther ones. This means that when using global, traditional regression models in order to predicted ridership, errors are generated due to the assumptions that observation are geographically independent while they are actually not. As explained above, taking into consideration geographic location of the observation at the time was both time and cost consuming. Today, with the development of GIS, Geographical Information System models and softwares during the last decade, makes these excuses obsolete. These new softwares can acknowledge the geographic locations of the observations while developing a regression model. The Geographically Weighted Regression is one of these new methods that can be used in order to explain and (maybe) predicted ridership.
Objectives and Goals
The objective of this thesis is to determine if Geographically Weighted Regression method can be used in order to predict the station-level ridership at future metro stations.
Three kind of stations are to be compared, stations with different geographic locations compared to the rest of the network and built in different changing environment: Stations close to the existing metro stations of the network that will not experience major changes in their surrounding environment, stations built close to the existing metro station that will, however, experience considerable change to their surrounding environment and, finally, stations that will be built far away from the actual network that will experience big change to their surrounding environment.
The comparison between these three types of station will determine if and/or when is are
predictions possible following the method use in this study. The case of Stockholm is studied here.
Thesis Structure and Flow
This paper is divided into seven parts. First, the literature review talks the way predictions were done until today, while introducing GWR, both theoretically and with regards to previous studies using it. Second, a methodology is presented in which detailed steps performed in this study are described in order to reach the results that are needed. Third, Stockholm as a city and its metro network is presented describing the extension of the metro network. This part also lists the candidate variables as well as the assumptions laid for this study. Fourth, results are presented (the models used, parameters, significance of the models and finally the ridership predictions). Fifth,
6
the analysis of these results is done extensively dividing the analysis into general, regional and station-specific. Sixth, the limitations of the studies are described mainly in order to present what should be avoided for future similar studies that are to be performed. Finally, the conclusions are presented both case study specific and generally.
7
ABSTRACT 1
English/ Engelska/ Anglais 1
SAMMANFATTNING 2
Swedish/ Svenska/ Suédois 2
RESUME 3
French/ Franska/ Français 3
INTRODUCTION 4
Background 4
Objectives and Goals 5
Thesis Structure and Flow 5
LITERATURE REVIEW 10
Old Forecasting Methods 10
Geographically Weighted Regression 11 Description 11 Difference between Spatial Autocorrelation and Spatial Non-Stationality: Accuracy of the model 14 Spatial Autocorrelation 15 Spatial Non-Stationality 16
Geographically Weighted Regression as Forecasting Method 17 Previous works and fields in use 17 Comparison between OLS and GWR 18 Predictions using GWR compared to other methods 18
METHODOLOGY 21
STOCKHOLM’S CASE STUDY AND DATA COLLECTION 24
Stockholm’s Metro System: Present Situation, Forecasting and Future Development 24 Stockholm today with ridership and population 24 Preliminary studies with Sampers for Stockholm’s new station 25 General idea with map of the planned extension 26 Transit-oriented development: Stockholm Case Study 28
Data and Candidate Variables 29 Line chosen 29 Candidate Variable 29 Prediction assumptions for the candidate variables 32
8
RESULTS 32
GWR Equations with Existing Conditions 32
Predictions Using the Determined GWR Equations 41
ANALYSIS 45
Division between the North and the South 45
The Model in Numbers 45
General Reasons 48
Bus Assumptions and Effect on Predictions 48
Special Case of Sofia 48
LIMITATIONS OF THE STUDY 49
CONCLUSION 51
REFERENCES 52
APPENDIX 56
Appendix I: Method and Tools in Determining Data in ArcGIS (ArcMap) 56 Income, Workers, Population and Age 56 Road density (m/m2) 56 Number of bus lines at a 200-meter buffer around the entrances of the metro 56 Terminal Station 56 Type of change 57 Commuting distance 57
Appendix II: Table of GWR Equations and Predictions for the 2016 Situation 58
Appendix III: Table of GWR Equations and Predictions for the 2016 Situation with New Stations 73
Appendix IV: GWR Prediction by ArcGIS (ArcMap) 89
9
FIGURE 1: DIAGRAM ILLUSTRATES TRAVELERS PER DAY BY CARS AND PUBLIC TRANSPORTATION IN THOUSANDS (TRAFIKFÖRVALTNINGEN, N.D.) 4
FIGURE 2: TOTAL FUNDING IN PUBLIC TRANSPORTATION IN THE USA (IN BILLIONS OF 2017 DOLLARS) 4 FIGURE 3:TRADITIONAL FOUR-STEP TRANSPORT MODEL (ADAPTED FROM WHITEHEAD & BUTTON, 1977, P.117) 10 FIGURE 4: DIFFERENT TYPES OF THE KERNEL FUNCTION (INSTITUT NATIONAL DE LA STATISTIQUE ET DES ETUDES
ECONOMIQUES, 2018) 13 FIGURE 5:100-METER SQUARES IN RENNES, SAMPLED IN RED (FLOCH, 2015) 20 FIGURE 6: BOX-PLOT OF THE RCEQMR FOR THE HORWITZ-THOMPSON (1), REPRESSION (2) AND GWR (3)
ESTIMATORS (FLOCH, 2016) 21 FIGURE 7: CATCHMENT AREA (SERVICE AREAS) FOR EACH STATION (OLD STATIONS IN BEIGE, NEW STATIONS IN
PURPLE). 22 FIGURE 8: SERVICE AREA FOR THE BLUE LINE STATION TO BE PREDICTED FOR 2030 24 FIGURE 9: METRO NETWORK WITH THE ADDITIONAL STATIONS IN DASHED LINE. 26 FIGURE 10: PROJECTED RIDERSHIP ON THE BLUE LINE DURING THE MORNING PEAK HOUR WITH A FOUR-MINUTE
HEADWAY (NYLÉN, 2017; HARDERS AND BJÖRKMAN, 2016) 27 FIGURE 11: PREDICTED RIDERSHIP ON THE YELLOW LINE DURING THE MORNING PEAK HOUR BY 2030. 28 FIGURE 12: RIDERSHIP VS. NUMBER OF WORKERS 36 FIGURE 13: OLS (UP) AND GWR (DOWN) STANDARD DEVIATIONS 37 FIGURE 14: GWR STANDARD DEVIATION WITH NEW STATIONS 40 FIGURE 15: THE DISTRIBUTION OF THE INTERCEPT OVER THE STUDIED AREA (FROM THE LOWEST IN BLUE TO THE
HIGHEST IN RED; THIS APPLIES TO ALL DISTRIBUTIONS TO FOLLOW) 45 FIGURE 16: THE DISTRIBUTION OF THE WORKER’S COEFFICIENTS OVER THE STUDIED AREA 46 FIGURE 17: THE DISTRIBUTION OF THE BUS’S COEFFICIENTS OVER THE STUDIED AREA 47 FIGURE 18: THE DISTRIBUTION OF THE CHANGE’S COEFFICIENTS OVER THE STUDIED AREA 47 TABLE 1: EQUIVALENT PRESENT STATIONS FOR FUTURE STATIONS (WSP ANALYS & STRATEGI, 2013) ..................... 26 TABLE 2: CANDIDATE VARIABLES EVALUATION AND SELECTION ................................................................................ 32 TABLE 3: SUMMARY OF MULTICOLLINEARITY ............................................................................................................. 33 TABLE 4: PERCENTAGE OF SEARCH CRITERIA PASSED ................................................................................................. 34 TABLE 5: MORAN’S I TESTS ON THE DEPENDENT AND CANDIDATE VARIABLES ......................................................... 34 TABLE 6: OLS EQUATION ............................................................................................................................................. 38 TABLE 7: GWR EQUATIONS ......................................................................................................................................... 38 TABLE 8: PREDICTED RIDERSHIP USING OLS EQUATION FOR NEW STATION IN 2016 ................................................ 39 TABLE 9: GWR EQUATIONS WITH THE NEW STATIONS ............................................................................................... 40 TABLE 10: NUMBER OF ADDITIONAL POPULATION AND WORKERS BY 2030 ............................................................. 41 TABLE 11: COEFFICIENTS AND GWR ESTIMATIONS FOR NEW STATIONS ................................................................... 43 TABLE 12: NUMBER OF ADDITIONAL POPULATION AND WORKERS BY 2030 FOR BARKARBYSTADEN AND BARKARBY
STATION ............................................................................................................................................................. 44 TABLE 13: COEFFICIENTS AND GWR ESTIMATIONS FOR NEW STATIONS AFTER THE CHANGE IN WORKERS ............. 44
10
Literature Review
Having efficient and reliable ridership estimation is important for all stakeholders. Passengers can plan their trips by choosing confidently the time and route of their choice will be sure of their time of arrival to their destination. It can also create a routine when it comes to regular trips, mainly work-bound trips in the morning, increasing adequate planning and thus efficiency and productivity. Transit operators can plan efficiently for the needed capacities and frequencies by securing funds early on while spending them in the required areas and departments. Public authorities and operators can forecast resourcefully the funds needed for the future, the dispatching and evolution of jobs and population on the interested region as well as implementing policies and strategies in order to lead the region towards a more sustainable future.
Old Forecasting Methods
Transit ridership are usually estimated using comparison methods to equivalent situations, professional and elasticity analysis and travel demand models (Litman, 2004; Boyle, 2006). The first models are typically employed for route evaluation while the latter is used in assessing new amenities providing transit ridership only as a part of the prediction with no particular focus on transit ridership and public transportation travel that is treated as another mode (Zhang & Wang, 2014).
Transport forecasting and modelling took a serious turn in the fifties’ when the four-step
model, a method that predicts traffic patterns at an aggregate level, was created (Horowitz, 1984). It has since been the dominant model for transport modelling and was adopted for transit ridership as well over the years (McNally, 2007). The four-step model is characteristic by its four step process by first generating the demand for travel in specific region, second distributing this demand by creating Origin-Destination region pairs, third assigning a mode of transport the travelers are going to use (public transport, car, walking, biking, etc.), and finally by assigning a specific route to each trip.
Figure 3:Traditional four-step transport model (adapted from Whitehead & Button, 1977, p.117)
11
Activity-based models are used for forecasting and prediction. This method focuses on predicting individual travel behavior at a disaggregated level (Hildebrand, 2003).
However, these overall travel demand forecasting methods require a huge number of
surveys, data collection and processing. These are only a couple of reasons why this method is costly to implement and maintain (Marshall & Grady, 2006). They also fail to capture subtle land-use characteristics in specific areas that might influence ridership more or less than in the other region (Cervero, 2006). For these reasons, other methods, such as regression models, were developed in order to have efficient and reliable forecasts and predictions of transit ridership. They are also faster and cheaper to develop. Creating a straight relationship between a couple of predefined factors (independent variables) with transit ridership (dependent variable), regression models are easy to use with less trouble in defining them providing a rapid alternative. According to consulted papers, these predefined factors are usually grouped in 4 categories: Demographic features, socio-economic indexes, land-use arrangement, geographic information.
Nevertheless, most current regression models accept spatial-independence in ridership estimation. The problem with this assumption is the fact that many (if not all) factors are spatially correlated (Zhang & Wang, 2014). New methods must then be utilized to acknowledge this spatial dependency between the different observations.
An Ordinary Least Squared regression (OLS) is a regression method that determines the
parameters of the linear regression model by minimizing the square of the errors. The following equation summarizes this regression.
Geographically Weighted Regression
In the following section, the technical information of the geographically weighted regression is described as well as the different ways to evaluate its significance. Description
Global calibrating models, using all the observations provided for a concerned region, predict global estimates for the whole interested region, while local models, using a handful of observation like GWR predict local models that for each interested observation. The main difference between the two methods is that the first emphasizes on the spatial similarities while the latter emphasizes on the spatial differences in the interested region.
As a reminder, the OLS equation is presented below:
! = # + #1'( + #2'*+. . . +#,'- +./
Where y is the dependent observed variable, xj is the j-th independent observations,
variables or predictors (j = 1, ..., p), βj is the j-th model parameters to be estimated (j = 0, 1, ..., p) and epsilon y is the error at for observation y. An Ordinary Least Squared regression (OLS) is a
12
regression method that determines the parameters of the linear regression model by minimizing the square of the errors. The following equation summarizes this regression. It is important to note that OLS the βj are the same for all the studied region and do not change with space or dependent observation.
In fact, traditional global regression models, like OLS, assume that the whole studied
region can be explained and predicted using one common equation with common parameters on the whole region. This rational is easily discredited by local regression models, like GWR (Fotheringham, Brunsdon & Charlton, 2002). This has been explained in section SOMETHING with case study examples from previous studies.
An advantage of GWR over other spatial methods, like multi-level modelling is that each
calibration yields equations for each observation where each one of them is treated independently from the others, capturing geographic heterogeneity (Zhang et Wang, 2014).
The main idea behind the GWR method is the fact that each point i to be predicted is
surrounded by an area of influence that decreases the farther the sampled observations are from point i, thus creating as many regression equations as there are observations to predicted. This is done by incorporating the geographical coordinates of each observation in its equation.
As stated before, GWR is inspired by OLS. OLS can be, actually, seen as an exception of
GWR where all function is constant over space. Indeed, GWR uses a weighted least squares method to predict the parameters (Fotheringham and Charlton, 1998). The following function summarizes the GWR model:
!0 = #1(30, 50) +78
#8(30, 50)'08 + .0
Where !0 is the dependent observed variable at location (ui,vi), #1(30, 50) is the intercept parameter at location (ui,vi), #8(30, 50) are the independent parameters for observation (ui,vi), '08 are the observation k at (ui,vi) and .0 is the error term for observation (ui,vi). Beta best is thus equivalent to this equation:
#9(:) = (;<=(30, 50);)>(;<=(30, 50)? Where Wi (uivi) is the weighting function. This weighting function is spatially dependent providing a weight for the observation depending on its location from (and thus distance to) the observation that is to be predicted. It can be represented in a diagonal matrix where the primary diagonal line represents the weight function at location i (Fotheringham et al, 1998). Here is the weight matrix used:
13
=(:) = @
A0( 0 … 00 A0* … 00 0 … 00 0 … A0D
E
Where A0F(G = 1, 2, … , H) is the weight given at location j
Coming back to the fact that OLS is an exception of GWR, it is clearer now that the weight function is introduced. In fact, one can assume that the weight function is equal to 1 for all points in the studied area.
These functions vary depending on the predicted information. There are several options from Gaussian, exponential, bi-square and kernel just to name the most used ones.
One can differentiate between continuous weight functions where a weighting value is given to each observation in the studied area from weight function with compact support where the latter tends to a zero-value reached at the determined bandwidth value and assigned to observations having a distance greater than this bandwidth (Institut national de la statistique et des études économiques, 2018). However, according to Brunsdon, Fotheringham, and Charlton in 1998, there are the choice weight function has no significant effect on the results.
Figure 4: Different types of the kernel function (Institut national de la statistique et des études économiques, 2018)
The following curves are written explicitly in the following equations, respectively uniform kernel, Gaussian kernel and Exponential kernel.
AIJ0FK = 1
AIJ0FK = L>(*(MNOP )Q
AIJ0FK = L>(*(RMNORP )
14
Another way to differentiate them is by classifying them in either a fixed or adaptive weight function. The main difference between the two is observation density and sample size and whether the bandwidth is constant or variable. The first kind determines the spread of the weight function according to a fixed distance (bandwidth) identical in the whole studied area to be used when one has high density of observation and sample points. The second determined the spread of the weight function according to the number of neighboring observations a point of interest has led to a greater spread when the density is low and varying the distance (bandwidth) (Fotheringham et al., 2002). The changing function can also determine the optimal bandwidth for highly dense observations. In fact, this optimal value of bandwidth, that is a variable in the weight function equations with compact support, can be determined in other ways for a stated weighting function. The bandwidth value has the biggest influence on the results (Institut national de la statistique et des études économiques, 2018). Examples of these data-driven criteria are cross-validation (CV), generalized cross-validation (GCV), Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). The most in use are CV and AIC and are, thus, presented here (Fotheringham et al., 2002).
ST =7
D
0U(
[!0 − !XY0(ℎ)]*
Where !XY0(ℎ) is the value of y at i predicted where developing the model with all observation except !0. The optimum value of the bandwidth would be 0 if all the observations are used to estimate the model, meaning that the only available point in the model is !0, leading to !0 = !X. Generally, the bandwidth that minimizes CV is the one that maximizes the predictive capacity of the model.
\]S(ℎ) = 2H ^H ^H(_X) + H ^H ^H(2`) + H aH + (b)
H − 2 − (b)c
Where n is the sample size, _X the estimate of the standard deviation of the error term, (b) the trace of the projection matrix of the observed variable y on the estimated variable !X.
When using one of these two statistical criteria, the bandwidth is determined when minimizing their values. The main difference between the two is that CV maximizes the predictive power of the model while AIC compromises between this predictive power and the model’s complexity. The weaker the bandwidth, the more the global model is complex. In general, AIC determines larger bandwidth than CV. Difference between Spatial Autocorrelation and Spatial Non-Stationality: Accuracy of the model
The estimated GWR model needs to be evaluated and diagnosed passing by the same process of global models.
The estimated GWR can be evaluated thanks to coefficient of determination (R-Square), t-values and p-values.
15
d* = 1 −∑0 .*
∑0 (!0 − !)*
Where . is the residual between the observed value and the predicted one by the model, !0 is the observed value i (or at location i if dealing with GWR) and ! is the average of the observed values. As stated before, the Akaike Information Criterion (AIC) can determine the optimal bandwidth for a given weighting function in the studied area. However, AIC can also evaluate the goodness-of-fit of the model. In general, an AIC value greater than 3 suggest a good fit of the model (Fotheringham et al, 2002).
In addition to these estimations and because local statistics are considered spatially disaggregated compared to global models, new evaluation statistics must also be performed to evaluate the model for local characteristics.
These characteristics can be classified into two categories: spatial autocorrelation and spatial
non-stationality (Anselin, 1999). They have been challenging issues to deal with according to Fotheringham in 2002 and GWR permits to consider while considering the coordinates of the observation for spatial autocorrelation when calculating the intercept and spatial non-stationality when estimating the parameters. On one hand, spatial autocorrelation refers to an interaction in space, in other words the value of a certain variable compared to the value of its neighbours. On the other hand, spatial non-stationality refers to the structure in space (Anselin, 1999).
Spatial Autocorrelation
Spatial autocorrelation can be tested using the error of the GWR model created. In fact, GWR
hypothesis that the error terms are identically distributed. Thus, a test for validating or not this independence of the distribution is the establishment of a hypothesis test (Leung, Mei and Zhang, 2000).
Ho: No spatial autocorrelation among the disturbance.
H1: There exists either positive or negative spatial autocorrelation among the disturbances with respect to a specific spatial weight matrix W.
In order to accept or reject the null hypothesis, the test statistics Moran’s I is used. The values
range between -1 and 1, where 0, theoretically, represent no spatial correlation (Rosenberg, 2010). However, for a defined sample size, the value representing no spatial correlation is -1/(N-1) where N is the number of spatial observations: it is the expected value. This value (or 0) is the expected value of the null hypothesis. Moran’s I is given as follows:
]1 =.̂<g<=g.̂.̂<g<g.̂
16
Where W = a specific symmetric spatial weight matrix of order n; and N
g = ] − h = ] −
⎣⎢⎢⎡;(
<[;<=(1);]>(;<=(1);*<[;<=(2);]>(;<=(2)
…;D<[;<=(H);]>(;<=(H)⎦
⎥⎥⎤
A probability must be thus determined as well in order to accept or reject the null
hypothesis. The equation given below calculates theoretically this probability when the Moran’s I value is less than a given value p (Leung, Mei and Zhang, 2000).
o(]1 ≤ q) =12−1`rs
1
t:H t:H[u(v)]vw(v)
Jv
Where u(v) = (
*∑D0U( xqyvxH xqyvxH(z0v), w(v) = ∏D
0U( (1 + z0*v*)1,*| and z0 =
g<(= − q])g.
This equation can be simplified after multiple assumption to the following expression (Leung, Mei and Zhang, 2000):
r}
1
t:H t:H[u(v)]vw(v)
Jv
Spatial Non-Stationality
For GWR, the dependent and a given independent variable are linked geographically. In other
words, the hypothesis for GWR is that the variables are stationary in a given geographic area. In order to evaluate the efficiency of the model, it might be interesting to test the spatial non-stationality of the variables (Institut national de la statistique et des études économiques, 2018). Technically, the calculation of the variance of the variables for a given variable should be able to give a satisfying answer (Fotheringham et al, 2002):
Txq~#9(:)� = [(;<=(:);)>(;<=(:)][(;<=(:);)>(;<=(:)]<_*
Where _* = ∑0 (!0 − !X0)/(H − 25( + 5*)
However, the theoretical distribution of each variable is unknown leading to a difficulty in
using the above method. Thus, another method, the Monte Carlo Simulation, is used to help reject or accept the null hypothesis of the following hypothesis test:
H0 : ∀k, βk(u1,v1) = βk(u2,v2) = ... = βk(un,vn)
H1 : ∃k, all βk(ui,vi) are not equal.
17
In fact, if there is spatial non-stationality in the studied area, the locations (coordinates) of the observations are irrelevant and changing them will yield the same value of the variance. When dealing with the Monte Carlo Simulation, the geographical coordinates of the observations are permuted n times, finding n spatial variance estimations of the observations. The p-value of the spatial variability of the coefficients is then estimated. This p-value can determine if the null hypothesis should be accepted or rejected (Institut national de la statistique et des études économiques, 2018). Finally, the bandwidth can give an indication of the efficiency and reliability of the model. Its value is very important and when compared with the extent of the studied area can give, even if not precise, important information about the model nay if GWR should be even used in the first place.
On one hand, if the bandwidth yields toward the maximum value possible (over the whole studied area), local autocorrelation and spatial stationality are weak and GWR should not be used. On the other hand, if the bandwidth is really small, it is important to check for randomness in the process (Gollini et al, 2015).
Geographically Weighted Regression as Forecasting Method
Geographically Weighted Regression or GWR, developed by Fotheringham and Charlton in 1998, is a new regression model inspired by the usual Ordinary Least Square (OLS) method regression that, however, takes into consideration spatial dependency when forecasting equations and its parameters. The authors refer to a family of “spatial adjusted” regression. Previous works and fields in use
Since its introduction as a spatial data analysis in the late 1990’s, GWR has been used in a large number of areas. From health and healthcare (Zhang, Wong, So & Lin, 2012) and forestry (Pineda, Bosque-Sendra, Gómez-Delgado & Franco, 2010) to real estate (Dimopoulos & Moulas, 2016; Institut national de la statistique et des études économiques, 2018), passing by land and urban space use (Luo & Wei, 2009; Tu & Guo, 2008) and poverty rates (Floch, 2016), GWR has been more and more present in transport science, mainly as an explanatory method.
The field of transportation was no exception. Determining the explanatory variables in order to identify potential causes and relations with transportation related issue is the main use today of this relatively new technique. Traffic accidents, average commute distances, transport-land use interaction and influence and public transport share (Chow et al, 2006) are only a couple of fields that were tackled by GWR over the years. In 2015, Qian and Ukkusuri evaluated after a comparison of the performances of the Ordinary Least Square (OLS) and GWR, the causes behind the taxi ridership in New-York city. Liu, Ji, Shi and Gao presented a research on the effect of the built environment on student’s metro commuting to their schools and back home in Nanjing, China.
However, ridership forecasting on a station-based level is not developed enough when it
comes to utilising the GWR method. A couple of preliminary studies in this field were, nevertheless, presented having mainly as the core subject the comparison of the level of efficiency and reliability between OLS and GWR. In 2015, Chiou, Jou and Yang determined the predictors
18
for ridership data for the state of Taiwan with the help of both OLS and GWR. Another study conducted by Blainey and Mulley in 2013 also determines the predictors for ridership data for railway stations in the Sydney region of New South Wales by also comparing both OLS and GWR methods. Studies in Madrid, Spain, Sydney, Australia and Adelaide, Australia conduct similar studies by examining the local characteristics of their respectable studied areas. They also predict some ridership data with the models that they created (Cardozo, García-Palomares, Javier Gutiérrez, 2012; Somenahalli, 2011; Blainey, Mulley, 2013). Comparison between OLS and GWR
As stated above, global regression models, like OLS, assumes that the relationship between the dependent and independent variables are uniform over the study area, ignoring spatial characteristics such as distances to stations. OLS does not consider variations due to spatial autocorrelation (Fotheringham, Brunsdon and Charlton, 2000; Lloyd and Shuttleworth, 2005; Cardozo, 2012). In 2010, Harris et al. compared multiple methods, mainly variations of Kriging, multiple regression methods and GWR and concluded that GWR-based models out-performed MLR models.
Previous studies that have focused on comparing OLS and GWR methods have frequently
discovered that GWR provides more predictability than GWR due to this concern for spatial correlation.
In fact, according to Hadayeghi, Shalaby and Persaud, estimation errors are smaller in a
majority of cases when using GWR compared to OLS. They explain this result by claiming that the problem of spatial autocorrelation is reduced nay eliminated. This is the case for Cardozo, García-Palomares and Javier Gutiérrez when, in 2012 they compare station-level ridership forecasting between the OLS and GWR methods. Their analysis of the residuals proved better results with GWR than with OLS. In addition, when comparing Moran’s I values that were calculated, the one generated for GWR was closer to the expected value than the OLS one, concluding that spatial autocorrelation was reduced. In this same study, GWR performed better with p-values and z-values, describing less variance and greater likelihood for random distribution.
Another study, already presented above, compares a global regression model with GWR
in Taiwan and found an adjusted R-Squared for GWR more than 0,2 units higher than the traditional MLR model in addition to a better performance with regards to spatial autocorrelation for the GWR model (Chiou, Jou and Yang, 2015). Zhao et al. also, in 2005, found a better prediction performance with GWR with regards to OLS for the county of Broward, Florida.
Predictions using GWR compared to other methods
With the establishment of the formulas and equations for each observation, one can use the
calculated coefficients to analyze how relationships vary across the studied area and study any possible patterns they might create. These coefficients can provide local understanding of the observed dependent variable (Fotheringham et al, 2002). In fact, these coefficients, seeing that they are localized, can provide information on the influence of changes in the station’s direct
19
environment will have on the dependent variable, in this case station-level ridership. Any evolution of population, jobs or even land use can provide detailed and specific information for the station.
This leads to a more accurate and detailed forecast, compared to a generalized forecast with
standardized coefficients for the whole studied region when dealing with global regression models (Lloyd, 2010). The model(s) can be evaluated over the whole studied area, determining the areas with better fits, variations of estimated coefficient and significance.
Any future development, such as new residential buildings or workplaces can be thus
evaluated independently for each station environment and area. In fact, for example, an increase in the number of jobs in one area can lead to an increase in ridership while in another it can have the opposite effect. This is where localized policies and plans, land-use for example, come in action and can be implemented, with more insurance and less risks with these more realistic predictions, in the area in order to improve ridership and/or the level of service of the station and/or public transport network. De Smith et al. (2009) call in “place-based” techniques.
GWR can also be used in predicting models in areas where dependent variable observations
are not available. The use of approaches based on the use of “Best Linear Unbiased Predictors (BLUP) estimators is more and more common nowadays (Chambers and Clark, 2012). This method is based on the replacement of non-observed dependent variables by predicted values thanks to a model where the parameters are estimated from observed dependent variables. Recent literature seems to prefer the use of GWR in estimation methods rather than methods originated from other methods. The fact that GWR considers spatial heterogeneity is regarded as theoretically improve the precision of the estimators.
Floch presented in 2016 at the JMS (Journée de Méthodologie Statistique) a study of
Rennes’ Iris zones where they wanted to predict the number of households with low incomes in areas of the city that was not observed. For that reason, and having the number of households beneficiation from free health care due to low incomes in all the studied area, they calculated first the GWR models for all the observed Iris zones determining the equations.
20
Figure 5:100-meter squares in Rennes, sampled in red (Floch, 2015)
Each Iris zone is divided in 100-meter squares. Presenting the notation used, y represents the number of people with low income, x is the number of people having free health care. In U, all squares (N=2141) are assigned their coordinates, the number of low-income people, the number of people with free health care and the Iris it belongs to. A sample s of n/N=40% is selected. r is the complement of s in U where all yi’s are known in i ∈ s and xi are known for all squares in i ∈ U.
In order to determine the predicted number of households with low incomes, three different estimation methods (Horvitz-Thompson, the classic regression one and the GWR one respectively) were calculated.
Determining K=1000 estimators by Iris, the relative quadric average error of Monte Carlo
estimation is calculated before calculating its square root (RCEQMR).
ÑÖÜ áv/à (G)â = ä>(7ã
8U(
(v/à (G)8 − v/(G))*
dSÑÖÜd áv/à (G)â = åÑÖÜáv/à (G)â
v/(G)
Where v/à (G)8 is the estimator for the total of the variable y of the Iris j and the simulation k.
21
Figure 6: Box-Plot of the RCEQMR for the Horwitz-Thompson (1), repression (2) and GWR (3) estimators (Floch, 2016)
One can realize that the best outcome was the GWR estimation. In fact, with a RCEQMR of 0.4, the Horwitz-Thompson estimator was the least performing of the three. The RCEQMR of both the “classic” regression and the GWR estimators are quite similar with 0,178 and 0,156 respectively. This is where the box-plot gives more information about these two RCEQMR showing a smaller spread regarding GWR.
It can be thus safely said that the results showed the GWR estimator as the best compared to the other methods.
Methodology A precise and extensive methodology was developed in order to determine the GWR
equations for each station, for present as well as future conditions. The software used is ArcGIS (ArcMap).
The first step is to choose a line to investigate thoroughly. In fact, even if evaluating the
whole network would also yield interesting findings, limited time and resources constrained to selecting one specific line.
The next step is to gather the data needed to produce the analysis, required by the candidate
variable such as population, income, age, type of station and distance to the center of the network. Having gathered all the data needed to create both OLS and GWR equations, catchment
areas (or service areas) around each metro station were defined. In fact, a catchment area is considered as the area around the station where the walking distance to the station is 800 meters or less, representing 10 minutes or less of walking time. This number was determined after previous literature and research review. It was evaluated that a station’s “neighborhood” or the
22
willingness distance to walk to and from a rail (and thus a metro) station is 800 meters (O’Neill et al., 1992; Hsiao et al., 1997; Murray, 2001; Zhao et al., 2003; Kuby et al., 2004; Sallis, 2008; Gutiérrez et al., 2011). In 2008, Gutiérrez et al. proved that network distance provides better estimates than Euclidean distances. Following this reasoning, determining the 800-meter catchment area following the road network around the respective stations was executed. It was also proven that riders are more willing to walk a larger distance at end stations.
Figure 7: Catchment Area (Service areas) for each station (old stations in beige, new stations in purple).
New station names are also presented on this map.
T-Centralen is present here and on all following map by a black point for orientation.
At some locations, mainly in the central area of the urbanized region and of the network,
the dense metro network lead to stations being situated at less than 1600 meters, meaning that an overlap in catchment areas is unavoidable. Following the reasoning described above, possible riders have more than one station to choose from within walking distance in order to travel. In order to avoid double counting when determining the equations and forecasting the ridership and following the steps of various previous studies, Thiessen polygons as catchment areas were generated for these special cases, meaning that possible riders always choose the closest stations at walking distance (Cardozo et al., 2012; Zhang and Wang, 2014). This method is also reinforced by Wardman’s 2004 study in Manhattan that suggest that riders are more likely to choose the closest station due to a higher out-of-vehicle value a time compared with in-vehicle value of time.
The next step is to determine the significant variables from the previously presented
candidate variables as well as the best OLS equation. In fact, ArcGIS permits the evaluation candidate variables by creating OLS regression equation. The most significant equation is chosen in order to proceed with the analysis. The choice is made after comparing the significant indicators such as p-value, R-Squared, the variance inflation factor and the global Moran’s I p-value. Moran’s
23
I is also performed individually on each candidate variable determining the most significant variables with regards to spatial correlation.
Using ArcGIS, the now determined OLS equation is used to forecast ridership for the whole
network, at the future stations that are being built, as if they were existent today. Even if these ridership computations are completely hypothetical, they are necessary in determining GWR equations for these stations (and all station more generally).
In order to formulate GWR equations, the weight function must be determined. Multiple
functions are adequate for the job, however, according to Zhao et al. in 2005, adaptive kernel does not have a limited number of observations meaning an advantage for observation on the limits of the study area.
Having the knowledge of all dependent and independent variables on the whole system for 2016 including the hypothetical ridership for the new stations on the blue line as well as the most adequate weighted function for the presented situation, GWR equations can be computed by ArcGIS for all stations on the blue line for current conditions. These different equations are then evaluated and compared with regards to significance and goodness-of-fit using R2, AIC, p-value and t-test methods with regards to the previously computed OLS equation.
Having the GWR equations for the whole network, the chosen line with its new stations is
isolated and updated with the predicted data for the time of prediction in question. Depending on the variables that were chosen, the data needed would be different. Some data might be lacking. Some assumptions can thus be made in order to remedy this. Population data, when detailed estimations are lacking, can be determined by multiplying the official number of apartments planned to be built by the average number of persons per household. When this is also lacking, the projected evolution of the population in the region can be used for the service areas. Regarding workers, they can be obtained by multiplying the population by a ratio Population to Workers determined by the stations’ equivalent stations (stations that are present today in the network with similar characteristics to a future stations). Regarding the median income, a trend line is determined from past statistics and interpolated into the time of prediction. The change in land use incorporated and updated to the state of the time of prediction should also be considered. Other information is assumed to remain unchanged from today.
24
Figure 8: Service Area for the Blue Line Station to be predicted for 2030
Having determined the ridership according to GWR models, ridership is compared to official forecasts and analyzed.
Finally, each new station is evaluated with regards to its GWR model and the ridership it
forecasted with regards to the official forecast. The determined parameters can help determine policies that can be proposed. For instance, the choice of building new apartments in specific areas can also be analyzed and other areas, not considered, can be proposed.
Stockholm’s Case Study and Data Collection The following section presents the present metro situation in Stockholm as well as the plans for a future extension of this current network. It also presents the data collected through candidate variable that can explain the ridership for current and future states.
Stockholm’s Metro System: Present Situation, Forecasting and Future Development
Stockholm’s metro network developed over the years starting in the 1950’s to become a complex start network with T-Centralen in its core. Forecasting for the ridership has been done using a special model, Sampers, developed by Trafikverket, the transportation authority in Sweden. This model has been used in preliminary studies to forecast the ridership on stations and lines to be constructed by 2030.
Stockholm today with ridership and population
In the 1940s, Sweden decided to build a metro network in its capital, even if, at the time, the number of inhabitants of Stockholm did not technically require, this new form of public
25
transportation. The green line in the 50s, the red one in the 60s and the blue one in the 70s were constructed with gradual extensions over time.
Today the system welcomes almost a million passengers every day (MTR, 2018). This
number is expected to increase by 170 000 in 2030 (Stockholms läns landsting, 2016). In fact, Stockholm is expected to welcome between 30 000 to 35 000 new residents every year until 2030, making it the fastest-growing European capital. The last seven years saw a growth by 250 000 people in Stockholm county. The current capacity of the network would not sustain this influx of new residents.
In addition, Stockholm is facing, with the rest of Sweden, a housing scarcity problem. The
official rent-controlled queue has around 500 000 people waiting in line for an accommodation. The average waiting time is nine years with some neighborhoods having an average time up to 20 years. For this reason, Stockholm’s city council is backing the construction of 40 000 new permanent homes by 2020 and 100 000 more by 2030. Movable modular homes and a big co-living space for global entrepreneurs are also considered to be adopted to ease the housing problem (Savage, 2016).
For these reasons, Stockholm has decided to expand and enlarge its public transportation
network, namely its metro network, constructing new housing next to these new stations. Preliminary studies with Sampers for Stockholm’s new station
Preliminary studies were done starting from 2007 and were updated continuously with the progression and direction the metro expansion project took.
The forecasting part and projection of ridership for the new completed network by 2030,
including these new stations were determined thanks to Sweden’s national model system, Sampers. According to Trafikverket’s website, updated in 2018, it uses a cross-sectional analysis for determining future passengers and traffic volumes for different scenarios. Its main variables are GDP, fuel prices, employment and population growth. In addition to forecasting new ridership or/traffic flows, Sampers provides impact assessments and investment calculations for land-use or transport changes such as new residential project or infrastructure project for possible transport policy measures.
According to Prognos över resandeutveckling for both the extension of the blue line and
the construction of the new yellow line, the traffic analysis for 2030 was executed using both PTV Visum and Sampers. Ridership was thus found using these methods. However, the report presenting the results for the yellow line presents limits for PTV Visum. In fact, “VISUM is a generalizing model that includes the whole Stockholm län and whose strength is, first, providing a general analysis”. It continues in warning that detailed analysis of the results should be done in caution.
Equivalent present stations to potential stations during the preliminary analysis has been
determined by WSP in its Effekter på värdet på handelsfastigheter vid etablering av nya
26
tunnelbanelinjer i Stockholmsregionen report. The following table presents the equivalent station of the selected stations of the blue line:
Table 1: Equivalent Present Stations for Future Stations (WSP Analys & Strategi, 2013)
Station Equivalent Station
Station Equivalent Station
Kungträdgården (New)
Hötorget Järla Västra Skogen
Sofia Mariatorget Nacka Solna Centrum
Hammerby Kannal
Alvik Barkerbystaden Farsta strand
Sickla Järla Barkerby Station
Farsta Strand
General idea with map of the planned extension
As explained above, the new stations will be built on two different lines: the blue line and
a new yellow line.
Figure 9: Metro network with the additional stations in dashed line.
Notice that the current arm of the green line towards Hagsätra will be integrated to the blue line by 2030 (Stockholms läns landsting, 2016)
On one hand, the blue line, already existent, is facing an extension from both sides. In the
north, a small extension will see the line grow by two stations after Akalla, Barkarbystaden and Barkarby. It is in the south that the extension is going to be significant with the addition of five new stations: Sofia, Hammarby kanal, Sickla, Järla and Nacka. In addition, at Sofia, the blue line
27
is splitting in two different branches, one that goes until Nacka passing by all the named stations and another that is connecting to the present green branch to Hagsätra at Gullmarsplan, turning it blue. The new station, Slackhusetområdet, is replacing the present Globen and Enskede gård stations (Nylén, 2017). The below figure XX presents the forecasted ridership on the whole line, with an obvious emphasis on the new parts of the line.
Figure 10: Projected ridership on the blue line during the morning peak hour with a four-minute headway (Nylén, 2017; Harders
and Björkman, 2016)
On the other hand, the yellow line is going to be built from scratch from Odenplan to Arenastaden, with two stations between them, Hagastaden and Södra Hagalund. After Odenplan the line is joining the green line continuing to either Farsta Strand or Snarpnäck. The figure below (Figure XX) present the ridership between Odenplan and Arenastaden. It is important to note however that the ridership presented here was determined with the assumption that Odenplan is the terminal station and with a line having only three stations instead of the current four (Harders and Björkman, 2016). In fact, the decision to change the initial assumptions was made in 2017.
28
Figure 11: Predicted ridership on the yellow line during the morning peak hour by 2030.
The upper graph represents the ridership heading towards Odenplan and the lower one heading towards Arenastaden. Legend: Dark red: Boarding with a five-minute headway, Pink: Boarding with a 10-minute headway, Dark blue: Alighting with a five
minute headway, Light blue: Alighting with a 10 minute headway, Bold line: Load for a 5 minute headway, Light line: Load for a 10 minute headway (Harders and Björkman, 2016)
Transit-oriented development: Stockholm Case Study There have been multiple studies drawing a connection between transit use and Transit-Oriented Development (TOD). In Stockholm, the planned construction of multiple residences as well as workplaces around not only the new stations but current stations show the interest of Stockholm’s policy makers in TOD. By definition, TOS is the development of urbanized neighborhood where transit is easily accessible. Mixed land use as well as pedestrian oriented mobility in such regions is a core aspect in addition to a densely designed roads and buildings (Zhao et al., 2005).
Even if no clear conclusion can be drawn, some studies, like the one done by Parker et al in 2002, show transit ridership can be increased by up to 40% at individual stations after TOD was implemented around them. A case study of Portland asserts this finding after a survey in 1994 concluded that transit share is higher and car ownership lower in TOD in comparison to traditionally developed neighborhoods (Lawton, 1997).
29
With regards to Stockholm län a goal of 140 000 new residences by 2030 is shared among the different municipalities, mainly Nacka, Järfalla and Solna as the latter will have the new stations built within their municipal boundaries. A total of 78 000 of these new residences will be built around these stations (Stockholms läns landsting, 2016). For example, around the future Hammerby Kannal station, around 2140 residences and a total of 73 000 square meters of new locals and offices will be built (Stockholm växer, 2018).
However, in order to strengthen and increase the share of transit in the region, existing
stations like Kista, Rinkeby and Kristineberg will also see a densification of their neighborhoods with additional residences and workplaces. As an example, the area around Kista will see the construction of a 1600 new residential building and a couple of new office buildings.
These projects are intended to densify and diversify, if not already present TOD’s, potential
one in order to increase the share of transit in the whole Stockholm region in general and more specifically, lead to relatively high transit use around future stations.
This is where GWR comes in. In fact, according to Somenahalli, in 2011, GWR were better
in developing the relationship between transit use and TOD’s.
Data and Candidate Variables
Line chosen
The blue line was selected to be investigated in this case. There are mainly two reasons behind this bias. On one hand, the fact that the blue line is indeed expanding greatly, will thus allow the careful analysis and discussion over both old and new stations, discarding both green and red lines. On the other hand, even if one can consider the newly constructed yellow line as part of the green line, the preliminary studies were conducted as if the line would end at Odenplan with a missing station leaving behind problematic results for this study. The complexity and uncertainty of the newly constructed green line regarding, for example, headway, the number of stations and the number of branches were also additional reasons that favored the use of the blue line for this study.
All stations, as explained in the methodology were assigned an 800-meter service area
each, except for terminal stations, in this case, Nacka Barkarby Station and Hjulsta where they were assigned a 1000-meter service area each.
Candidate Variable
The choice of candidate variables was established thanks to studies while insuring the logic
with the case in hand, i.e. Stockholm. The dependent variable will be the number of boarding passengers at each station during
the morning peak hour, meaning between 7:30 and 8:30. This criterion was selected as a simple goal to be able to compare predicted data from the model developed here with official predictions. The present (2016) data was available in the yearly published report by SL, AB Storstockholms
30
Localtrafik: SL och ländet 2016. This data is the base for the OLS and GWR equations developed for forecasting and evaluation for 2030 situation.
Multiple independent variables were chosen in order to explain the ridership starting with
the socio-economic factors and ending with accessibility ones. Population, income, age distribution and workers for 2016 were all acquired from the
Läntmateriet and GeoData Portal website, respectively the official Swedish authority in gathering statistical and infrastructural information and the website where this information is published. Both these websites make this data public for research purposes. This data sets were already divided into small area, called SAMS in the data. The road network was found on NVDB website. It provided the road network for the whole Stockholm Län. Finally, the metro network for 2030 was provided by Torbjörn Ekerot and Henrik Sarri, respectively an IT-manager from SLL and Metria. The shape file had also the network of the other mode of public transport for the county, such as the bus network and stops as well as the commuter rail (pendeltåg) network and stops. It also assigned each station, present and future, if it is to be considered a significant changing point, a regional one, or not at all. The latter information was taken as a base to determine the type of change that takes place at each station. Unfortunately, detailed land use was only available for current state and only for the municipality of Stockholm. It was thus disregarded as a candidate variable.
The first explanatory variable is the density and size of the population living around the stations. According to Messenger and Ewing in 2007, the relationship between the public transport ridership and population density, even if not direct, exists and is regularly used a factor to justify transportation station expansion or upgrades. Multiple other authors have also asserted the existence of this relationship (Javier Gutiérrez et al, 2011; Sekhar Somenahalli, 2011). In 1996, Seskin et al. presented sufficient evidence to establish a positive relationship between the two.
The second potential explanatory variable is the number of workers around a station. In
fact, even if one can easily hypothesis and deduce it from what was explained with the population explanatory variable, Murray et al. (1998) discovered that the more workers live around a transit service the greater the probability of the latter will be used.
The third potential significant variable is the income of the population respectively to where they live. In fact, it was used in Chow et al., in 2006 and proven to be a significant variable. In addition, an increase of income in specific areas leads to a decrease in transit use for the benefit of the car (Gómez-Ibáñez, 1996; Wachs, 1989; Kitamura, 1989).
The fourth but last socio-economic variable considered is age. In fact, multiple previous
studies (Cristaldi, 2005) consider it and even incorporate it in their respective models like in Bernetti et al, in 2008. Age groups can be created, with the first one between 0 and 19 years old, representing mainly the minor, non-active and unlicenced population, the second between 20 and 64, representing the active population and possibly licenced population and the last third group from 64 onwards representing the retired population. These groups were defined accordingly given the available data for Stockholm and the way the age groups are defined in previous literature.
31
The fifth candidate but first accessibility variable is road density. It was used in multiple previous studies such as Cardozo et al in 2012 and Zhao et al in 2005. In fact, road density can determine to a certain extent the accessibility of the metro station and consequently the number of alternatives to reach this station leading to a shorter and easier reach with high road density.
The sixth one would define the number of bus lines in a 200-meter radius around each
station. A study in 2004 by Kuby et al discovered a connection between feeder modes, for instance bus stops or bus lines accommodating a specific station and ridership at the station in question.
The seventh candidate variable is the type of station itself one is dealing with, mainly with
regards to terminal stations and change and transfer stations. The first type of stations tends to attract residents for larger areas than intermediate stations due to the fact that this station is the closest one to the network inclining riders to walk more than for other stations (O’Sullivan and Morral, 1996). The second type of stations attracts more riders than normal stations (Gutiérrez, Cardozo and García-Palomares, 2011). In fact, be it an interchange station or an intermodal one, they both usually have higher boarding than non-interchange non-intermodal stations. Dummy variables for both these kinds of stations can be used as it was done in Kuby et al, in 2004.
The last accessibility and eighth candidate variable is the commuting distance to the central
business district (CBD) or the central region of the network. In Stockholm case, T-centralen was assumed to be the central point of the network, being the start of the star network system of Stockholm and where all line meet. This was chosen after studying both the papers of Pushkarev and Zupan (1982) and Kuby et al (2004) where it was defined that passengers usually commute to the central part of the network, especially during the morning’s peak hour to reach their places of work.
Finally, the last candidate variable is the land use around the station. According to multiple
studies, such as the ones by Parsons Brinckerhoff in 1996, land use plays a role in transit ridership. In a study by Bhat and Gossen in 2004, an equation was developed in order to quantify the type of land use present in the area of interest, categorising land use in three different groups: residential, commercial/industrial/office and other types.
32
The value ranges between 0 and 1 with 0 being no land use diversity and 1 being perfect land use diversity. This equation was used in the study as multinomial logit model variable for the San Francisco area. Prediction assumptions for the candidate variables
When the predictions are made, these assumptions are taken for the following independent variables. Regarding the population data, lacking detailed and precise number of inhabitants around the stations, except for Barkarbystaden and Barkarby Station where the exact number of inhabitants is known, the number of apartments planned to be built by 2030 are multiplied by the average number of persons per household. An increase in the population of 5% and this according to RUFS (Regional utvecklingsplan för Stockholmsregionen, 2010) is done on remaining stations where there is a lack of information with regards to specific population evolution and the number of apartments. Workers also lack detailed predictions. They are thus multiplied by a ratio Population to Workers determined by the stations’ equivalent stations, presented in an earlier part. Regarding the median income, a trend line is determined from past statistics and interpolated into 2030. Other information is assumed to remain unchanged from today.
Results The main results are split into two parts, the first being for present conditions where the GWR equations were determined for the blue line and the second presenting the results and ridership by station for 2030.
GWR Equations with Existing Conditions
The presented candidate variables were analyzed on ArcMap using a tool called Explanatory variables.
Table 2: Candidate variables evaluation and selection
(AdjR2 is Adjusted R-Squared, AICc the Akaike's Information Criterion, p- value the Koenker Statistic p-value, VIF the Max Variance Inflation Factor and the variable’s significance at 0,01 is in yellow)
Number of Variables
Highest AdjR2
AICc p-value
VIF Model
1 of 11 0,33 1552,13 0,01 1,00 Bus
0,23 1565,55 0,10 1,00 Workers
0,22 1566,38 0,09 1,00 Age 2
2 of 11 0,47 1530,64 0,01 1,04 Workers Bus
33
0,46 1531,71 0,01 1,04 Age 2 Bus
0,45 1534,35 0,01 1,04 Pop Bus
3 of 11 0,52 1522,11 0,00 1,33 Workers Bus Change
0,51 1524,02 0,00 150,46 Pop Age 2 Bus
0,51 1524,07 0,00 1,32 Age 2 Bus Change
4 of 11 0,56 1515,00 0,00 150,49 Pop Age 2 Bus Change
0,54 1518,24 0,00 7,65 Age 1 Age 2 Bus Change
0,54 1518,46 0,00 38,62 Pop Workers Bus Change
5 of 11 0,56 1514,84 0,00 152,03 Pop Income Age 2 Bus Change
0,56 1514,89 0,00 162,54 Pop Age 2 Buses Change Dist
0,56 1516,48 0,00 308,66 Pop Workers Age 2 Bus Change
Table 2 presents ArcGIS’s explanatory variable analysis in which the software analyses all
giving variables, in this case all candidate variables. The analysis presents each possible equation for a specific number of variables in these equations in function of the three highest adjusted R-Squared. It also provides a multicollinearity table in which it presents each candidate variable’s covariates. The table goes even forward in explicitly stating that a combination of variables was not possible due to perfect multicollinearity.
Table 3: Summary of Multicollinearity
Variable VIF Violations
Covariates
Pop 577,62 321 Workers (98,47), Age 1 (93,13), Age 2 (93,13), Age 3(93,13)
Worker 222,84 309 Age 2 (98,47), Age 3(98,47), Pop (98,47), Age 1 (54,96)
Inc 3,44 0
34
Age 1 46,83 201 Pop (93,13), Age 2 (77,10), Workers (54,96), Age 3 (38,93)
Age 2 770,17 312 Workers (98,47), Age 3 (93,13), Pop (93,13), Age 1 (77,10)
Age 3 31,80 279 Workers (98,47), Age 2 (93,13), Pop (93,13), Age 1 (38,93)
Road density
1,09 0
Buses 1,55 0
Terminal 1,14 0
Change 1,40 0
Distance 3,35 0
The following table presents the number and percentage of equations generated and passed according the presented criteria.
Table 4: Percentage of Search Criteria Passed
Search Criterion
Cutoff Trials # Passed % Passed
Min Adjusted R-Squared
> 0,50 1015 115 11,33
Max Coefficient p-value
< 0,05 1015 69 6,80
Max VIF Value < 7,50 1015 417 41,08
Moran’s I test was also done in order to evaluate autocorrelation with regards to ridership. The following table presents the outcomes as well as the estimated Moran’s I value.
Table 5: Moran’s I tests on the Dependent and Candidate Variables
Variable Moran's Index
Expected Index
Variance
z-score p-value Pattern
Ridership 0,019397 -0,010204
0,00308 0,533356 0,593787 Clustered
Population 0,513214 -0,010204
0,00463 7,69199 0 Clustered
35
Age 1 0,375009 -0,010204
0,005029 5,435004 0 Clustered
Age 2 0,548009 -0,010204
0,004924 7,957756 0 Clustered
Age 3 0,675734 -0,010204
0,003933 10,929235 0 Clustered
Workers 0,551187 -0,010204
0,004596 8,28109 0 Clustered
Med Inc 0,659357 -0,010204
0,004869 9,5952 0 Clustered
Road Density
0,268342 -0,010204
0,004724 4,052722 0,000051 Clustered
Buses 0,149382 -0,010204
0,004242 2,450214 0,014277 Clustered
Change -0,08105 -0,010204
0,004801 -1,022436 0,306575 Random
Terminal 0,009537 -0,010204
0,004629 0,290162 0,771692 Mixed
Distance 0,872311 -0,010204
0,005166 12,281491 0 Clustered
The best fit was determined to include the number of workers in the station’s proximity,
the type of station when it comes to changes and the number of bus lines in a 200-meter radius. However, the initial equation presented both the p-value and the t-value of the number of
workers insignificant. Plotting the ridership as a function of the number of workers, it was shown that T-Centralen, Slussen and Gullmarsplan were in all of them huge outliers and when out of the data, the worker variable is significant with an R-Squared of 0,02166 before removing them and 0,23722 after removing them. It was thus decided not to include these three stations in future studies both for OLS and GWR.
36
Figure 12: Ridership vs. Number of Workers
(R-Squared values show a better fit when T-centralen, Slussen and Gullmarsplan are taken out of the expression)
After taking these outliers out, both OLS and GWR analysis were executed again leading to the following residual maps.
y = 0,1758x + 1006,7R² = 0,0283
y = 0,2083x + 619,99R² = 0,236
0
2000
4000
6000
8000
10000
12000
14000
16000
0 2000 4000 6000 8000 10000 12000
Ride
rshi
p
WorkersRidership with Outliers Ridership without outliers
Linear (Ridership with Outliers) Linear (Ridership without outliers)
37
Figure 13: OLS (up) and GWR (down) Standard Deviations
38
The significance of the OLS parameters as well as for both equations are presented in the following tables.
Table 6: OLS Equation
Variables Coefficients Standard Error
t-Statistic Probability VIF
Intercept -552,050799 351,437275 -1,570837 0,119624 -
Workers 0,311537 0,054621 5,703612 0,000000 1,044079
Buses 88,258312 18,137573 4,866049 0,000005 1,327416
Change 1038,326730 283,930296 3,656978 0,000429 1,277995
Number of Observation
97
Number of Variables
4
Adj R-Squared
0,546342
Residual Square
33224590,808728
AICc 1631,798893
Table 7: GWR Equations
Variables Minimum Maximum Mean Standard Deviation
Intercept -2446,301797 657,6131415 -563,7553011 716,045689
Workers -0,004097741 0,258132412 0,149459243 0,050484
Buses -2,954762007 88,47400346 39,9531998 23,725989
Change -130,9206048 2793,393959 883,5952445 738,663661
Number of Variables
4
Number of Neighbors
42
39
Adj R-Squared
0,683548
Residual Square
16766876,5866
AICc 1500,09803 Looking at the maps, one can see different cluster in the OLS residual map that were reduced nay sometimes corrected in the GWR residual map. Looking closely at the value, one can safely say that the GWR has a higher predictability with an overall adjusted R-Squared of 0,68 compared to 0,55 for the normal OLS model. In addition, the residual square is almost half for the GWR model compared with the OLS one. It is important to specify that when the OLS equation was determined, ArcGIS launched a warning stating that the model should be checked for spatial autocorrelation, hinting that a spatial independent model might not be the best fit. This was proven with both Moran’s I test as well as the comparison between the OLS equation and GWR ones. The OLS equation was thus used in order to predict the ridership at the new stations for current situation. The following table presents the predicted ridership.
Table 8: Predicted Ridership using OLS Equation for New Station in 2016
Station Predicted Ridership
Nacka 2940
Järla 1053
Sickla 1118
Hammarby Kannal 1854
Sofia 2026
Barkarbystaden 265
Barkarby Station 608 The following map show the overall standard deviation of residual when GWR equations are calibrated for the new stations with this OLS estimated ridership.
40
Figure 14: GWR Standard Deviation with New Stations
The coefficients of the GWR equations as well as their significances are presented bellow.
Table 9: GWR Equations with the New Stations
Variables Minimum Maximum Mean Standard Deviation
Intercept -2367,47257 625,462321 -570,298174 670,481597
Workers 0,009195 0,24977 0,151743 0,043329
Buses 7,585861 86,22674 42,060432 19,308832
Change -120,334075 2667,703535 873,7549 688,076229
41
Number of Variables
4
Number of Neighbors
46
Adj R-Squared
0,69915
Residual Square
17569889,2664
AICc 1600,959011
Predictions Using the Determined GWR Equations
Having determined the GWR equations, future ridership can be thus determined. The future number of workers in the respective station’s service area should be thus determined. For the presented stations, the number of planned residences has been multiplied by the average number of persons per household. The average number of persons per household depends on the municipality and is respectively 2,12 and 2,47 for Stockholm and Nacka in 2018 (Statistiska centralbyrån (SCB), n.d.). Each new station’s equivalent station is also presented as well as the ratio Population to Workers. Finally, this ratio is multiplied to the number of people living in the area, leading to the additional number of workers living in the area. Adding these numbers to the current number of workers in the area gives the information needed to input into the equation.
Table 10: Number of Additional Population and Workers by 2030
(*Note: For Barkarbystaden and Barkarby Station, the total population and the number of workers is presented)
Station Number of residences
Household
Additional Population
Equivalent Station
Ratio Pop/Wor
Additional Workers
Nacka 6500 2,47 16055 Solna Centrum
0,540875 8684
Järla 950 2,47 2347 Västra Skogen 0,462653 1086
Sickla 2000 2,47 4940 Liljeholmen 0,555743 2746
Hammerby Kannal
2140 2,12 4537 Alvik 0,577865 2622
Sofia 155 2,12 329 Mariatorget 0,580132 191
42
Kungträd- gården
0 2,12 0 Hötorget 0,563254 0
Fridhems- plan
134 2,12 285 - 0,612255 174
Stadshagen
1975 2,12 4187 - 0,630363 2640
Rinkeby 1100 2,12 2332 - 0,294039 686
Tensta 1030 2,12 2184 - 0,347523 759
Kista 1600 2,12 3392 - 0,467844 1587
Akalla 1000 2,12 2120 - 0,443841 941
Barkarby- Staden*
- - 4946 Farsta Strand 0,462653 2288
Barkarby Station*
- - 1755 Farsta Strand 0,462653 811
Regarding Barkarbystaden and Barkarby Station, the future number of residences is already known by SAM. Adding the population to each service area by proceeding in the same manner done for the current situation, one can proceed in multiplying this population data by the ratio Population to Workers and getting the number of workers in the service area. Remaining stations on the blue line where no significant information has been found saw an increase of 5% of its worker’s population and this according to RUFS’ estimated increase of the population by 2030. T-Centralen’s service area remained untouched, however. Using the GWR equations, the ridership of the chosen blue line stations is predicted. The following table presents the GWR equation for each station as well as the final determined ridership. Having all the needed data and statistics, a deep analysis and discussion of these results is discussed in the following section.
43
Table 11: Coefficients and GWR Estimations for New Stations
Name Intercept Worker Coeff.
Bus Coeff.
Change Coeff.
Predicted by GWR
Predicted by SLL
Nacka -587,71361
32 0,13881581 36,517037
07 949,625133
7 4200
3000
Järla -514,98852
84 0,14376295
6 38,812961
65 844,623781
4 1169
600
Sickla -628,53274
11 0,15307304
8 36,370452
15 942,764456
8 1444
800
Hammarby Kanal
-1667,6785
14 0,17223955
6 21,962985
43 2007,36610
6 2298
1200
Sofia -1487,5451
22 0,16621622
7 18,791536
61 1860,82295
4 1632
1600
Barkarbystaden
-6,7022540
07 0,12151317
6 52,664246
18 282,178624
6 554
2300
Barkarby station
-28,110621
92 0,11704238
7 54,708347
43 306,278404
3 701
1200
After looking at the results, one can see that the ridership predicted for Barkarbystaden and Barkarby Station were very much under predicted, less than a quarter of the official prediction for Barkarbystaden. For this reason, the way the workers were determined for these two stations was reevaluated and calculated the same way it was done for other stations with information about the number of planned residences to be built by 2030. The following table shows the newly calculated additional workers.
44
Table 12: Number of Additional Population and Workers by 2030 for Barkarbystaden and Barkarby Station
Station Number of residences
Household
Additional Population
Equivalent Station
Ratio Pop/Wor
Additional Workers
Barkarbystaden
16593 2,42 40156 Farsta Strand 0,46265 18578
Barkarby Station
5078 2,42 12289 Farsta Strand 0,46265 5686
Recalculating the ridership for these two stations, the final ridership predictions for all stations new stations are as presented below.
Table 13: Coefficients and GWR Estimations for New Stations after the change in workers
Name Intercept Worker Coeff.
Bus Coeff.
Change Coeff.
Predicted by GWR
Predicted by City
Nacka -587,71361
32 0,13881581 36,517037
07 949,62513
37 4200
3000
Järla -514,98852
84 0,14376295
6 38,812961
65 844,62378
14 1169
600
Sickla -628,53274
11 0,15307304
8 36,370452
15 942,76445
68 1444
800
Hammarby Kanal
-1667,6785
14 0,17223955
6 21,962985
43 2007,3661
06 2298
1200
Sofia -1487,5451
22 0,16621622
7 18,791536
61 1860,8229
54 1632
1600
Barkarbystaden
-6,7022540
07 0,12151317
6 52,664246
18 282,17862
46 2566
2300
Barkarby station
-28,110621
92 0,11704238
7 54,708347
43 306,27840
43 1342
1200
45
Analysis Analyzing the predicted ridership for the new stations, one can see that they are, generally overestimated compared to official results, thus with regards to the Sampers model used in Vissum. This can be explained by multiple reasons both general, regional and station specific and the assumptions behind them.
Division between the North and the South
One can see a clear schism between the north stations (Barkarbystaden and Barkarby Station) and the south stations (Sofia, Hammerby Kannal, Sickla, Järla and Nacka). The first mentioned have ridership estimates that are close to the estimations made by SLL, even if slightly over estimated.
However, it is important to stress that the models of the northern stations has been developed with only one ridership that was predicted by the OLS model while in the southern station’s models, this number is definitely higher, especially when moving further away from the center of the study area towards Nacka.
The Model in Numbers
The model’s coefficients are very broad and different. Looking into the intercepts, stations like Sofia and Hammerby Kannal have values well
below the average of -570 with -1487,54 and -1667,67 while the stations in the north (Barkarbystaden and Barkarby Station).
Figure 15: The distribution of the intercept over the studied area (from the lowest in blue to the highest in red; this applies to all
distributions to follow)
46
Looking into the workers coefficients, these values seem to decrease the further one moves away from the city center towards the end of the lines. This means that, at least for the blue line, the further one goes from the city center, the less workers use the metro to get to their respective places of work.
Figure 16: The distribution of the worker’s coefficients over the studied area
Looking into the bus coefficient, south stations have values below the average of 42,06 with respectively from Sofia to Nacka, 18,79, 21,96, 36,37, 38,81 and 36,52. With regards to northern stations, they are slightly above average with respectively 52,66 and 54,70 for Barkarbystaden and Barkarby Station. The same pattern can be observed than the one observed for workers, yet inversely. In fact, the farther the station is compared to the central area the greater the bus coefficient influences the estimated ridership. This is quite common for metro network, especially star network like the one in Stockholm: in central area the dense metro stations presents does not present the need to use buses to get to a specific station while the further one goes from the center, the lower the density of metro stations, leading to the use of buses to reach these stations.
47
Figure 17: The distribution of the bus’s coefficients over the studied area
Looking into the change coefficient, the station in the north have coefficients that are below average while the south stations have coefficient above the 873-mean value. The same pattern can also be deduced. The change coefficient in stations in the central part of Stockholm increase the number of passengers in each specific station compared to stations outside of the city center.
Figure 18: The distribution of the change’s coefficients over the studied area
48
General Reasons
In general, this might be due to the use of the average number of persons per household. In fact, even if these new station with the exception of Sofia and Hammerby Kannal are outside of municipal Stockholm, they are relatively close to it. This means that the average number of persons per household in this area might be closer and thus lower than their average municipal one, i.e. Nacka and Järfalla.
The major change of the areas between 2016 and 2030 might affect the way the GWR
equations are predicted leading to a change in the calibration of the variables that are geographically dependent. The GWR equations were estimated for a spatial area that is different in 2030 than it is today, like Sickla, Järla and Nacka. Sofia are predicted relatively accurately due to the fact that the environment has not changed much.
Bus Assumptions and Effect on Predictions
Looking closer into the difference between the north and south stations in the first place might explain the difference seen between GWR values and the one by SLL: this is mainly due to the assumption taken with regards to the number of bus lines. In fact, the number of buses in the areas of interests were left as is. Although, in 2030, with new development and metro stations built, the current state will definitely evolve. On one hand, the stations in the north of the blue line see a slight overprediction of respectively 11,6% and 11,8% for Bakarbystaden and Barkarby Station compared to SLL estimations. This is mainly due to the fact that a limited number of bus lines were found around the future stations, none for Barkarbystaden and six for Barkarby Station. An increase in the number of residence and two new metro stations will definitely lead to new lines stopping in the area, especially in Barkarbystaden. On the other hand, the stations in the south and in municipal Nacka are overwhelmingly overestimated. In fact, the predicted ridership at Järla by the GWR model is almost twice as the ridership predicted by SLL. Here, a huge number of bus lines that pass in the region that lead to Nacka and beyond east will be removed by 2030. Today, these lines mainly start at Slussen. In 2030, it is planned that the lines that lead to Nacka will drastically decrease and the ones that leads farther east moved to Nacka’s bus terminal. This means that a lot of the buses accounted for when doing the predictions will be gone in the future and thus should not be accounted for in order to get a closer projected ridership to the SLL one.
Unfortunately, no data is available today on the number of bus lines that are planned to cover both areas.
Special Case of Sofia
Sofia is the only station that has an estimation that is almost identical to the one predicted by SLL.
49
A simple explanation might be the fact that the region barely changes with regards to the socio-economic dynamic as well as the transit characteristic with regards to 2016. In fact, hardly any residence will be built in the area, only 155, a thin increase of less than 200 workers. The area has already a high-service when it comes with bus stops and bus lines that pass through it. A change in the number of lines will thus be limited in the area.
Limitations of the Study Multiple limitations can be drawn in this study dipping a toe into the multiple defects of the model and the study itself.
First, land use variables are, according to preliminary studies an important component that can explain a part of ridership. However, accurate land use data was not available for the whole study area. Only municipal Stockholm had land use data with the required accuracy. For this reason, land use was disregarded when calculation both OLS and GWR equations. Lack of land use data and variable automatically puts limits for this study as it is a crucial information to incorporate in the study, especially when debating on a subject that has the change of land use for the future service areas of the new stations in its core.
Second, when service areas were drawn on ArcGIS, three main critics can be said. First, when Thiessen polygons were used for close stations, the latter assumes, at least in the central part of Stockholm that people always choose the closest station independently of the type of station. Behavioral studies show that this is not always the case. Second, problems with generating the service area on ArcGIS led to having some stations with either service areas way bigger or way smaller than 800 meters road length wise and the fix was thus done manually and might not be precis. Third, the network layer used in order to create the service areas had missing pedestrian tracks for some regions, meaning that some service areas might have been created with flawed road and pedestrian network especially at the outskirts of the metro network where the problem was considerable. Manual fixes were thus undertaken in order to resolve the problem.
Third, assumptions and hypotheses taken in order to be able to pursue the study can definitely be questioned.
The way the GWR equations for the new stations were determined is completely problematic. Determining ridership for stations that do not yet exist in a year they definitely will never do can never be observed. This hypothetical situation has led to error being included and dragged in the study that can never be identified.
The assumption about the population and workers being homogeneously distributed over the multiple SAMS is far from the situation on the ground and can lead to inaccuracy and misrepresentation, especially in relatively large zones (which are mainly clustered in the outskirts of the metro network) where irregularities can be more pronounced. As already discussed above, the unchanged bus network between 2016 and 2030 is a huge drawback for accurate prediction at the new stations.
Finally, the choice of the number of persons per household and of the equivalent stations while calculating the number of workers in service areas is unrealistic. First, the exact number of persons per household cannot be known and the value that was used was for the current state of 2018. Second, to assume that the ratio of workers to population in new stations is identical to their
50
chosen current equivalent station cannot be proven yet and is totally hypothetical done for the purpose of easing the calculation of the potential number of workers in the service areas.
The ridership predicted by SLL was assumed to be correct and unfaulty, However , this
predictions cannot be observed yet. Comparing the GWR predictions with other predictions cannot prove for sure how GWR performs with regards to the situation on the ground.
In future studies, these limits should be considered and tackled in order to improve the
potential predictiveness of the model.
51
Conclusion
After going through the analysis of these results, one case-specific conclusion and two
main conclusions were deduced from this study.
In this case, GWR equations overpredicted the ridership for future stations with regards to the official predictions. However, a generalization of this results cannot be drawn due, first, the many case-specific assumptions taken here and, second, the lack of other studies available today using the same process.
The general conclusion are as follows. First, GWR is a good predicting tool for future
stations that are closely located from most present stations. Second, GWR is a good predicting method for stations where limited changes in the future environment will occur.
These conclusions help in realizing when to use GWR. In fact, according to these results
the conclusion determined, Geographically Weighted Regression can be used as both an explanatory and predictive tool but in limited cases. With regards to its explanatory power, the model can help explaining in details and area specific the current situation or the possibility of building one or more stations for a given area which will not change drastically. This can also be taken further in determining the ridership using the GWR equation.
GWR is thus performs good for a limited number of stations built next to an existing, stable
environment. For example, densely populated cities with an existing but limited public transportation or metro network are ideal cases for the use of this method. In addition, developed countries with limited funds are good situation where an area/station-specific equation explains and predicts the ridership reassuring policymakers and investors. The determined parameters can also help determine area-specific and thus more accurate and precise policies that can be proposed. For instance, the choice of building new apartments in specific areas can also be analyzed and other areas, not considered, can be proposed.
However, this study is still preliminary and experimental in the field of the
predictive power of GWR with regards to new metro stations. More studies should be performed for different contexts, tackling the limitations in order to get more accurate and significant results.
52
References American Public Transportation Association. (2019). Public Transportation Facts - American Public Transportation
Association. [online] Available at: https://www.apta.com/news-publications/public-transportation-facts/ [Accessed 5 Sep. 2019]. Anselin, L. (1999). Interactive techniques and exploratory spatial data analysis. Longley, P., Goodchild, M., Maguire, D., Rhind, D. (Eds.), Geographical Information Systems, second ed. John Wiley, pp.252–264. Bernetti, G., Longo, G., Tomasella, L. and Violin, A. (2008). Sociodemographic Groups and Mode Choice in a Middle-Sized European City. Transportation Research Record: Journal of the Transportation Research Board, 2067(1), pp.17-25. Bhat, C. and Gossen, R. (2004). A mixed multinomial logit model analysis of weekend recreational episode type choice. Transportation Research Part B: Methodological, 38(9), pp.767-787. Blainey, S. and Mulley, C. (2013). Using Geographically Weighted Regression to forecast rail demand in the Sydney Region. Australasian Transport Research Forum 2013. Brisbane. Boyle, D. (2006). Fixed-route Transit Ridership Forecasting and Service Planning Methods. Transportation Research Board Reports, 66. Cardozo, O., García-Palomares, J. and Gutiérrez, J. (2012). Application of geographically weighted regression to the direct forecasting of transit ridership at station-level. Applied Geography, 34, pp.548-558. Cervero, R. (2006). Alternative Approaches to Modeling the Travel-Demand Impacts of Smart Growth. Journal of the American Planning Association, 72(3), pp.285-295. Chambers, R. and Clark, R. (2019). An introduction to model-based survey sampling with applications. Oxford University Press. Chiou, Y., Jou, R. and Yang, C. (2015). Factors affecting public transportation usage rate: Geographically weighted regression. Transportation Research Part A: Policy and Practice, 78, pp.161-177. Chow, L., Zhao, F., Liu, X., Li, M. and Ubaka, I. (2006). Transit Ridership Model Based on Geographically Weighted Regression. Transportation Research Record: Journal of the Transportation Research Board, 1972(1), pp.105-114. Cristaldi, F. (2005). Commuting and Gender in Italy: A Methodological Issue. The Professional Geographer, 57(2), pp.268-284. Data.worldbank.org. (2019). Urban population growth (annual %) | Data. [online] Available at: https://data.worldbank.org/indicator/SP.URB.GROW [Accessed 5 Sep. 2019]. De Smith, M., Goodchild, M. and Longley, P. (n.d.). Geospatial analysis. A comprehensive guide to principles, techniques and software tools. Dimopoulos, T. and Moulas, A. (2016). A Proposal of a Mass Appraisal System in Greece with CAMA System: Evaluating GWR and MRA techniques in Thessaloniki Municipality. Open Geosciences, 8(1). Floch, J. (2016). Prendre en compte l’hétérogénéité spatiale pour calculer des estimateurs : l’apport de la geographically weighted regression.
53
Fotheringham, A., Charlton, M. and Brunsdon, C. (1998). Geographically Weighted Regression: A Natural Evolution of the Expansion Method for Spatial Data Analysis. Environment and Planning A: Economy and Space, 30(11), pp.1905-1927. Fotheringham, A., Brunsdon, C. and Charlton, M. (2000). Quantitative geography. Los Angeles: Sage Publications. Fotheringham, A., Brunsdon, C. and Charlton, M. (2002). Geographically weighted regression. Chichester: Wiley. Geodata.se. (2019). Startsida. [online] Available at: https://www.geodata.se [Accessed 20 May 2019]. Gollini, I., Lu, B., Charlton, M., Brunsdon, C. and Harris, P. (2015). GWmodel: AnRPackage for Exploring Spatial Heterogeneity Using Geographically Weighted Models. Journal of Statistical Software, 63(17). Gómez-Ibáñez, J. (1996). Big-City Transit Rider snip, Deficits, and Politics: Avoiding Reality in Boston. Journal of the American Planning Association, 62(1), pp.30-50. Gutiérrez, J., Cardozo, O. and García-Palomares, J. (2011). Transit ridership forecasting at station level: an approach based on distance-decay weighted regression. Journal of Transport Geography, 19(6), pp.1081-1092. Hadayeghi, A., Shalaby, A. and Persaud, B. (2010). Development of planning level transportation safety tools using Geographically Weighted Poisson Regression. Accident Analysis & Prevention, 42(2), pp.676-688. Harders, C. and Björkman, J. (2016). Prognos över resandeutveckling. Stockholms läns landsting. Harris, P., Fotheringham, A., Crespo, R. and Charlton, M. (2010). The Use of Geographically Weighted Regression for Spatial Prediction: An Evaluation of Models Using Simulated Data Sets. Mathematical Geosciences, 42(6), pp.657-680. Hildebrand, E. (2003). Dimensions in elderly travel behaviour: A simplified activity-based model using lifestyle clusters. Transportation, 30(3), pp.285-306. Horowitz, A. (1984). Simplifications for single-route transit-ridership forecasting models. Transportation, 12(3), pp.261-275. Hsiao, S., Lu, J., Sterling, J. and Weatherford, M. (1997). Use of Geographic Information System for Analysis of Transit Pedestrian Access. Transportation Research Record: Journal of the Transportation Research Board, 1604(1), pp.50-59. Institut national de la statistique et des études économiques (2018). Handbook of Spatial Analysis. Montrouge: Jean-Luc Tavernier, pp.231-254. Kitamura, R. (1989). A causal analysis of car ownership and transit use. Transportation, 16(2), pp.155-173. Kuby, M., Barranda, A. and Upchurch, C. (2004). Factors influencing light-rail station boardings in the United States. Transportation Research Part A: Policy and Practice, 38(3), pp.223-247. Lantmateriet.se. (2019). Lantmäteriet – vi känner till varenda plats i Sverige.. [online] Available at: https://www.lantmateriet.se [Accessed 22 May 2019]. Lawton, K. (1997). Metro 1994 Travel Behavior Survey. Portland: Portland Metro. Leung, Y., Mei, C. and Zhang, W. (2000). Testing for Spatial Autocorrelation among the Residuals of the Geographically Weighted Regression. Environment and Planning A: Economy and Space, 32(5), pp.871-890. Litman, T. (2004). Transit Price Elasticities and Cross - Elasticities. Journal of Public Transportation, 7(2), pp.37-58.
54
Liu, Y., Ji, Y., Shi, Z. and Gao, L. (2018). The Influence of the Built Environment on School Children’s Metro Ridership: An Exploration Using Geographically Weighted Poisson Regression Models. Sustainability, 10(12), p.4684. Lloyd, C. (2010). Local models for spatial analysis. Lloyd, C. and Shuttleworth, I. (2005). Analysing Commuting Using Local Regression Techniques: Scale, Sensitivity, and Geographical Patterning. Environment and Planning A: Economy and Space, 37(1), pp.81-103. Luo, J. and Wei, Y. (2009). Modeling spatial variations of urban growth patterns in Chinese cities: The case of Nanjing. Landscape and Urban Planning, 91(2), pp.51-64. Marshall, N. and Grady, B. (2006). Sketch Transit Modeling Based on 2000 Census Data. Transportation Research Record: Journal of the Transportation Research Board, 1986(1), pp.182-189. McNally, M. (2007). The four step model. Handbook of transport modeling, pp. 35-52. MTR (2018). Hållbarhetsredovisning 2017. Hållbarhetsredovisning. Murray, A. (2001). Strategic analysis of public transport coverage. Socio-Economic Planning Sciences, 35(3), pp.175-188. Nylén, A. (2017). Prognos över resandeutveckling Akalla Barkarby. Stockholm: Stockholms läns landsting. O’Neill, W., Ramsey, R. and Chou, J. (1992). Analysis of transit service areas using geographic information systems. Transportation Research Record, (1364), pp.131-138. O'Sullivan, S. and Morrall, J. (1996). Walking Distances to and from Light-Rail Transit Stations. Transportation Research Record: Journal of the Transportation Research Board, 1538(1), pp.19-26. Parker, T., McKeever, M., Arrington, G. and Smith-Heimer, J. (2002). Statewide Transit-Oriented Development Study – Factors for Success in California. Sacramento: Business, Transportation and Housing Agency and California Department of Transportation. Parsons Brinckerhoff, 1996. Transit and Urban Form, TCRP Report 16, vol. 1. Transportation Research Board, National Research Council, Washington, DC. Pineda Jaimes, N., Bosque Sendra, J., Gómez Delgado, M. and Franco Plata, R. (2010). Exploring the driving forces behind deforestation in the state of Mexico (Mexico) using geographically weighted regression. Applied Geography, 30(4), pp.576-591. Pushkarev, B.S., Zupan, J.M., 1982. Where transit works: urban densities for public transportation. Levinson, H.S., Weant, R.A. (Eds.), Urban Transportation: Perspectives and Prospects. Eno Foundation, Westport, CT. Qian, X. and Ukkusuri, S. (2015). Spatial variation of the urban taxi ridership using GPS data. Applied Geography, 59, pp.31-42. Regional utvecklingsplan för Stockholmsregionen (2010). RUFS 2010. Stockholm. Rosenberg, M. (2010). The Bearing Correlogram: A New Method of Analyzing Directional Spatial Autocorrelation. Geographical Analysis, 32(3), pp.267-278. Sallis, J. (2008). Angels in the details: Comment on “The relationship between destination proximity, destination mix and physical activity behaviors”. Preventive Medicine, 46(1), pp.6-7.
55
Savage, M. (2016). The city with 20-year waiting lists for rental homes. [online] Bbc.com. Available at: https://www.bbc.com/worklife/article/20160517-this-is-one-city-where-youll-never-find-a-home [Accessed 19 Apr. 2019]. Somenahalli, S. (2011). Stop-level Urban Transit Ridership Forecasting – A case Study. Journal of the Eastern Asia Society for Transportation Studies, 9, pp.422-436. Statistiska centralbyrån (SCB). (n.d.). Antal personer och hushåll samt antal personer per hushåll efter region.. [online] Available at: http://www.statistikdatabasen.scb.se/pxweb/sv/ssd/START__BE__BE0101__BE0101S/HushallT09/table/tableViewLayout1/ [Accessed 13 Aug. 2019]. Stockholms läns landsting (2016). Everything you need to know about Stockholm’s new Metro. Stockholm. Trafikverket. (2019). Nationell vägdatabas, NVDB. [online] Available at: https://www.trafikverket.se/tjanster/system-och-verktyg/data/Nationell-vagdatabas/ [Accessed 1 May 2019]. Trafikverket. (2018). Sampers. [online] Available at: https://www.trafikverket.se/tjanster/system-och-verktyg/Prognos--och-analysverktyg/Sampers/ [Accessed 12 Apr. 2019]. Trafikförvaltningen (n.d.). Procurement of public transport in Stockholm. Stockholm: Stockholms Läns Landsting. TU, J. and XIA, Z. (2008). Examining spatially varying relationships between land use and water quality using geographically weighted regression I: Model design and evaluation. Science of The Total Environment, 407(1), pp.358-378. Vaxer.stockholm. (2018). Omvandling av industribyggnader i Hammarby Sjöstad - Stockholm växer. [online] Available at: https://vaxer.stockholm/projekt/omvandling-av-industribyggnader-i-hammarby-sjostad/ [Accessed 8 Jul. 2019]. Vaxer.stockholm. (n.d.). Start - Stockholm växer. [online] Available at: https://vaxer.stockholm [Accessed 8 Jul. 2019]. Wachs, M. (1989). U.S. Transit Subsidy Policy: In Need of Reform. Science, 244(4912), pp.1545-1549. Wardman, M. (2004). Public transport values of time. Transport Policy, 11(4), pp.363-377. Whitehead, C. and Button, K. (1977). Urban Economics: Theory and Policy. Economica, 44(175), p.313. WSP Analys & Strategi (2013). Effekter på värdet på handelsfastigheter vid etablering av nya tunnelbanelinjer i Stockholmsregionen. Stockholm: WSP. Zhang, D. and Wang, X. (2014). Transit ridership estimation with network Kriging: a case study of Second Avenue Subway, NYC. Journal of Transport Geography, 41, pp.107-115. Zhang, P., Wong, D., So, B. and Lin, H. (2012). An exploratory spatial analysis of western medical services in Republican Beijing. Applied Geography, 32(2), pp.556-565. Zhao, F., Chow, L., Li, M. and Liu, X. (2005). A Transit Ridership Model Based on Geographically Weighted Regression and Service Quality Variables. Miami: Lehman Center for Transportation Research. Zhao, F., Chow, L., Li, M., Ubaka, I. and Gan, A. (2003). Forecasting Transit Walk Accessibility: Regression Model Alternative to Buffer Method. Transportation Research Record: Journal of the Transportation Research Board, 1835(1), pp.34-41.
56
Appendix
Appendix I: Method and Tools in Determining Data in ArcGIS (ArcMap)
Income, Workers, Population and Age
All of the above was done by a method of intersect and dissolve like in NYC. The service area was divided on the given SAM’s areas. The proportional population (with the respective age groups) and workers were then added to each division that fell within each respective SAM’s area. It was assumed that the population and the number of workers were homogeneously distributed on the SAM’s area and thus on the division of the service area that fell in each SAM area.
A step further was taken in calculating the median income. A weighted average in function of the number of workers was done. The following equation explains how the median income was determined for each service area.
∑=çqéLqt_ê!_ë:5:t:çH ∗ ÜLJ]Hy∑=çqéLqt_ê!_ë:5:t:çH
Where the Workers_by_Division was calculated as explained above.
Road density (m/m2) The layer was intersected using the tool “Intersect” then the summation of the length of roads was done in the area, the all divided by the area of the service area.
Number of bus lines at a 200-meter buffer around the entrances of the metro
The number of buses located at a 200-meter radius was considered. The buffer tool was used. Some adjustments were done manually in order to correct any mistakes. The number of bus lines were finally divided by 2 in order to account for only the feeder type of the buses in the morning leading to the station.
Terminal Station
Terminal station were added regarding their respective years. In 2016, Akalla, Hjulsta and Kungsträdgården on the blue line, Hässelby Strand, Hagsätra, Farsta Strand and Skarpnäck on the green line and Mörby Centrum, Ropsten, Norsborg and Fruängen on the red line were assigned the dummy value of 2 while all other stations were assigned 1. In 2030, Kungsträdgården is replaced by Nacka and Akalla by Barkarby. All other stations remain the same.
57
Type of change
The type of station was just translated to a dummy variable using the “bytespunkt” column from SLL of future stations.
Commuting distance
The point to distance tool was used to determine the distance from a station to T-centralen. Euclidian distances were used as the Stockholm metro network has a star shape and limited curves in its network.
58
Appendix II: Table of GWR Equations and Predictions for the 2016 Situation
Stat
ion
Obs
erve
d
Loca
l R2
Pred
icte
d
Inte
rcep
t
Wor
kers
Bus
Chan
ge
Resid
ual
Std
Erro
r
Std
Err_
Int
Std
Err
Wor
kers
St
d Er
r Bus
Std
Err
Chan
ge
StdR
esid
Äng
bypl
an
400,
0000
0000
0000
000
0,74
5808
5427
3200
0 59
5,87
4423
8949
9995
0 - 57
,098
7578
9659
9999
- 0,
0040
9774
1460
700
71,1
2596
8780
4000
01
514,
8025
9472
6000
050
- 195,
8744
2389
5000
010
440,
9814
5246
6000
010
381,
0086
0604
3999
980
0,08
2431
8699
9290
0 27
,907
9925
6609
9999
31
9,18
3645
8380
0002
0 - 0,
4441
7837
2581
000
Aka
lla
800,
0000
0000
0000
000
0,67
1086
8392
7900
0 93
3,80
5454
0240
0003
0 44
,963
6597
0650
0003
0,
1150
8926
5256
000
51,5
2738
5300
7999
99
249,
9766
1361
1999
990
- 133,
8054
5402
4000
000
390,
0152
8203
1000
030
302,
6912
2643
2000
010
0,05
9364
5491
7470
0 11
,278
7089
5919
9999
22
9,47
4690
8640
0000
0 - 0,
3430
7746
4369
000
Fittj
a
600,
0000
0000
0000
000
0,06
7274
1394
5750
0 91
3,86
1968
7460
0000
0 59
5,27
1498
8180
0005
0 0,
1173
6388
3033
000
21,4
9569
4744
4000
01
- 105,
8248
0043
0000
000
- 313,
8619
6874
6000
000
426,
6336
1665
2000
000
415,
9698
4168
0999
990
0,13
9072
9690
1400
0 36
,149
3607
5380
0003
30
4,26
7640
4229
9999
0 - 0,
7356
7097
5037
000
Brom
map
lan
3000
,000
0000
0000
0000
0,
6896
8980
3382
000
2594
,867
6502
7999
9900
- 48
6,80
5166
8219
9999
0 0,
0969
5239
9292
300
54,6
5231
8977
1000
02
831,
2969
5161
8999
970
405,
1323
4971
8000
000
166,
4111
4393
5000
010
382,
8379
8985
4999
990
0,06
1248
8680
1600
0 28
,419
4697
4320
0001
35
0,35
5978
8910
0001
0 2,
4345
2655
9560
000
Lilje
holm
en
4200
,000
0000
0000
0000
0,
8292
7186
9873
000
3755
,908
7095
8000
0000
- 22
79,9
4394
0149
9999
00
0,15
5161
9448
8700
0 37
,011
1633
7999
9999
26
42,3
2660
0120
0000
00
444,
0912
9042
4000
020
203,
5263
5604
2000
000
363,
3476
5862
9000
020
0,05
0474
9233
0690
0 27
,175
9831
9360
0000
34
2,91
8497
3170
0000
0 2,
1819
8418
6520
000
59
Örn
sber
g
900,
0000
0000
0000
000
0,75
0006
1087
9600
0 80
4,33
3990
0229
9996
0 - 14
35,5
2524
8859
9999
00
0,17
2199
0080
1700
0 73
,274
2496
7349
9995
16
73,4
8241
4460
0000
00
95,6
6600
9977
2000
05
438,
8380
2864
2999
970
384,
6743
9868
3999
980
0,05
8211
1461
4880
0 30
,089
8171
4140
0001
31
9,91
4168
8880
0001
0 0,
2179
9844
9845
000
Kär
rtorp
900,
0000
0000
0000
000
0,49
6910
9286
1900
0 65
2,07
0832
7229
9997
0 - 78
,980
2940
9580
0005
0,
1510
3419
2061
000
34,4
9912
6787
5999
98
396,
4958
0479
9999
970
247,
9291
6727
7000
000
450,
8317
2477
8000
020
367,
2393
1636
8000
000
0,06
1071
0451
8290
0 26
,322
7332
7100
0001
28
7,48
6143
4050
0001
0 0,
5499
3726
8499
000
Stur
eby
500,
0000
0000
0000
000
0,43
8176
8446
7900
0 66
1,16
4172
2509
9997
0 - 24
1,95
6260
9360
0001
0 0,
1445
6254
8257
000
59,0
2243
5228
5000
01
503,
3231
5617
6000
000
- 161,
1641
7225
1000
000
449,
2946
1506
4000
030
346,
2971
2979
2000
020
0,06
5277
5087
6440
0 28
,539
7469
6699
9999
27
5,96
5378
8439
9999
0 - 0,
3587
0488
2826
000
Glo
ben
300,
0000
0000
0000
000
0,60
2723
7730
7900
0 55
0,35
2684
1719
9995
0 - 10
93,4
9460
7679
9999
00
0,15
6560
1973
7600
0 40
,755
0528
0210
0003
13
45,9
2159
1990
0000
00
- 250,
3526
8417
2000
010
458,
2904
1066
5000
020
359,
4516
8974
0000
010
0,06
2474
3132
5000
0 28
,705
3812
9879
9999
28
5,98
7991
0729
9999
0 - 0,
5462
7519
6569
000
Huv
udsta
600,
0000
0000
0000
000
0,63
1011
1965
7000
0 92
6,23
8926
6249
9998
0 - 77
0,26
1623
8099
9995
0 0,
1936
6421
3989
000
30,6
8981
6188
1999
98
928,
3731
7775
0999
960
- 326,
2389
2662
4999
980
448,
4876
0379
9000
000
328,
1205
7522
7000
020
0,03
9174
7078
1510
0 20
,981
5975
9949
9999
29
6,87
3984
7019
9997
0 - 0,
7274
2016
4706
000
Uni
vers
itete
t
500,
0000
0000
0000
000
0,59
7904
8678
4400
0 82
2,45
7341
4369
9997
0 - 75
4,89
3479
6590
0004
0 0,
1456
0841
2684
000
33,0
7125
7853
2000
02
1061
,770
4260
0000
0000
- 32
2,45
7341
4370
0003
0 42
9,22
3630
4269
9999
0 28
2,02
0067
7340
0001
0 0,
0342
2284
9464
000
11,9
6863
9394
2000
00
257,
3876
2027
0000
010
- 0,75
1257
1968
9700
0
Ham
mar
byhö
jde
n 90
0,00
0000
0000
0000
0 0,
5605
2280
1223
000
988,
6445
0089
1000
010
- 558,
0317
1047
1000
000
0,15
8867
3267
6700
0 32
,349
9297
5929
9997
84
9,76
1543
5879
9999
0 - 88
,644
5008
9049
9998
46
0,33
7759
2919
9999
0 36
9,98
9969
1730
0002
0 0,
0579
7470
5077
500
26,7
3637
8512
0000
02
292,
3979
0117
4000
030
- 0,19
2564
0447
7200
0
60
Joha
nnel
und
200,
0000
0000
0000
000
0,64
0398
6013
4900
0 49
9,50
3251
4120
0000
0 67
,535
0184
4340
0006
0,
0621
5205
2741
000
71,5
7583
5008
9000
00
254,
3221
7367
9999
990
- 299,
5032
5141
2000
000
431,
6275
9706
0000
030
372,
7506
0285
0000
010
0,08
5770
0649
7600
0 25
,259
0797
9370
0000
30
2,57
1291
7180
0000
0 - 0,
6938
9272
9410
000
Hag
sätra
1200
,000
0000
0000
0000
0,
3862
2809
6510
000
676,
4634
4108
4000
010
- 112,
0135
4510
6000
000
0,24
5129
7689
3300
0 66
,796
5999
2480
0006
24
7,35
6807
5990
0001
0 52
3,53
6558
9159
9999
0 44
0,89
9094
6920
0001
0 37
7,71
3177
6389
9998
0 0,
0879
7411
8596
600
33,1
7154
7694
2000
01
301,
2679
3064
8999
990
1,18
7429
4259
6000
0
Mäl
arhö
jden
700,
0000
0000
0000
000
0,59
1631
5791
4000
0 84
2,25
4058
6970
0000
0 - 75
6,55
7067
2400
0004
0 0,
2223
8022
7498
000
61,5
0530
5715
6999
97
975,
1882
7678
2999
990
- 142,
2540
5869
7000
000
450,
1293
6263
0000
000
398,
5358
2521
2000
020
0,06
0560
8743
5650
0 26
,617
5537
5150
0001
34
0,10
5192
4720
0000
0 - 0,
3160
2928
0707
000
Isla
ndsto
rget
800,
0000
0000
0000
000
0,73
7669
6804
9100
0 65
1,84
3732
7159
9998
0 - 9,
3220
0403
0680
001
0,00
4697
7212
5907
0 75
,088
3297
7170
0003
43
0,01
9200
4150
0000
0 14
8,15
6267
2839
9999
0 45
2,24
9941
5880
0001
0 38
0,76
0145
6049
9999
0 0,
0851
5253
3152
700
27,4
6532
0656
3000
01
320,
5244
6289
0999
980
0,32
7598
2010
4000
0
Berg
sham
ra
1100
,000
0000
0000
0000
0,
6004
3372
3983
000
881,
9106
3576
3999
950
- 451,
5865
0102
1999
990
0,15
7550
4087
9300
0 46
,272
0497
5779
9998
62
6,46
3725
5960
0002
0 21
8,08
9364
2359
9999
0 44
0,65
8763
2839
9997
0 29
4,61
0310
5440
0001
0 0,
0350
2860
9122
800
12,0
0177
2224
3000
00
244,
6588
2296
1999
990
0,49
4916
6620
7000
0
Skan
stull
2200
,000
0000
0000
0000
0,
8114
3694
1582
000
1765
,562
6392
7000
0100
- 24
46,3
0179
7000
0001
00
0,18
3633
7602
4000
0 7,
9875
0168
4050
000
2793
,393
9585
8999
9800
43
4,43
7360
7279
9999
0 39
1,82
7655
6060
0001
0 37
4,22
2224
1540
0000
0 0,
0473
7066
5587
000
25,1
4669
1876
9999
99
382,
6324
4440
8000
030
1,10
8746
0379
9000
0
Hök
arän
gen
1000
,000
0000
0000
0000
0,
3892
8858
3502
000
1029
,237
1728
9999
9900
80
,839
6649
9000
0003
0,
1636
2916
9078
000
40,7
4980
3809
2999
98
228,
7396
4856
6999
990
- 29,2
3717
2903
0000
01
390,
8668
4041
1000
000
343,
4460
4783
0999
990
0,06
7499
2722
1130
0 30
,265
2233
7910
0000
27
1,53
7452
5249
9999
0 - 0,
0748
0085
2567
300
61
Mas
mo
200,
0000
0000
0000
000
0,08
3483
1172
4680
0 50
9,47
6224
3810
0001
0 54
1,50
9618
9930
0000
0 0,
1531
7899
3087
000
17,4
5271
8908
2000
01
- 70,9
0715
0568
0999
93
- 309,
4762
2438
1000
010
273,
3393
8040
4999
990
410,
8456
5528
0000
020
0,13
3257
7566
6200
0 36
,835
2035
1939
9997
29
9,15
3271
2210
0001
0 - 1,
1322
0504
0940
000
Frid
hem
spla
n
5700
,000
0000
0000
0000
0,
6663
4433
4885
000
4208
,887
3744
9000
0200
- 13
50,1
6061
6590
0000
00
0,19
4762
8960
8300
0 5,
0839
5325
0760
000
1677
,082
3235
8000
0100
14
91,1
1262
5510
0000
00
346,
5620
0091
6999
980
305,
5899
1338
7000
020
0,03
6604
3898
3010
0 19
,167
4191
2159
9998
27
1,90
6498
3530
0002
0 4,
3025
8545
8200
000
Väs
terto
rp
700,
0000
0000
0000
000
0,49
1330
4378
2300
0 10
59,1
2492
4700
0001
00
- 562,
7378
7546
7999
970
0,21
5492
4289
5000
0 68
,419
0258
5459
9994
80
0,83
5927
2190
0004
0 - 35
9,12
4924
7010
0000
0 45
2,74
5656
5449
9997
0 45
0,73
0805
5199
9999
0 0,
0798
7705
7606
200
32,1
5520
4800
1000
02
371,
0645
0334
4000
000
- 0,79
3215
6156
7600
0
S:t E
riksp
lan
2500
,000
0000
0000
0000
0,
6433
2856
4656
000
1932
,574
6979
2000
0100
- 11
57,4
3066
3079
9999
00
0,19
0841
4393
3600
0 7,
0729
2396
0580
000
1477
,790
8623
5000
0000
56
7,42
5302
0840
0001
0 38
8,09
7128
3090
0003
0 29
4,84
3901
9809
9999
0 0,
0351
0352
6465
900
18,8
2842
1751
4000
01
265,
6807
3801
8999
990
1,46
2070
3444
9000
0
Med
borg
arpl
atse
n 14
00,0
0000
0000
0000
00
0,73
3958
9552
7200
0 15
14,6
1706
4780
0000
00
- 1853
,530
8792
5999
9900
0,
1802
9045
6263
000
0,56
6547
3646
7300
0 22
64,7
4579
2679
9998
00
- 114,
6170
6477
6000
010
417,
5447
7483
4000
010
329,
7309
8872
8999
990
0,04
0793
8139
2130
0 22
,230
3613
6800
0001
34
8,40
8574
1330
0000
0 - 0,
2745
0245
2633
000
Tallk
roge
n
400,
0000
0000
0000
000
0,42
9981
0130
6300
0 65
2,34
0974
5510
0004
0 - 2,
4700
6563
3230
000
0,15
1186
4304
6000
0 43
,373
4863
6230
0000
30
1,82
4763
4190
0002
0 - 25
2,34
0974
5509
9999
0 45
6,27
6065
2189
9997
0 34
9,30
9235
6899
9998
0 0,
0661
7163
6248
100
28,8
6297
7895
8999
99
274,
4862
7283
5000
020
- 0,55
3044
5135
9000
0
Gam
la st
an
800,
0000
0000
0000
000
0,68
2681
8835
0500
0 78
9,45
7852
5050
0001
0 - 15
39,9
3719
6180
0000
00
0,17
3117
0632
9700
0 - 1,
1926
6970
3250
000
1991
,201
6753
7000
0000
10
,542
1474
9540
0000
44
6,78
6228
7739
9999
0 32
5,11
0693
3899
9999
0 0,
0376
2659
6004
900
20,7
8133
2843
3000
00
325,
5149
2999
3000
010
0,02
3595
5067
9150
0
62
Kar
lapl
an
900,
0000
0000
0000
000
0,64
2292
4919
5600
0 88
3,79
9495
0779
9995
0 - 13
17,6
4404
6570
0000
00
0,13
3094
5795
6000
0 - 2,
9547
6200
6600
000
1929
,611
6710
7000
0100
16
,200
5049
2200
0000
45
5,63
6586
2870
0000
0 32
9,90
5841
5939
9998
0 0,
0357
1895
7838
800
20,8
3672
9038
3000
00
341,
7842
4538
5000
020
0,03
5555
7595
8910
0
Ensk
ede
Går
d
300,
0000
0000
0000
000
0,55
6628
3665
9000
0 49
1,52
3876
5060
0002
0 - 83
1,98
0303
4980
0001
0 0,
1485
0654
6788
000
48,5
4254
7944
1000
01
1078
,952
0831
4999
9900
- 19
1,52
3876
5059
9999
0 45
6,17
6692
5680
0002
0 35
5,42
9116
2409
9997
0 0,
0635
6034
4701
200
29,0
1608
1353
2999
99
283,
4826
3484
4000
020
- 0,41
9845
8176
9700
0
Soln
a str
and
300,
0000
0000
0000
000
0,60
6490
6666
5900
0 64
6,57
6918
0830
0001
0 - 51
7,27
8313
8139
9994
0 0,
1643
9450
1222
000
45,9
9407
9434
5000
01
675,
6210
7330
5999
970
- 346,
5769
1808
3000
010
444,
0319
1135
4999
980
331,
6748
1975
4000
000
0,04
1360
2214
7470
0 19
,916
8004
5909
9999
28
1,84
5395
3179
9999
0 - 0,
7805
2254
6287
000
Rågs
ved
800,
0000
0000
0000
000
0,37
2720
4580
1400
0 68
3,73
6196
4169
9997
0 - 55
,356
2952
6470
0002
0,
2141
6002
3184
000
55,3
5556
7651
1999
98
264,
2166
3197
9999
990
116,
2638
0358
3000
000
429,
3414
9354
8000
020
355,
0990
0310
1999
980
0,07
5373
2544
2780
0 31
,200
0023
7620
0000
28
0,59
4341
2390
0002
0 0,
2707
9563
7809
000
Hal
lonb
erge
n
1000
,000
0000
0000
0000
0,
6191
6785
9742
000
905,
0584
7614
8999
940
- 155,
1842
0301
0000
000
0,14
2157
6634
9400
0 50
,761
209 8
1630
0003
35
7,60
1190
6310
0001
0 94
,941
5238
5079
9996
44
4,64
4807
2019
9998
0 28
7,11
1511
0200
0002
0 0,
0417
8824
2271
300
10,8
9299
9018
2999
99
225,
3909
3778
3000
000
0,21
3522
1694
1000
0
Skär
mar
brin
k
1200
,000
0000
0000
0000
0,
6064
1997
1857
000
487,
9943
8567
8000
010
- 1022
,411
7719
0000
0000
0,
1624
1728
6180
000
35,2
1323
2194
6999
98
1283
,702
9449
7000
0100
71
2,00
5614
3220
0004
0 44
5,59
1825
0180
0001
0 34
7,34
7431
0400
0000
0 0,
0572
4866
1050
800
26,7
4310
9434
6000
01
272,
9692
5251
4999
990
1,59
7887
5157
6000
0
Hju
lsta
300,
0000
0000
0000
000
0,62
7908
1661
0200
0 68
4,64
7146
7320
0001
0 14
,031
0835
3280
0000
0,
1045
0301
4549
000
53,9
6024
9604
9000
00
293,
4902
1923
3000
000
- 384,
6471
4673
2000
010
451,
0624
7093
7999
990
312,
2835
4249
0999
990
0,06
5401
1438
8840
0 12
,431
9061
5860
0000
23
3,83
2270
9599
9999
0 - 0,
8527
5803
5783
000
63
Zink
ensd
amm
800,
0000
0000
0000
000
0,73
3892
8462
3200
0 12
45,3
3463
1290
0001
00
- 1897
,091
3488
3000
0000
0,
1742
3203
2490
000
6,02
1349
6488
8000
0 23
13,2
4975
2759
9999
00
- 445,
3346
3129
0000
000
461,
1714
8034
5000
020
328,
6251
0933
9999
990
0,04
0306
6239
2920
0 20
,934
8729
6140
0000
32
7,98
8444
6919
9997
0 - 0,
9656
5952
2044
000
Hus
by
1100
,000
0000
0000
0000
0,
6711
2589
5414
000
810,
3657
8627
6999
980
46,2
4468
6972
4999
98
0,11
3087
3431
7300
0 51
,399
7137
1250
0002
24
8,37
5657
0810
0001
0 28
9,63
4213
7230
0002
0 44
8,08
2918
7159
9999
0 29
8,63
8869
9490
0002
0 0,
0576
9778
5725
000
11,2
0379
5838
4000
00
226,
9760
1918
0000
010
0,64
6385
3042
0400
0
Tele
fonp
lan
2400
,000
0000
0000
0000
0,
6730
7590
1797
000
1195
,448
8602
5000
0100
- 11
90,6
5843
3580
0001
00
0,14
1858
7724
0600
0 88
,474
0034
6230
0004
14
59,5
7328
6289
9999
00
1204
,551
1397
4999
9900
44
8,48
3774
7049
9997
0 39
2,90
7068
8620
0002
0 0,
0720
3341
5731
500
40,1
7040
0242
6000
03
336,
3780
9314
3999
990
2,68
5829
9178
0000
0
Band
hage
n
900,
0000
0000
0000
000
0,40
7945
7669
2500
0 85
7,23
6786
5400
0003
0 - 12
8,73
2604
9700
0001
0 0,
1605
2239
6335
000
55,1
8735
1483
6999
99
391,
2028
0211
4000
010
42,7
6321
3459
9000
01
458,
8593
9131
2000
010
341,
5204
7364
9999
990
0,06
5777
3722
1450
0 28
,765
3955
2699
9999
27
0,70
2483
3859
9998
0 0,
0931
9459
1348
800
Gub
bäng
en
600,
0000
0000
0000
000
0,40
3440
1567
7300
0 72
5,69
1115
6290
0002
0 36
,851
4265
1129
9998
0,
1586
8701
3575
000
42,6
8728
1261
4000
03
265,
4603
6385
1000
010
- 125,
6911
1562
9000
000
462,
1414
8677
8000
000
344,
1422
6395
9999
980
0,06
6346
4045
0760
0 29
,401
0037
3840
0000
27
1,24
4421
2660
0002
0 - 0,
2719
7539
9797
000
Thor
ildsp
lan
800,
0000
0000
0000
000
0,68
1898
0270
8000
0 59
2,18
5158
6020
0000
0 - 14
05,7
3642
6400
0000
00
0,19
8137
7222
2000
0 10
,531
2397
8030
0000
17
04,0
2346
1940
0001
00
207,
8148
4139
8000
000
436,
0171
7102
1000
020
308,
4896
8736
4000
020
0,03
7927
5520
6910
0 19
,995
1533
6509
9998
27
8,38
5205
4430
0002
0 0,
4766
2077
3700
000
Väl
lingb
y
1900
,000
0000
0000
0000
0,
6503
0087
0163
000
1505
,400
0363
0999
9900
59
,790
0463
1210
0003
0,
0511
2025
6270
600
71,9
8751
9605
0000
03
280,
3797
6409
4000
000
394,
5999
6368
8999
990
403,
6503
7259
3999
980
365,
5056
1171
3000
010
0,08
2238
0394
9570
0 25
,263
4383
1020
0002
30
0,01
8889
0179
9998
0 0,
9775
7859
3953
000
64
Skar
pnäc
k
1000
,000
0000
0000
0000
0,
4417
7639
3381
000
941,
1941
8006
5999
940
109,
1349
5601
1000
000
0,15
4234
2235
9300
0 32
,898
1035
2469
9998
23
1,49
6159
7610
0000
0 58
,805
8199
3380
0002
34
7,47
4097
0849
9998
0 35
7,32
6723
1930
0002
0 0,
0658
5474
2075
700
28,9
9987
4585
8999
99
281,
0703
4063
4999
980
0,16
9237
9962
3400
0
Kist
a
1200
,000
0000
0000
0000
0,
6755
5854
3280
000
1519
,245
7378
3999
9900
50
,124
2709
6100
0001
0,
1113
0190
7593
000
51,2
4187
8409
9999
98
241,
5173
3168
1999
990
- 319,
2457
3783
8000
030
423,
7988
7683
7000
020
296,
7883
9275
0000
010
0,05
5877
6784
9660
0 11
,146
1023
5920
0000
22
5,26
3483
2400
0000
0 - 0,
7532
9538
4406
000
Abr
aham
sber
g
1000
,000
0000
0000
0000
0,
6830
4213
8419
000
729,
2630
2882
1000
030
- 788,
4619
6097
4000
020
0,15
7739
5583
8400
0 46
,116
5846
5659
9997
10
46,7
0250
7190
0000
00
270,
7369
7117
9000
020
452,
5035
9347
5000
000
365,
5407
6469
7999
980
0,05
0356
2000
6370
0 27
,431
3092
0660
0002
35
0,25
5940
8290
0000
0 0,
5983
0899
7062
000
Kun
gsträ
dgår
den
50
0,00
0000
0000
0000
0 0,
6630
3481
9765
000
593,
0563
5542
0000
050
- 1424
,335
2154
7000
0100
0,
1616
1166
0971
000
- 1,42
1472
2853
8000
0 19
19,1
2102
6060
0001
00
- 93,0
5635
5420
1000
04
384,
7587
5872
7999
980
325,
2912
0876
1000
010
0,03
6347
8291
4010
0 20
,684
9987
0939
9999
32
4,61
3638
4370
0002
0 - 0,
2418
5636
6643
000
Björ
khag
en
700,
0000
0000
0000
000
0,52
4346
8813
4700
0 72
1,14
4402
2099
9995
0 - 16
9,06
1344
4910
0000
0 0,
1558
5358
1528
000
28,8
7600
5512
1999
99
497,
0137
3529
7999
970
- 21,1
4440
2209
5000
01
457,
1814
1953
4999
970
390,
4389
1385
7999
980
0,05
8442
4835
2050
0 26
,094
8518
2750
0001
31
1,88
3102
6909
9998
0 - 0,
0462
4947
8447
800
Hor
nstu
ll
2100
,000
0000
0000
0000
0,
7693
3980
3491
000
1652
,329
8697
2000
0000
- 20
46,3
1632
1230
0001
00
0,16
0295
7035
9500
0 21
,587
0928
8060
0000
24
36,9
4963
5999
9998
00
447,
6701
3028
3999
980
396,
9650
7086
2000
000
342,
8834
6423
3999
970
0,04
3563
5202
7500
0 22
,323
4481
2410
0000
32
2,68
6971
2779
9999
0 1,
1277
3179
0890
000
Vår
by g
ård
500,
0000
0000
0000
000
0,12
0337
9638
2200
0 77
7,30
9754
2440
0003
0 41
4,48
2591
1649
9997
0 0,
1624
5196
3342
000
34,2
1107
1513
5000
02
- 32,0
8699
9973
7000
03
- 277,
3097
5424
3999
980
449,
4829
5913
4999
990
396,
9569
0904
4999
970
0,12
2342
5389
3400
0 31
,523
2419
8730
0000
29
2,74
5306
9289
9998
0 - 0,
6169
5276
4522
000
65
Rådm
ansg
atan
1500
,000
0000
0000
0000
0,
6354
4591
1403
000
1406
,805
3608
4999
9900
- 12
26,2
3188
7520
0001
00
0,16
1028
8144
2300
0 4,
6481
9248
6900
000
1679
,811
5963
8000
0100
93
,194
6391
5369
9995
43
3,70
2479
9420
0002
0 30
5,85
0051
3020
0000
0 0,
0346
2502
7870
000
18,8
9791
6822
1000
01
281,
5518
0899
4999
990
0,21
4881
4993
3100
0
Mör
by
cent
rum
20
00,0
0000
0000
0000
00
0,62
4791
8185
1000
0 18
59,2
1131
7000
0000
00
- 215,
5682
6809
6000
000
0,15
1602
3847
4800
0 51
,592
2960
4080
0001
38
1,27
3326
9039
9999
0 14
0,78
8682
9970
0000
0 33
2,88
0547
4509
9998
0 31
7,19
9125
9889
9998
0 0,
0356
1571
7966
700
11,8
4814
7422
5000
00
233,
7430
0795
9000
010
0,42
2940
5535
2600
0
Axe
lsber
g
500,
0000
0000
0000
000
0,70
0621
1144
2000
0 73
5,47
8601
7530
0002
0 - 11
94,2
2763
1329
9999
00
0,20
0347
9839
3500
0 66
,752
4391
6319
9995
14
08,6
9179
3310
0001
00
- 235,
4786
0175
2999
990
445,
8762
6471
5999
980
392,
6047
1810
7000
000
0,05
6338
2976
1320
0 26
,915
1099
6639
9999
32
8,60
9790
6649
9999
0 - 0,
5281
2544
7320
000
Dan
dery
ds
sjukh
us
2800
,000
0000
0000
0000
0,
6056
5299
2824
000
2916
,349
2811
0000
0100
- 31
4,10
0111
8209
9998
0 0,
1489
4615
5225
000
49,1
1135
2470
5000
02
500,
0500
4856
5000
000
- 116,
3492
8110
0000
000
277,
5945
7805
3000
020
319,
3815
3943
6000
030
0,03
6088
4421
8560
0 12
,351
6435
5000
0000
24
9,69
3068
7329
9999
0 - 0,
4191
3383
8695
000
Alb
y
1000
,000
0000
0000
0000
0,
0558
8146
0074
800
839,
9106
0331
0000
060
657,
6131
4151
2999
960
0,09
7681
2289
8210
0 18
,129
0395
3229
9998
- 13
0,92
0604
8220
0000
0 16
0,08
9396
6900
0000
0 34
2,65
3564
4910
0000
0 42
4,24
9319
2920
0000
0 0,
1460
3021
3308
000
37,6
3459
9602
1999
98
308,
7095
1645
7000
010
0,46
7204
8193
2700
0
Näc
kros
en
1100
,000
0000
0000
0000
0,
6177
4243
0975
000
1243
,325
7538
5000
0000
- 36
6,68
9053
1110
0002
0 0,
1724
2812
0435
000
48,2
0034
9746
2999
99
483,
2498
0355
2000
000
- 143,
3257
5384
5000
010
419,
4136
0731
3999
990
303,
6644
2080
7999
990
0,03
6456
8340
5290
0 11
,051
4748
6810
0000
23
6,95
2014
8930
0001
0 - 0,
3417
2890
7565
000
Rink
eby
1000
,000
0000
0000
0000
0,
6331
6677
9528
000
977,
9547
4881
4000
030
- 23,2
1540
1972
7999
99
0,11
7597
1897
3600
0 52
,190
5823
6890
0003
30
5,72
4158
0900
0000
0 22
,045
2511
8610
0000
40
6,45
8201
5259
9998
0 28
7,14
8214
2489
9998
0 0,
0535
5232
9012
600
11,3
2652
7368
3000
01
221,
9714
4196
4000
010
0,05
4237
4372
1570
0
66
Riss
ne
900,
0000
0000
0000
000
0,60
5641
0105
2300
0 12
21,9
8273
4170
0000
00
- 59,4
0676
2512
9000
02
0,10
8503
1384
1000
0 53
,086
9703
7470
0002
34
6,91
5261
4650
0001
0 - 32
1,98
2734
1690
0003
0 41
6,42
0265
6609
9997
0 28
7,97
4917
1170
0002
0 0,
0533
1397
4660
000
12,2
6747
7916
2000
01
225,
8127
8756
8999
990
- 0,77
3215
8127
7500
0
Gär
det
1100
,000
0000
0000
0000
0,
6349
1064
7325
000
729,
8201
9431
1000
020
- 1265
,311
3791
5999
9900
0,
1237
5681
3877
000
- 2,42
8856
7278
6000
0 19
07,1
7167
6519
9999
00
370,
1798
0568
8999
980
391,
5896
2541
1999
980
333,
9916
5367
1999
980
0,03
5733
4392
2240
0 19
,436
7091
9399
9999
34
8,07
1722
2179
9999
0 0,
9453
2587
6035
000
Häs
selb
y gå
rd
900,
0000
0000
0000
000
0,65
2056
8142
3900
0 68
5,47
5127
4010
0001
0 57
,456
1728
5310
0000
0,
0654
6239
1076
800
73,5
1811
2400
8999
94
258,
3412
9047
2000
030
214,
5248
7259
8999
990
433,
1878
4009
9000
030
377,
9364
3674
7000
020
0,08
8088
4263
4960
0 25
,628
9925
8010
0000
30
6,68
4217
2199
9999
0 0,
4952
2367
1445
000
Rådh
uset
1000
,000
0000
0000
0000
0,
6754
3340
4266
000
1210
,704
1577
6000
0000
- 14
84,5
4407
7609
9999
00
0,18
1313
9904
0600
0 2,
6348
3549
3130
000
1881
,368
3457
4000
0000
- 21
0,70
4157
7560
0000
0 46
2,59
3480
3470
0002
0 31
9,09
2095
1560
0003
0 0,
0373
8232
6200
400
19,5
6152
1195
2000
01
291,
7235
4606
0999
980
- 0,45
5484
4949
3400
0
Kris
tineb
erg
1100
,000
0000
0000
0000
0,
6982
0055
5401
000
1354
,258
1557
0000
0100
- 13
51,5
6860
6170
0001
00
0,19
9061
7186
0500
0 20
,894
3680
6390
0000
16
08,4
2084
4369
9999
00
- 254,
2581
5570
2000
010
406,
5484
2431
9999
980
327,
7684
8696
8999
980
0,04
0694
1304
0420
0 22
,430
7539
2349
9999
30
4,82
3742
2440
0002
0 - 0,
6254
0681
6241
000
Hög
dale
n
1000
,000
0000
0000
0000
0,
3827
1581
8095
000
1357
,863
8938
3000
0100
- 54
,161
5567
0879
9999
0,
1823
5871
6736
000
53,4
7779
8853
1999
99
299,
8720
9499
4000
010
- 357,
8638
9383
4000
010
351,
3720
8787
7000
010
349,
6717
4465
6999
980
0,06
9977
4109
2540
0 30
,309
5275
6180
0000
27
6,84
7060
6550
0001
0 - 1,
0184
7558
8080
000
Alv
ik
2100
,000
0000
0000
0000
0,
7071
3817
5110
000
2072
,576
8252
3999
9800
- 12
33,4
9705
7200
0000
00
0,19
6079
9048
6800
0 29
,841
1756
8850
0002
14
63,8
4381
2679
9999
00
27,4
2317
4761
3999
99
235,
0801
4513
0000
010
354,
8642
9502
8000
020
0,04
4752
7382
2610
0 25
,579
9076
1570
0002
34
1,94
5324
3080
0001
0 0,
1166
5457
6448
000
67
Blås
ut
600,
0000
0000
0000
000
0,53
2092
7067
6400
0 67
3,42
5759
1069
9998
0 - 39
7,08
2289
1259
9998
0 0,
1463
8760
9063
000
41,0
1774
3950
2000
00
680,
5317
7823
6000
040
- 73,4
2575
9107
1000
06
454,
6266
3597
8000
020
364,
0224
8547
2000
030
0,06
1923
5150
7050
0 27
,574
7192
0849
9999
28
3,41
1733
0070
0001
0 - 0,
1615
0782
4875
000
Höt
orge
t
500,
0000
0000
0000
000
0,64
7984
9880
9000
0 75
0,25
2931
1920
0001
0 - 13
21,2
7422
7140
0000
00
0,16
3042
7052
0700
0 1,
6552
4606
6950
000
1790
,218
1634
2000
0100
- 25
0,25
2931
1920
0001
0 45
2,21
4145
6990
0003
0 31
3,84
1218
8950
0000
0 0,
0353
0847
7255
500
19,5
5383
0174
4000
02
293,
3310
2328
0000
010
- 0,55
3394
7435
5700
0
Baga
rmos
sen
1100
,000
0000
0000
0000
0,
4612
8435
1609
000
702,
8904
1280
7999
950
56,3
4777
3490
1000
03
0,15
3108
5533
0500
0 32
,325
5333
1770
0000
28
0,26
7722
6710
0000
0 39
7,10
9587
1919
9999
0 44
9,37
2364
9890
0000
0 35
9,92
3517
0719
9998
0 0,
0624
2623
6952
700
27,2
0641
2238
1000
00
282,
2313
3316
3999
980
0,88
3698
2826
0800
0
Sand
sbor
g
700,
0000
0000
0000
000
0,49
6369
0024
1000
0 40
5,06
3923
3289
9998
0 - 24
0,82
0095
2320
0000
0 0,
1428
0533
9955
000
44,5
4452
9705
8000
02
526,
2346
1640
0000
050
294,
9360
7667
1000
020
447,
5850
7703
7000
020
355,
9756
5532
3000
010
0,06
2275
1254
3780
0 26
,966
5394
1320
0000
27
7,02
6952
7610
0002
0 0,
6589
4975
4588
000
Sätra
800,
0000
0000
0000
000
0,33
5085
9182
8300
0 60
7,80
5541
8679
9998
0 - 22
,957
6845
6349
9999
0,
2581
3241
1632
000
58,7
1683
5610
8999
99
220,
4866
2224
8000
000
192,
1944
5813
1999
990
419,
1426
9541
8000
020
441,
1052
5263
4000
010
0,10
9618
7322
2500
0 26
,932
5834
9110
0001
32
1,03
3321
9639
9998
0 0,
4585
4182
8912
000
Fars
ta
1200
,000
0000
0000
0000
0,
3754
1773
0892
000
1024
,581
7853
7000
0000
13
8,03
0743
4339
9999
0 0,
1676
2843
4815
000
38,1
6107
6433
2000
02
183,
6035
5974
6000
000
175,
4182
1462
6000
010
309,
3195
9797
2000
000
345,
6227
6719
8000
020
0,06
9781
3410
4490
0 31
,574
6529
8929
9999
27
4,76
0177
6970
0002
0 0,
5671
0992
7001
000
Fruä
ngen
1800
,000
0000
0000
0000
0,
3924
6150
9290
000
1905
,784
5264
6999
9900
- 30
6,27
4415
9389
9999
0 0,
2349
0979
9387
000
61,2
6044
4889
5000
00
548,
4402
0356
2999
950
- 105,
7845
2647
3000
000
279,
9935
3354
5999
980
448,
9628
1957
3999
980
0,09
2025
7868
3400
0 36
,669
9310
4619
9999
37
3,01
5693
2690
0000
0 - 0,
3778
1060
5601
000
68
Duv
bo
600,
0000
0000
0000
000
0,59
6077
6887
9100
0 99
6,11
8595
5680
0005
0 - 19
9,28
1545
8840
0000
0 0,
1354
8263
8000
000
50,9
2788
8303
8999
98
419,
8163
7323
7999
980
- 396,
1185
9556
7999
990
426,
2139
8529
7000
020
291,
1977
9820
7000
010
0,04
3367
8445
2250
0 11
,841
1486
1049
9999
23
1,83
1583
1470
0000
0 - 0,
9293
8901
4046
000
Soln
a ce
ntru
m
1600
,000
0000
0000
0000
0,
6258
5369
6892
000
1954
,171
8613
9000
0000
- 53
4,39
4878
0619
9998
0 0,
1905
2300
6555
000
40,7
8159
1501
1000
03
634,
6324
9951
7000
040
- 354,
1718
6139
0999
990
399,
2983
0639
1999
970
311,
5335
8747
3000
010
0,03
5995
2077
6770
0 12
,322
3951
8510
0000
25
8,90
6398
9840
0002
0 - 0,
8869
8563
3850
000
Ode
npla
n
2100
,000
0000
0000
0000
0,
6294
1536
2418
000
3753
,596
0217
1000
0100
- 11
27,0
7904
4740
0000
00
0,17
0892
2608
4200
0 9,
2039
3238
0120
000
1506
,544
1289
4000
0100
- 16
53,5
9602
1710
0001
00
376,
6129
1069
9999
990
296,
9431
2837
6999
990
0,03
4166
1561
9490
0 17
,678
7249
9610
0001
26
9,58
4777
4669
9998
0 - 4,
3907
0455
2950
000
Öste
rmal
msto
rg 80
0,00
0000
0000
0000
0 0,
6470
6971
2298
000
1381
,977
9862
5000
0000
- 13
34,0
0684
0940
0001
00
0,15
1357
6683
7500
0 - 0,
1382
8457
9452
000
1860
,610
3508
3999
9900
- 58
1,97
7986
2460
0000
0 44
0,49
1250
5479
9998
0 31
6,37
3822
9400
0003
0 0,
0352
0405
0758
500
19,9
1592
8182
9000
00
308,
4746
0624
2999
980
- 1,32
1202
1476
5000
0
Stad
shag
en
700,
0000
0000
0000
000
0,67
3518
9274
0500
0 10
43,6
0550
9120
0001
00
- 1273
,859
9671
5000
0100
0,
2134
5029
1695
000
6,90
6748
5433
7000
0 15
22,1
3412
2080
0000
00
- 343,
6055
0911
8999
980
440,
6945
9321
8000
020
313,
1008
9751
1000
030
0,03
7937
0926
5410
0 20
,354
5415
7559
9999
28
2,45
0160
3659
9998
0 - 0,
7796
9077
5442
000
Tekn
iska
högs
kola
n 20
00,0
0000
0000
0000
00
0,62
4253
2855
9500
0 27
85,3
9694
7790
0000
00
- 1162
,273
4856
8000
0000
0,
1442
8047
6550
000
6,66
3985
4240
9000
0 16
62,4
2796
5509
9999
00
- 785,
3969
4779
0000
010
369,
1189
5173
9000
010
310,
0142
1823
5999
980
0,03
4220
3902
5360
0 16
,724
3808
5690
0002
29
3,93
2742
6070
0002
0 - 2,
1277
6110
2730
000
Sund
bybe
rg
1500
,000
0000
0000
0000
0,
5935
5462
1426
000
2113
,684
1935
6000
0000
- 34
9,83
6604
2000
0001
0 0,
1540
6081
6299
000
48,8
0956
1265
7999
99
520,
0809
6552
2000
040
- 613,
6841
9356
4000
000
400,
8963
7748
0000
010
311,
9165
7746
9000
000
0,04
0106
4207
8610
0 13
,199
0997
6410
0000
24
7,36
8569
9670
0001
0 - 1,
5307
8009
1910
000
69
Åke
shov
500,
0000
0000
0000
000
0,71
0611
1012
3600
0 58
1,86
0181
3620
0005
0 - 17
5,30
1210
1180
0000
0 0,
0287
9733
6472
600
64,8
4085
7528
3000
06
600,
8421
4018
6000
050
- 81,8
6018
1362
0000
06
436,
0845
9174
2999
980
377,
2296
2683
9999
980
0,07
2810
0863
3230
0 27
,776
7244
1820
0001
32
1,78
9331
1620
0000
0 - 0,
1877
1628
9252
000
Sved
myr
a
400,
0000
0000
0000
000
0,45
1790
6354
8900
0 64
1,67
2465
7390
0002
0 - 20
9,28
1569
5220
0001
0 0,
1408
8003
0524
000
54,1
5733
7540
2999
99
479,
2972
4570
5000
020
- 241,
6724
6573
8999
990
438,
7935
0462
0000
020
349,
8867
1455
8999
980
0,06
4577
2159
1700
0 28
,063
3116
7830
0000
27
6,26
5738
1760
0001
0 - 0,
5507
6582
3092
000
Fars
ta st
rand
700,
0000
0000
0000
000
0,36
9912
5191
5600
0 10
20,8
7316
1870
0000
00
155,
1699
4445
3999
990
0,16
9378
8682
8900
0 37
,454
0626
5679
9998
17
0,02
7653
7860
0000
0 - 32
0,87
3161
8709
9998
0 34
5,17
0703
3710
0000
0 34
6,27
6223
0089
9998
0 0,
0706
1912
2909
800
32,0
232 7
0303
4999
99
275,
9345
3968
5000
000
- 0,92
9607
1733
1400
0
Sock
enpl
an
500,
0000
0000
0000
000
0,49
1468
8287
9200
0 58
1,14
3183
8610
0001
0 - 37
4,94
9291
1510
0001
0 0,
1335
6366
2847
000
55,8
0957
8995
8000
03
636,
8018
5336
8000
020
- 81,1
4318
3860
7000
06
461,
9641
4607
8000
000
357,
0885
9251
8999
970
0,06
5957
0629
1710
0 29
,079
5765
3260
0001
28
2,03
0503
4819
9997
0 - 0,
1756
4822
8439
000
Bred
äng
1100
,000
0000
0000
0000
0,
4419
9738
9828
000
1076
,979
6052
6000
0000
- 22
3,70
4196
2479
9999
0 0,
2357
9020
6964
000
67,4
5337
4499
6000
02
432,
6531
7313
8000
000
23,0
2039
4736
8000
00
445,
4306
1250
0999
980
429,
6112
6843
5000
000
0,08
6332
4378
2030
0 27
,336
0487
8479
9999
34
1,91
2517
3560
0002
0 0,
0516
8121
3842
700
Skog
skyr
kogå
rde
n 40
0,00
0000
0000
0000
0 0,
4665
5515
7685
000
741,
6258
0585
9000
020
- 101,
9519
9617
1000
000
0,14
7651
9894
1700
0 41
,736
1158
1899
9998
40
0,47
4721
3570
0002
0 - 34
1,62
5805
8590
0002
0 46
0,52
1819
8499
9999
0 35
1,17
6656
7970
0002
0 0,
0618
4739
1114
100
26,5
8259
4784
2000
01
273,
8246
1389
5000
030
- 0,74
1823
2777
1200
0
Mid
som
mar
kran
sen
900,
0000
0000
0000
000
0,76
3772
7983
7500
0 88
6,38
5013
8999
9999
0 - 17
95,8
9534
8380
0001
00
0,15
2855
6614
7000
0 65
,085
9725
1320
0005
20
95,0
3945
0190
0000
00
13,6
1498
6100
1999
99
374,
5766
0104
1999
990
382,
6848
4910
7999
980
0,06
6216
4951
0140
0 36
,614
4113
6599
9999
33
1,39
2967
7159
9999
0 0,
0363
4766
8440
500
70
Tens
ta
900,
0000
0000
0000
000
0,63
0302
7641
0500
0 70
8,47
1445
9450
0003
0 13
,064
5065
5670
0000
0,
1057
1943
1990
000
53,1
4901
1678
9999
97
292,
4380
5827
0999
990
191,
5285
5405
5000
000
455,
0566
5826
6999
990
304,
2419
1955
9000
000
0,06
2207
4865
1610
0 11
,953
7810
5330
0000
22
9,54
9677
5180
0001
0 0,
4208
8946
6345
000
Stor
a m
osse
n
500,
0000
0000
0000
000
0,68
9576
4162
7800
0 59
9,83
4146
0399
9995
0 - 10
11,6
6904
5660
0001
00
0,18
1706
0148
7300
0 37
,727
4445
6299
9999
12
42,0
0107
7550
0000
00
- 99,8
3414
6039
4999
98
442,
6444
7279
6000
000
356,
9096
7013
5000
000
0,04
6169
8214
7990
0 26
,446
9982
6890
0000
34
6,24
5833
5200
0002
0 - 0,
2255
4025
2223
000
Stad
ion
700,
0000
0000
0000
000
0,63
5028
6707
6000
0 11
10,8
2656
2120
0001
00
- 1261
,170
1128
8999
9900
0,
1396
8471
3089
000
0,83
8112
5463
8100
0 18
24,2
9288
0840
0000
00
- 410,
8265
6211
8000
030
462,
3062
3559
6000
020
317,
9646
2515
7999
990
0,03
4812
7206
1840
0 18
,876
1427
0199
9999
31
2,78
4702
5739
9999
0 - 0,
8886
4594
6098
000
Skär
holm
en
800,
0000
0000
0000
000
0,22
2348
2304
9300
0 12
55,3
8527
3680
0000
00
169,
4783
0307
9999
990
0,22
6824
5455
8500
0 45
,939
0767
9969
9997
96
,406
7294
0189
9995
- 45
5,38
5273
6760
0000
0 34
9,02
9954
1320
0000
0 39
8,75
6201
6410
0002
0 0,
1105
3916
3734
000
27,5
8571
1070
3999
99
288,
2830
0686
7999
970
- 1,30
4716
8825
6000
0
Vår
berg
1100
,000
0000
0000
0000
0,
1686
0012
4614
000
526,
0413
0552
5999
970
326,
6084
8999
9000
030
0,17
6835
6165
4600
0 44
,198
5566
0780
0001
1,
7512
9128
6740
000
573,
9586
9447
4000
030
398,
1110
4286
6999
980
395,
0872
5776
4000
010
0,11
7854
8832
8800
0 28
,577
2395
0540
0001
29
0,43
6883
5889
9999
0 1,
4417
0503
3700
000
Häs
selb
y str
and
700,
0000
0000
0000
000
0,67
4047
4395
1500
0 71
6,33
3303
7260
0005
0 46
,333
7686
9290
0001
0,
0672
5061
5545
600
76,8
5473
5540
9999
97
262,
1754
2385
8999
980
- 16,3
3330
3725
4999
99
451,
9986
0165
5000
020
388,
9909
5735
6000
020
0,09
4088
9903
5150
0 26
,098
0883
6630
0001
31
5,10
6225
2460
0001
0 - 0,
0361
3573
9503
900
Häg
erste
nsås
en 10
00,0
0000
0000
0000
00
0,59
7229
9785
7800
0 79
6,12
1720
1040
0002
0 - 89
7,01
5075
5670
0005
0 0,
1782
1895
6362
000
80,1
8026
6424
3000
01
1135
,526
2090
9000
0100
20
3,87
8279
8960
0001
0 44
0,01
3351
7169
9998
0 41
0,36
4521
1269
9999
0 0,
0686
7531
4011
400
35,5
2756
6082
6999
98
347,
6052
1554
7999
990
0,46
3345
6668
9900
0
71
Väs
tra sk
ogen
1300
,000
0000
0000
0000
0,
6459
0222
8646
000
600,
7215
1355
9999
950
- 897,
5524
7891
0999
980
0,21
3675
1157
9900
0 16
,091
1945
0620
0001
10
79,2
0852
2220
0001
00
699,
2784
8644
0000
050
443,
0670
1547
7999
970
315,
3539
7270
9000
000
0,03
7702
5212
1500
0 19
,955
9913
5450
0000
29
3,37
9526
3289
9999
0 1,
5782
6798
6580
000
Mar
iato
rget
1900
,000
0000
0000
0000
0,
7393
5191
5749
000
1756
,509
7653
6000
0100
- 19
24,5
9825
5950
0001
00
0,17
8761
5085
1500
0 2,
1592
2020
9290
000
2334
,935
2874
4000
0100
14
3,49
0234
6410
0000
0 40
2,72
2681
5240
0000
0 33
5,93
9103
6719
9999
0 0,
0410
2789
3464
900
21,9
8474
2328
5000
01
349,
7227
7308
8999
990
0,35
6300
3556
1500
0
Blac
kebe
rg
500,
0000
0000
0000
000
0,72
3517
9132
0600
0 56
5,56
3354
0159
9995
0 9,
5658
4797
3550
000
0,01
7669
9300
7300
0 76
,242
3953
8999
9999
38
2,98
0256
5179
9998
0 - 65
,563
3540
1619
9995
44
9,98
4071
5870
0002
0 38
0,32
5816
8699
9998
0 0,
0864
1032
4258
000
27,0
3241
8951
8999
99
320,
1894
6008
5000
010
- 0,14
5701
4995
7800
0
Råck
sta
1100
,000
0000
0000
0000
0,
6749
2348
0757
000
629,
9899
6982
6999
980
42,1
8093
4125
5999
97
0,03
4374
8337
6040
0 73
,023
1463
6229
9997
32
5,56
4805
4120
0000
0 47
0,01
0030
1730
0002
0 45
6,74
6585
5560
0001
0 36
4,04
1598
1040
0000
0 0,
0814
5529
5604
100
25,6
0122
3515
8999
99
302,
5591
8535
0000
010
1,02
9038
9573
5000
0
Asp
udde
n
1000
,000
0000
0000
0000
0,
7768
7771
2488
000
1053
,380
7790
5000
0000
- 16
78,9
0187
6680
0000
00
0,14
8994
2297
8100
0 73
,704
3187
0439
9995
19
50,4
5106
8800
0000
00
- 53,3
8077
9053
6000
01
453,
0881
5845
2000
020
374,
2404
7 716
7999
980
0,06
3776
9112
8500
0 35
,170
1655
3999
9999
31
2,97
2036
3910
0002
0 - 0,
1178
1543
6263
000
Rops
ten
3800
,000
0000
0000
0000
0,
6216
4459
2548
000
2664
,041
7735
8000
0200
- 11
06,7
0397
4340
0001
00
0,11
2785
5568
8600
0 10
,618
8913
6289
9999
16
97,7
6131
2760
0000
00
1135
,958
2264
2000
0100
34
2,73
9677
7110
0001
0 32
0,10
4193
3560
0000
0 0,
0352
8835
7934
900
14,1
5582
1723
7000
01
319,
3052
1740
0000
000
3,31
4347
0111
5000
0
Nor
sbor
g
400,
0000
0000
0000
000
0,06
5499
4193
8290
0 77
5,04
5439
2310
0005
0 62
3,28
7192
7899
9995
0 0,
0634
9577
1962
700
31,2
5174
0709
6000
00
- 129,
7298
1167
3000
000
- 375,
0454
3923
0999
990
259,
9591
5393
0000
010
414,
7859
1884
5000
030
0,14
6985
2282
5200
0 34
,235
9850
7919
9999
31
0,83
1266
1769
9997
0 - 1,
4427
0910
8570
000
72
Hal
lund
a
300,
0000
0000
0000
000
0,06
1613
7368
9130
0 89
2,98
0589
6530
0004
0 62
5,46
5650
4700
0001
0 0,
0790
2522
6269
800
27,5
6382
2097
3000
01
- 127,
9415
0668
6000
000
- 592,
9805
8965
3000
040
418,
0844
7536
5000
000
417,
0960
7203
7000
000
0,14
5125
9733
4300
0 35
,000
0691
4490
0003
30
9,27
4601
4269
9999
0 - 1,
4183
2721
5180
000
73
Appendix III: Table of GWR Equations and Predictions for the 2016 Situation with New Stations
Stat
ion
Obs
erve
d
Loca
lR2
Pred
icte
d
Inte
rcep
t
Wor
kers
Bus
Chan
ge
Resid
ual
Std
Erro
r
Std
Err I
nt
Std
Err
Wor
kers
St
d Er
r Bus
Std
Err
Chan
ge
Std
Resid
Äng
bypl
an
400,
0000
0000
0000
000
0,72
7369
2341
9800
0 58
7,38
8321
5719
9999
5 - 81
,939
3552
3659
9994
0,
0139
8781
2387
000
70,0
6781
6737
8000
01
515,
2601
8219
5999
960
- 187,
3883
2157
1999
995
433,
1980
8421
5999
984
361,
2608
6556
0000
013
0,07
6271
1518
7810
0 26
,936
7417
9989
9998
30
6,74
6430
0500
0001
5 - 0,
4325
6959
8990
000
Aka
lla
800,
0000
0000
0000
000
0,69
9401
7182
6100
0 92
7,32
9708
6750
0000
6 0,
2248
9695
2387
000
0,12
6336
3590
1500
0 51
,782
3458
1449
9998
26
6,83
7991
5310
0000
0 - 12
7,32
9708
6750
0000
6 38
8,83
9882
6839
9997
4 28
2,96
1593
7410
0000
2 0,
0536
5378
3224
400
10,8
8269
1585
9000
00
221,
9512
7054
4000
010
- 0,32
7460
5161
2800
0
Fittj
a
600,
0000
0000
0000
000
0,07
8572
8103
4390
0 93
8,01
1866
1190
0004
7 53
7,23
9307
3380
0000
3 0,
1244
7410
4369
000
27,4
0168
2787
1999
99
- 86,1
8470
2012
9000
04
- 338,
0118
6611
8999
990
419,
2909
0940
6000
026
393,
9943
7839
0000
008
0,12
8862
2512
6800
0 33
,485
1190
1740
0002
29
2,55
9789
4320
0000
2 - 0,
8061
5119
1301
000
Brom
map
lan
3000
,000
0000
0000
0000
0,
6674
9263
7530
000
2576
,626
5592
2000
0217
- 53
6,56
5773
9329
9994
6 0,
1334
9341
4738
000
55,5
1024
5585
4000
00
816,
5178
6778
8999
979
423,
3734
4077
6999
985
182,
5143
2626
1000
008
338,
2088
0207
2000
026
0,05
1643
3753
7140
0 26
,135
7611
2559
9998
30
8,45
7002
0169
9997
9 2,
3196
7237
5590
000
Lilje
holm
en
4200
,000
0000
0000
0000
0,
8032
6067
1494
000
3606
,655
3792
9999
9822
- 21
39,2
6651
4139
9997
99
0,15
4922
8248
7900
0 37
,578
3648
1310
0002
24
94,1
9462
2619
9997
91
593,
3446
2070
4000
022
228,
2158
0217
1999
997
332,
3145
5162
4999
979
0,04
2765
4370
2670
0 21
,870
0962
8750
0001
30
8,87
4828
1809
9999
8 2,
5999
2785
3620
000
74
Örn
sber
g
900,
0000
0000
0000
000
0,74
9512
1592
5600
0 79
2,47
9968
4939
9999
1 - 14
49,2
4409
4910
0000
58
0,17
4508
9244
5300
0 66
,788
3121
9740
0003
16
97,6
2515
1510
0000
23
107,
5200
3150
5999
995
432,
9593
2172
4000
006
360,
3775
3392
0000
019
0,05
2807
8802
5180
0 27
,269
1036
6430
0001
30
3,75
4306
0709
9997
4 0,
2483
3749
0639
000
Kär
rtorp
900,
0000
0000
0000
000
0,69
6040
0775
3500
0 66
6,97
2846
1299
9999
4 - 11
1,08
0881
9939
9999
5 0,
1487
8941
2924
000
46,3
4812
7726
5999
98
386,
6620
4942
2999
985
233,
0271
5387
0000
006
452,
2995
9469
8000
021
337,
6761
8939
5999
984
0,04
2884
8066
7130
0 11
,763
3449
6460
0000
27
0,82
4391
7379
9997
4 0,
5152
0531
2146
000
Stur
eby
500,
0000
0000
0000
000
0,47
7068
9437
0800
0 67
1,22
7687
2759
9998
3 - 25
8,76
9051
9849
9997
6 0,
1539
5138
0013
000
54,3
8534
6745
6999
98
521,
1748
0612
5000
032
- 171,
2276
8727
6000
012
444,
4163
2152
1999
976
329,
9189
8468
6999
977
0,04
8650
8983
1290
0 21
,008
3100
4650
0000
25
9,29
2953
9250
0000
6 - 0,
3852
8667
5994
000
Glo
ben
300,
0000
0000
0000
000
0,65
4847
2756
8700
0 55
0,08
4690
4029
9996
8 - 10
89,6
5590
5110
0000
49
0,16
0818
9301
3800
0 43
,526
0309
7400
0001
13
28,7
2366
2979
9999
72
- 250,
0846
9040
2999
996
448,
7941
3339
6999
996
342,
6624
7206
7000
010
0,04
5933
4254
1710
0 19
,689
0683
6520
0001
26
8,18
1896
1600
0000
8 - 0,
5572
3698
6388
000
Huv
udsta
600,
0000
0000
0000
000
0,63
0480
4599
7900
0 92
8,76
1586
3700
0003
2 - 77
5,79
6494
9280
0001
6 0,
1920
0472
8508
000
31,6
3919
2601
2000
01
934,
5262
2374
2999
946
- 328,
7615
8636
9999
975
441,
6676
1863
6999
975
308,
2646
2831
2000
013
0,03
6903
9022
5290
0 18
,756
1503
2709
9999
27
7,19
9383
4350
0001
6 - 0,
7443
6425
1525
000
Uni
vers
itete
t
500,
0000
0000
0000
000
0,59
7722
2575
8200
0 82
2,85
8363
7010
0003
0 - 75
9,45
9815
3360
0002
0 0,
1476
0362
4606
000
33,6
1838
4413
5999
98
1058
,263
5271
3999
9951
- 32
2,85
8363
7009
9997
3 42
2,25
9521
9660
0002
3 27
0,94
4194
9999
9997
9 0,
0325
3104
0981
700
11,3
1799
1634
7000
00
240,
6181
0871
2000
009
- 0,76
4596
9999
6400
0
Ham
mar
byhö
jde
n 90
0,00
0000
0000
0000
0 0,
7002
0994
9697
000
1014
,155
3434
8000
0056
- 52
4,90
3935
2720
0001
2 0,
1569
4308
0148
000
40,5
7226
4056
9000
03
798,
9008
4574
1000
012
- 114,
1553
4347
9999
999
456,
4799
3781
7999
996
342,
4362
8395
1999
997
0,04
1564
3799
0430
0 12
,305
3828
0670
0001
28
2,21
6966
9840
0001
0 - 0,
2500
7746
0197
000
75
Joha
nnel
und
200,
0000
0000
0000
000
0,65
4873
5241
9300
0 47
7,50
4039
4889
9997
9 14
,734
6771
7910
0000
0,
0739
2667
4640
900
69,7
2890
9161
9000
06
282,
2822
3956
0999
983
- 277,
5040
3948
8999
979
432,
4002
1045
3999
989
341,
5417
0550
9999
986
0,07
9869
7916
8340
0 24
,115
5059
6340
0000
28
9,10
4519
2400
0000
2 - 0,
6417
7591
2176
000
Hag
sätra
1200
,000
0000
0000
0000
0,
3854
9536
9761
000
681,
7309
1320
1000
021
- 104,
7357
5369
3999
996
0,21
9770
7457
2800
0 62
,773
4343
9580
0003
29
2,66
5027
1469
9998
9 51
8,26
9086
7989
9997
9 43
4,81
2663
2729
9999
8 35
3,61
7627
8030
0000
5 0,
0724
9394
8129
700
30,5
0297
9870
0999
99
284,
7539
3565
6000
010
1,19
1936
5064
0000
0
Mäl
arhö
jden
700,
0000
0000
0000
000
0,59
5077
0418
7300
0 84
0,11
5278
3840
0002
1 - 76
5,91
7383
5679
9999
1 0,
2221
5421
2042
000
60,9
3848
1498
3000
00
985,
0608
8391
1000
019
- 140,
1152
7838
3999
993
440,
9052
3656
2000
027
384,
8715
3210
7000
007
0,05
7862
3862
3270
0 25
,904
5522
3660
0001
32
9,96
3545
7150
0000
9 - 0,
3177
9000
7388
000
Isla
ndsto
rget
800,
0000
0000
0000
000
0,73
1186
2311
2500
0 64
8,49
7797
5640
0005
2 - 23
,387
8223
7339
9999
0,
0091
9549
0242
980
73,7
7456
9299
4000
00
439,
0491
5825
5000
009
151,
5022
0243
6000
005
443,
2194
6377
0999
994
361,
7095
1728
5999
993
0,08
1006
1914
7180
0 26
,545
5289
4260
0001
30
8,36
0141
1020
0000
0 0,
3418
2208
7746
000
Berg
sham
ra
1100
,000
0000
0000
0000
0,
5974
5029
5291
000
874,
6628
8122
9000
050
- 490,
7604
5894
1000
024
0,15
9759
6330
3800
0 45
,663
9197
9699
9998
66
2,30
2363
6740
0004
8 22
5,33
7118
7710
0000
7 43
6,16
2801
1889
9999
3 27
5,58
4733
6339
9997
4 0,
0327
8500
6794
300
11,2
9692
2815
9000
00
224,
5318
1224
7999
994
0,51
6635
3438
5900
0
Skan
stull
2200
,000
0000
0000
0000
0,
8125
0395
3361
000
1892
,498
0732
7000
0077
- 23
67,4
7256
9500
0001
91
0,16
8581
3330
6500
0 26
,670
6613
2890
0000
26
67,7
0353
5399
9999
64
307,
5019
2672
7000
011
412,
5073
4397
2000
001
357,
1967
1504
7999
987
0,03
9399
6040
8300
0 19
,473
8446
1680
0001
35
4,61
8295
0940
0001
8 0,
7454
4594
4711
000
Hök
arän
gen
1000
,000
0000
0000
0000
0,
4821
8400
8004
000
1022
,036
5459
2999
9988
76
,951
9074
7190
0000
0,
1617
5590
3634
000
44,4
2546
5951
8000
00
219,
5378
4891
0000
008
- 22,0
3654
5929
5000
01
384,
5855
5419
7000
022
324,
4590
1551
2000
008
0,04
6478
0325
3370
0 17
,445
2809
0129
9999
25
2,95
7055
8229
9999
0 - 0,
0572
9946
3510
900
76
Mas
mo
200,
0000
0000
0000
000
0,09
4018
9376
5650
0 52
2,00
1586
0510
0000
4 48
0,57
5415
5129
9999
6 0,
1577
2639
8770
000
24,3
9600
4845
2999
98
- 51,8
6981
7153
6999
98
- 322,
0015
8605
1000
004
273,
8021
9410
2999
977
388,
4753
0541
2000
012
0,12
3335
1168
4000
0 33
,910
6493
4239
9999
28
7,05
2280
3900
0002
1 - 1,
1760
3727
4300
000
Frid
hem
spla
n
5700
,000
0000
0000
0000
0,
6615
6497
7584
000
4200
,304
0806
7000
0076
- 13
15,1
5695
0379
9998
98
0,19
2542
2047
6500
0 8,
7217
4666
9440
000
1634
,597
0635
5000
0030
14
99,6
9591
9329
9999
24
341,
3100
9839
0000
007
292,
6696
3877
6999
989
0,03
4966
6724
2610
0 17
,485
6811
5870
0000
25
5,51
5115
0039
9999
5 4,
3939
3948
9050
000
Väs
terto
rp
700,
0000
0000
0000
000
0,52
8774
7229
2500
0 10
36,1
6336
4289
9998
90
- 640,
7475
5218
2000
049
0,21
3929
9791
0900
0 61
,947
0310
4279
9999
88
5,74
0645
3679
9994
6 - 33
6,16
3364
2929
9997
2 44
6,57
2797
8549
9997
6 38
5,15
8093
4849
9999
5 0,
0602
8327
8249
100
27,8
3323
6612
0000
00
329,
5748
5083
3000
028
- 0,75
2762
7430
6900
0
S:t E
riksp
lan
2500
,000
0000
0000
0000
0,
6381
4913
0343
000
1954
,873
8397
1000
0084
- 11
37,1
1290
6390
0000
35
0,18
7445
9785
2400
0 10
,553
1547
1180
0000
14
54,3
6576
2919
9999
52
545,
1261
6028
5999
958
389,
3962
5912
2000
004
280,
7856
0278
8999
995
0,03
3470
4442
8740
0 17
,043
7146
3920
0001
24
7,11
6801
5560
0001
3 1,
3999
2654
6580
000
Med
borg
arpl
atse
n 14
00,0
0000
0000
0000
00
0,73
0662
6135
9400
0 16
63,2
0461
5860
0001
04
- 1755
,683
5218
1000
0002
0,
1693
8346
3295
000
18,8
6878
9803
7000
00
2107
,190
5180
3000
0021
- 26
3,20
4615
8620
0002
6 42
8,40
6439
0820
0002
0 30
8,72
7002
3010
0002
7 0,
0354
3821
6278
200
18,1
2953
7039
9999
99
318,
4652
0699
0000
013
- 0,61
4380
6251
5200
0
Tallk
roge
n
400,
0000
0000
0000
000
0,52
6711
2725
9100
0 64
9,72
5466
6950
0002
3 - 14
,515
3592
9249
9999
0,
1561
6640
0300
000
45,2
5542
3535
7999
99
298,
7209
4798
6999
988
- 249,
7254
6669
4999
994
450,
4884
8574
8000
016
329,
0423
2123
2000
006
0,04
5129
6256
1990
0 16
,391
0439
4170
0001
25
5,44
9955
0099
9999
6 - 0,
5543
4372
8187
000
Gam
la st
an
800,
0000
0000
0000
000
0,67
3663
1135
7700
0 79
0,78
7086
5459
9995
5 - 14
31,9
9442
6779
9999
12
0,16
9291
2379
9400
0 12
,492
2292
6829
9999
18
23,7
6938
5849
9999
35
9,21
2913
4544
2000
1 43
7,54
1984
1629
9999
5 30
4,73
4703
2220
0000
7 0,
0335
8324
7499
600
17,5
6388
4704
1999
99
297,
3605
8746
6000
027
0,02
1056
0672
7560
0
77
Kar
lapl
an
900,
0000
0000
0000
000
0,63
3848
9631
5200
0 92
4,83
6301
3580
0001
4 - 11
62,3
7617
6369
9999
39
0,13
3946
4871
9400
0 11
,939
4472
4340
0000
16
94,3
3569
8200
0000
24
- 24,8
3630
1357
6999
98
445,
7931
1463
9000
009
312,
8176
5391
5999
983
0,03
2100
7715
1230
0 15
,657
0050
0050
0000
31
8,37
2722
7300
0002
1 - 0,
0557
1261
7674
300
Ensk
ede
Går
d
300,
0000
0000
0000
000
0,60
5578
7463
7700
0 48
9,28
8469
8400
0000
5 - 84
4,02
8866
6059
9995
1 0,
1587
0331
7077
000
46,8
4500
3187
8999
98
1082
,034
9362
1000
0069
- 18
9,28
8469
8400
0000
5 44
6,72
3224
0189
9998
6 33
8,01
8685
3100
0002
3 0,
0462
9317
7779
000
19,8
7806
8332
1000
00
265,
1734
5975
6000
000
- 0,42
3726
5037
1100
0
Soln
a str
and
300,
0000
0000
0000
000
0,61
2393
1628
0300
0 64
8,81
7995
2550
0002
8 - 52
9,72
6435
2170
0001
7 0,
1694
9687
3637
000
45,2
9474
0153
1000
01
685,
7911
4122
1999
965
- 348,
8179
9525
5000
028
437,
9086
1352
1000
007
306,
8242
9952
6000
004
0,03
7928
3675
5310
0 16
,899
9365
1989
9999
25
8,25
1410
0579
9997
6 - 0,
7965
5431
4038
000
Rågs
ved
800,
0000
0000
0000
000
0,38
5590
7901
4000
0 69
2,39
2330
5570
0003
7 - 50
,546
3737
6310
0000
0,
1968
1659
4089
000
54,2
2977
8928
9999
98
289,
7344
6014
2999
978
107,
6076
6944
3000
006
424,
9341
1587
2000
007
335,
0828
6672
4999
974
0,06
0802
0519
0810
0 28
,119
5959
5369
9999
26
6,29
0860
4630
0000
1 0,
2532
3377
3011
000
Hal
lonb
erge
n
1000
,000
0000
0000
0000
0,
6232
1792
5670
000
904,
8147
5676
5999
959
- 171,
8351
6389
1999
997
0,14
6820
5453
0700
0 50
,659
0813
2620
0003
36
6,86
5465
8859
9998
1 95
,185
2432
3360
0005
43
6,83
1124
9980
0000
7 27
4,83
2352
8270
0002
2 0,
0393
3546
2286
500
10,5
6006
6075
2000
00
217,
6805
6353
3999
987
0,21
7899
4073
1500
0
Skär
mar
brin
k
1200
,000
0000
0000
0000
0,
6824
9961
6055
000
490,
8474
4452
5000
014
- 952,
2083
0909
5000
004
0,16
5130
8622
9700
0 38
,217
0090
7809
9998
12
10,1
4945
7889
9998
94
709,
1525
5547
4999
986
442,
6217
6064
9999
999
322,
6649
1669
9999
992
0,04
1486
0674
0940
0 14
,252
4813
0350
0000
25
9,33
7550
4730
0002
2 1,
6021
6378
5250
000
Hju
lsta
300,
0000
0000
0000
000
0,65
6582
7686
1200
0 66
8,82
1862
4840
0000
7 - 28
,399
4100
8560
0000
0,
1153
8247
4218
000
53,9
4159
6927
6999
97
309,
0154
8690
9999
993
- 368,
8218
6248
4000
007
449,
1269
4977
5000
014
289,
1951
1787
2000
026
0,05
9681
4218
9760
0 11
,961
2138
2360
0000
22
5,75
2328
8769
9999
7 - 0,
8211
9735
3374
000
78
Zink
ensd
amm
800,
0000
0000
0000
000
0,72
7431
8228
7500
0 12
78,6
9778
5540
0000
41
- 1824
,935
6262
5999
9935
0,
1692
9406
1539
000
15,7
4887
3325
9000
00
2208
,873
5661
3000
0199
- 47
8,69
7785
5370
0001
6 45
3,41
4195
4479
9998
7 31
1,93
2060
6829
9997
5 0,
0360
1700
6403
100
17,8
0372
9140
6000
02
304,
4702
6332
3999
973
- 1,05
5762
6786
8000
0
Hus
by
1100
,000
0000
0000
0000
0,
6960
3267
8910
000
800,
3932
9409
2000
019
1,24
6883
8238
9000
0 0,
1247
2557
1279
000
51,6
4355
6555
2000
00
265,
6449
5158
6999
980
299,
6067
0590
7999
981
445,
9346
5890
9000
007
277,
2202
1380
1999
989
0,05
1667
3017
2950
0 10
,781
8670
1029
9999
21
9,15
2984
3069
9999
7 0,
6718
6234
5575
000
Tele
fonp
lan
2400
,000
0000
0000
0000
0,
6738
9108
7019
000
1192
,525
7036
0000
0043
- 12
01,2
0713
8450
0000
02
0,14
4889
4850
4300
0 86
,226
7400
1020
0004
14
68,0
9132
9540
0000
61
1207
,474
2963
9999
9957
43
9,48
0338
3060
0002
1 37
8,18
5425
8760
0001
1 0,
0687
5988
5675
700
38,3
4099
8681
9999
99
324,
1352
2367
1000
006
2,74
7504
7030
7000
0
Band
hage
n
900,
0000
0000
0000
000
0,44
3131
0108
7000
0 85
7,29
2831
5550
0002
1 - 13
8,05
5238
3650
0000
8 0,
1623
2318
8408
000
52,7
6046
0378
2000
01
406,
0931
9941
5000
015
42,7
0716
8444
4999
99
451,
8905
2517
3000
015
324,
4377
3982
2000
026
0,05
0014
3938
6630
0 22
,226
1155
3370
0000
25
3,83
0076
7920
0000
0 0,
0945
0777
5811
700
Gub
bäng
en
600,
0000
0000
0000
000
0,48
9210
6577
8500
0 72
7,09
8544
6769
9995
0 28
,440
2609
9499
9999
0,
1600
7224
0360
000
44,8
1457
2764
8999
99
262,
8098
4586
0999
985
- 127,
0985
4467
7000
007
453,
9608
0624
6000
004
324,
6903
8653
6999
995
0,04
5873
6990
2620
0 17
,517
1595
4050
0000
25
2,17
3323
1860
0000
5 - 0,
2799
7691
1944
000
Thor
ildsp
lan
800,
0000
0000
0000
000
0,67
2272
2915
1200
0 59
9,54
2704
4230
0003
2 - 13
58,2
8803
5799
9999
89
0,19
6852
2368
6100
0 10
,595
8088
7640
0000
16
65,5
7359
2319
9999
89
200,
4572
9557
6999
996
430,
2004
1330
8999
998
291,
3071
0110
0000
011
0,03
5823
0731
2480
0 18
,073
3990
5430
0001
26
2,06
2924
0740
0002
3 0,
4659
6258
2498
000
Väl
lingb
y
1900
,000
0000
0000
0000
0,
6598
2285
4302
000
1509
,026
2115
9999
9897
11
,796
1738
9560
0001
0,
0622
6471
1720
000
70,5
2534
4618
3999
98
304,
0958
6814
0999
983
390,
9737
8839
9999
989
395,
4961
3839
5999
992
336,
6114
3934
5999
997
0,07
6416
8117
8450
0 24
,190
6863
4670
0002
28
7,36
8424
7090
0001
0 0,
9885
6537
5090
000
79
Skar
pnäc
k
1000
,000
0000
0000
0000
0,
6776
3472
1143
000
908,
0563
9971
0000
051
61,1
5796
1356
1999
98
0,14
8237
3656
1600
0 48
,310
1112
8350
0000
22
1,05
4001
2890
0001
3 91
,943
6002
9000
0006
35
1,34
6848
8149
9998
7 32
8,17
9598
2560
0002
0 0,
0434
6479
2483
900
11,8
0147
0020
8000
00
261,
0736
9359
6999
988
0,26
1688
9851
1600
0
Kist
a
1200
,000
0000
0000
0000
0,
6953
1519
2401
000
1525
,355
9628
9000
0001
8,
3630
0449
4900
000
0,12
2231
0578
5800
0 51
,474
1540
4899
9999
25
7,83
3828
3950
0001
2 - 32
5,35
5962
8859
9998
6 41
6,27
3826
9339
9999
9 27
3,42
6210
3629
9999
6 0,
0499
8328
9676
600
10,7
0450
2726
0999
99
217,
2806
7727
4999
988
- 0,78
1591
2071
2100
0
Abr
aham
sber
g
1000
,000
0000
0000
0000
0,
6648
4845
6624
000
737,
5971
1067
3999
964
- 761,
2060
9257
0999
999
0,16
3969
8540
6800
0 47
,040
3631
7269
9998
10
11,8
6969
1500
0000
44
262,
4028
8932
5999
979
447,
0056
4344
7000
011
323,
8156
8922
5000
028
0,04
5418
9370
2260
0 25
,181
6400
2360
0000
30
5,48
7065
3670
0001
4 0,
5870
2366
1051
000
Kun
gsträ
dgår
den
50
0,00
0000
0000
0000
0 0,
6535
3614
4873
000
730,
9420
4780
6000
005
- 1302
,544
7850
4000
0079
0,
1605
6764
8278
000
12,4
5033
4995
1000
00
1727
,911
7481
9000
0026
- 23
0,94
2047
8060
0000
5 39
7,40
2392
7270
0000
6 30
4,60
5745
1330
0002
8 0,
0325
5361
0284
600
17,3
0750
8349
3999
99
296,
0624
4896
9000
002
- 0,58
1128
9816
8900
0
Björ
khag
en
700,
0000
0000
0000
000
0,72
9604
4155
8700
0 70
5,39
6043
6690
0005
1 - 24
1,57
3862
9919
9998
8 0,
1488
3457
9705
000
44,6
2654
2553
0000
00
520,
3326
6072
3000
004
- 5,39
6043
6691
9000
0 45
3,86
0406
3989
9999
9 35
0,48
5030
2309
9999
6 0,
0418
4146
5841
900
11,6
7542
0836
2000
01
290,
4063
0701
8000
007
- 0,01
1889
2143
7760
0
Hor
n stu
ll
2100
,000
0000
0000
0000
0,
7439
3848
3300
000
1673
,435
7852
0999
9949
- 18
85,8
3616
2900
0000
90
0,16
2963
1455
8400
0 22
,274
1148
1980
0001
22
71,6
6286
1010
0001
42
426,
5642
1478
9000
005
408,
9640
6381
0000
027
313,
8206
6483
1999
977
0,03
7876
1332
4010
0 18
,419
0954
8170
0001
29
2,80
8379
7990
0002
2 1,
0430
3593
5270
000
Vår
by g
ård
500,
0000
0000
0000
000
0,13
1147
0483
2900
0 77
8,88
0509
5349
9998
7 38
1,79
8526
4590
0001
6 0,
1639
6689
2014
000
37,9
3137
9458
0000
02
- 18,5
2129
1410
0000
00
- 278,
8805
0953
4999
987
440,
7594
5204
9000
004
379,
5480
4559
2999
983
0,11
4627
8034
8600
0 29
,751
9774
2290
0001
28
2,51
1840
2680
0001
4 - 0,
6327
2723
5316
000
80
Rådm
ansg
atan
1500
,000
0000
0000
0000
0,
6310
6265
7970
000
1396
,527
9079
1000
0067
- 11
72,9
5668
2539
9999
74
0,16
2721
8904
1600
0 11
,366
6487
9270
0000
15
86,2
2785
7269
9999
59
103,
4720
9209
0000
004
426,
0437
5611
6999
987
294,
2541
4106
1999
974
0,03
2707
1739
5800
0 16
,964
1154
6390
0001
26
2,03
6039
1780
0001
0 0,
2428
6728
9108
000
Mör
by
cent
rum
20
00,0
0000
0000
0000
00
0,61
7217
1970
3900
0 18
41,6
0097
8510
0000
04
- 256,
0249
0742
2000
013
0,15
5824
1897
0000
0 51
,183
4944
6620
0003
41
4,17
0735
6719
9997
8 15
8,39
9021
4940
0001
0 33
1,65
4984
6269
9997
6 29
3,47
4771
6530
0000
5 0,
0335
9343
4974
600
11,3
1187
9601
5999
99
220,
1978
4099
8000
004
0,47
7601
8116
3100
0
Axe
lsber
g
500,
0000
0000
0000
000
0,70
1824
1319
6900
0 73
2,52
1932
0649
9998
7 - 12
03,4
9918
2699
9998
92
0,20
1354
8960
4400
0 64
,868
8337
7530
0004
14
19,0
4542
4879
9999
28
- 232,
5219
3206
4999
987
437,
1278
8204
8999
993
378,
5044
8239
9999
972
0,05
3810
2228
0320
0 25
,972
3701
3750
0000
31
8,04
6321
4029
9997
9 - 0,
5319
3113
8722
000
Dan
dery
ds
sjukh
us
2800
,000
0000
0000
0000
0,
6002
9412
3682
000
2923
,118
3183
8999
9786
- 35
7,32
2433
7139
9999
9 0,
1545
8886
4297
000
48,7
7456
0385
5999
97
530,
0409
5977
4000
044
- 123,
1183
1839
2999
996
282,
1961
6954
2999
996
292,
5798
3684
9999
992
0,03
3622
3588
5030
0 11
,584
1567
2360
0000
22
8,62
2733
2169
9999
0 - 0,
4362
8628
4793
000
Alb
y
1000
,000
0000
0000
0000
0,
0596
4531
8732
500
858,
5034
1643
9000
034
625,
4623
2127
3000
043
0,10
2040
2304
8600
0 21
,318
0274
0730
0001
- 12
0,33
4074
8309
9999
5 14
1,49
6583
5609
9999
4 33
7,72
3901
2659
9998
4 40
8,00
7976
7569
9999
4 0,
1391
3447
3725
000
35,7
3145
4396
6000
00
299,
4130
0102
5000
028
0,41
8971
1863
1600
0
Näc
kros
en
1100
,000
0000
0000
0000
0,
6190
9293
0887
000
1246
,364
9713
4000
0011
- 36
9,19
6412
6529
9998
2 0,
1731
0486
8114
000
48,1
7722
8011
6000
00
485,
4355
3643
6000
007
- 146,
3649
7134
4999
987
411,
4523
8792
9999
986
294,
7466
3992
7999
979
0,03
5411
2557
8960
0 10
,775
1627
6619
9999
23
0,57
3022
9140
0000
6 - 0,
3557
2760
2119
000
Rink
eby
1000
,000
0000
0000
0000
0,
6547
5128
0871
000
975,
7750
8361
6999
950
- 39,0
6055
8956
5999
98
0,12
0493
8562
6300
0 52
,328
9833
7889
9999
31
0,62
4976
5349
9997
3 24
,224
9163
8289
9998
39
9,09
5526
1569
9999
5 27
0,29
0085
6549
9998
4 0,
0510
4468
1728
800
10,9
8072
6414
6999
99
215,
9309
9165
6999
993
0,06
0699
5438
3140
0
81
Riss
ne
900,
0000
0000
0000
000
0,62
0787
5226
8700
0 12
30,2
9521
3690
0000
82
- 97,2
0022
8937
2999
97
0,12
0457
8467
5100
0 52
,790
4137
8190
0000
36
2,18
2452
2000
0000
0 - 33
0,29
5213
6910
0001
5 40
9,81
2099
3620
0002
6 26
6,75
9528
0950
0000
7 0,
0474
7910
4357
100
11,5
6691
2494
7000
00
216,
3102
6427
8000
005
- 0,80
5967
4524
1400
0
Gär
det
1100
,000
0000
0000
0000
0,
6301
5015
7918
000
678,
3231
5674
3000
027
- 1139
,863
0316
3999
9917
0,
1285
0585
4736
000
10,6
4790
8095
4000
00
1700
,510
9811
4000
0013
42
1,67
6843
2569
9997
3 39
0,47
9263
3569
9997
9 31
1,14
5876
2949
9999
4 0,
0322
7603
0602
900
14,9
0758
9906
9000
00
316,
9543
3983
8000
010
1,07
9895
6124
6000
0
Häs
selb
y gå
rd
900,
0000
0000
0000
000
0,66
2381
8285
0600
0 68
2,63
5977
3489
9997
2 4,
8295
2518
3890
000
0,07
7235
6246
3270
0 71
,181
7423
5610
0003
28
8,31
8472
1829
9998
3 21
7,36
4022
6509
9999
9 42
5,31
2635
8040
0002
4 34
5,14
5331
0790
0001
6 0,
0814
5097
3423
100
24,5
5564
1516
9000
00
292,
6793
0908
3000
021
0,51
1068
8099
8300
0
Rådh
uset
1000
,000
0000
0000
0000
0,
6669
0630
9102
000
1241
,950
7750
6000
0069
- 14
11,7
8048
4769
9999
30
0,17
7517
4977
0600
0 10
,201
7958
3420
0000
17
88,2
9384
2309
9999
45
- 241,
9507
7505
8999
994
454,
9744
5425
9000
026
298,
3438
1622
0000
008
0,03
4093
2500
1950
0 16
,950
5773
4900
0000
26
5,58
6090
4909
9999
3 - 0,
5317
8980
2249
000
Kris
tineb
erg
1100
,000
0000
0000
0000
0,
6756
6277
1194
000
1355
,970
4177
9000
0056
- 13
01,2
9096
5570
0000
26
0,19
6510
0912
4600
0 16
,198
4192
4629
9999
15
87,2
0667
1319
9999
41
- 255,
9704
1778
9000
010
414,
8368
9597
9000
019
297,
6182
8500
6000
008
0,03
6367
2052
8780
0 19
,293
9946
9530
0000
27
9,75
7685
3620
0001
8 - 0,
6170
3869
7065
000
Hög
dale
n
1000
,000
0000
0000
0000
0,
4150
6863
9527
000
1387
,211
2569
5000
0006
- 73
,290
6119
0359
9995
0,
1731
8865
6008
000
52,3
1109
7876
2000
02
337,
6822
8865
8000
004
- 387,
2112
5695
1999
985
353,
5193
9480
9999
994
326,
7589
0495
9000
006
0,05
3290
4622
1840
0 24
,619
6456
9150
0001
25
6,64
2442
4090
0001
2 - 1,
0953
0414
0690
000
Alv
ik
2100
,000
0000
0000
0000
0,
6873
7529
6342
000
2056
,946
2953
3999
9779
- 11
87,3
3726
1220
0000
73
0,19
1292
0336
7600
0 28
,322
7147
0499
9999
14
39,1
4796
0629
9999
43
43,0
5370
4661
4999
99
254,
5595
8017
8000
004
324,
3226
4839
4999
987
0,04
0997
1690
5640
0 23
,351
2057
0659
9998
31
3,91
6396
5960
0002
7 0,
1691
3016
8393
000
82
Blås
ut
600,
0000
0000
0000
000
0,65
8486
8410
6900
0 67
9,64
8367
6120
0001
5 - 37
7,39
5860
0409
9999
2 0,
1541
8316
3884
000
43,5
1560
0402
0999
99
644,
7328
8965
6999
987
- 79,6
4836
7612
0000
01
453,
7842
4634
4999
985
337,
5799
8313
6000
010
0,04
3115
4372
1540
0 13
,329
4407
5790
0000
26
7,06
6740
0219
9997
6 - 0,
1755
2034
5304
000
Höt
orge
t
500,
0000
0000
0000
000
0,64
1820
3934
3800
0 78
3,05
9456
1149
9998
9 - 12
49,3
9124
5540
0000
00
0,16
3826
5666
1600
0 10
,698
4132
9400
0000
16
68,4
7237
2149
9999
56
- 283,
0594
5611
4999
989
443,
4912
4580
5000
005
299,
6939
7204
6999
988
0,03
2822
8658
5820
0 17
,213
3112
7699
9999
27
1,15
5171
8399
9998
0 - 0,
6382
5263
4731
000
Baga
rmos
sen
1100
,000
0000
0000
0000
0,
7028
2011
2134
000
665,
0614
7206
1000
018
10,9
301 5
9599
0000
00
0,14
7129
1288
1800
0 48
,083
5587
7849
9999
26
8,11
9811
1329
9998
5 43
4,93
8527
9389
9998
2 45
1,06
7141
1729
9998
2 33
1,12
6773
9499
9997
2 0,
0430
4531
6512
800
11,6
5380
7366
6000
01
264,
2145
1037
3999
985
0,96
4243
4312
7400
0
Sand
sbor
g
700,
0000
0000
0000
000
0,61
0664
6847
7800
0 39
0,88
8249
0610
0001
0 - 25
1,46
7982
5960
0001
3 0,
1530
9023
1237
000
45,4
1023
8499
0000
02
518,
7744
4517
4999
983
309,
1117
5093
8999
990
442,
4650
0803
7000
018
331,
3724
8385
3999
995
0,04
3023
1609
0120
0 13
,634
4477
5620
0000
25
9,10
3352
1140
0001
8 0,
6986
1287
4067
000
Sätra
800,
0000
0000
0000
000
0,34
8179
8779
3700
0 60
4,09
9005
7770
0000
2 - 59
,292
2080
8819
9999
0,
2497
7027
6870
000
60,3
2399
6948
5999
98
255,
8775
8189
7999
988
195,
9009
9422
2999
998
415,
5800
6751
4000
007
402,
8469
3567
8000
023
0,09
4967
3032
3920
0 25
,009
2116
0180
0001
29
9,47
7515
3229
9997
7 0,
4713
9169
9306
000
Fars
ta
1200
,000
0000
0000
0000
0,
4901
3891
4833
000
1079
,601
8724
2000
0063
13
9,43
8492
1330
0001
1 0,
1622
1842
7919
000
44,4
4445
8194
2000
03
162,
2078
8370
8999
987
120,
3981
2758
0999
997
365,
2851
2274
7999
992
327,
5636
3649
8999
983
0,04
7821
8799
7800
0 16
,984
8355
7320
0002
25
7,39
8017
3460
0001
7 0,
3296
0041
3714
000
Fruä
ngen
1800
,000
0000
0000
0000
0,
4133
1892
3033
000
1920
,731
1707
1999
9909
- 33
4,73
1365
2770
0000
9 0,
2301
4673
1275
000
61,1
0297
2580
3999
99
577,
1091
5171
5000
053
- 120,
7311
7071
9999
994
288,
0024
2557
2999
982
411,
6713
5729
3000
028
0,07
8533
4554
4320
0 32
,844
2085
6860
0003
34
4,59
2599
4550
0002
7 - 0,
4192
0192
3317
000
83
Duv
bo
600,
0000
0000
0000
000
0,60
0329
5026
8200
0 10
06,5
0881
6740
0000
43
- 212,
4035
3140
9000
010
0,13
9722
9591
9500
0 50
,779
1371
1990
0000
42
6,02
5188
9090
0000
8 - 40
6,50
8816
7420
0002
2 41
8,86
9962
2500
0001
5 28
0,50
7302
1919
9999
7 0,
0411
2388
6531
500
11,4
0779
2552
6000
00
224,
1651
4790
9000
012
- 0,97
0489
3006
8400
0
Soln
a ce
ntru
m
1600
,000
0000
0000
0000
0,
6265
3293
3903
000
1975
,175
3278
5999
9925
- 55
0,60
6895
9020
0001
9 0,
1897
8328
3619
000
41,4
3628
8596
6000
01
649,
5095
0167
5000
024
- 375,
1753
2785
8000
003
396,
7354
9110
6999
973
295,
2704
9632
6999
989
0,03
4409
7145
8860
0 11
,598
4727
6760
0001
24
3,64
8470
2879
9999
9 - 0,
9456
5607
6322
000
Ode
npla
n
2100
,000
0000
0000
0000
0,
6266
7199
8870
000
3762
,555
6837
3000
0055
- 11
01,5
7395
8929
9998
90
0,17
0993
6327
4600
0 13
,334
9591
8500
0001
14
60,5
8739
4409
9998
88
- 1662
,555
6837
3000
0055
37
0,38
5491
9139
9999
9 28
4,42
8368
2640
0002
8 0,
0324
9652
7623
000
15,8
0663
8362
1999
99
249,
8354
8468
6000
001
- 4,48
8717
0799
9000
0
Öste
rmal
msto
rg 80
0,00
0000
0000
0000
0 0,
6386
0300
1678
000
1370
,642
7963
5999
9920
- 12
26,2
1551
8980
0000
70
0,15
2417
3230
0700
0 11
,594
0100
0600
0000
16
88,5
6231
3709
9998
98
- 570,
6427
9636
3000
002
433,
8193
4177
1999
973
301,
3096
4610
5000
013
0,03
2082
7613
5600
0 16
,801
8550
0700
0000
28
6,63
6626
6829
9997
4 - 1,
3153
9270
2480
000
Stad
shag
en
700,
0000
0000
0000
000
0,66
3662
6874
1700
0 10
44,2
8762
7449
9999
45
- 1242
,925
2897
7000
0063
0,
2079
8913
4913
000
7,58
5860
7325
2000
0 15
09,6
6276
4420
0000
30
- 344,
2876
2744
9000
013
435,
5795
9326
0000
024
293,
1409
0029
2000
026
0,03
5687
3840
5960
0 18
,536
1535
1319
9999
26
5,82
0785
3900
0002
6 - 0,
7904
1266
5736
000
Tekn
iska
högs
kola
n 20
00,0
0000
0000
0000
00
0,61
9785
4413
2000
0 28
07,5
3364
7090
0001
59
- 1105
,081
3424
4999
9966
0,
1480
4892
6099
000
12,5
9660
5684
3000
00
1564
,784
5715
5000
0010
- 80
7,53
3647
0910
0003
4 36
3,01
7295
6029
9998
0 29
7,49
7347
9550
0000
9 0,
0323
8096
1983
300
15,2
7068
8535
6000
00
274,
1065
4226
7999
998
- 2,22
4504
6086
5000
0
Sund
bybe
rg
1500
,000
0000
0000
0000
0,
6033
4474
9451
000
2133
,152
3063
8999
9922
- 36
4,34
1549
9270
0001
7 0,
1601
1580
1319
000
48,4
7010
3599
9000
00
528,
2879
5632
1000
024
- 633,
1523
0639
3000
003
398,
8174
0388
1000
018
293,
3505
3199
1000
025
0,03
7341
6437
7150
0 12
,172
3009
0330
0000
23
4,52
3587
2680
0000
0 - 1,
5875
7441
4330
000
84
Åke
shov
500,
0000
0000
0000
000
0,68
0543
2531
0400
0 54
6,91
7333
4800
0002
5 - 25
4,89
8140
0220
0000
7 0,
0753
7058
5746
400
64,2
8512
4321
4000
05
603,
5274
3304
3999
963
- 46,9
1733
3480
2000
00
431,
4255
9145
0000
013
346,
4732
1697
0000
010
0,06
2147
5152
3650
0 26
,395
4545
7890
0001
30
2,59
3018
8629
9999
8 - 0,
1087
4953
7371
000
Sved
myr
a
400,
0000
0000
0000
000
0,50
9545
3465
7100
0 62
4,60
9137
2839
9998
5 - 22
7,69
2789
7520
0001
0 0,
1536
8670
3729
000
49,9
0744
6041
9000
02
492,
7250
1695
6999
980
- 224,
6091
3728
4000
013
441,
4444
4196
1999
985
332,
8632
1556
6000
008
0,04
6850
7758
0740
0 19
,596
2860
4459
9999
25
8,80
5065
4589
9999
1 - 0,
5088
0499
5451
000
Fars
ta st
rand
700,
0000
0000
0000
000
0,49
0912
6413
8900
0 10
42,8
1648
3920
0001
10
162,
0357
9170
5999
998
0,16
2815
5093
1900
0 44
,386
0512
4059
9997
14
1,64
2460
9059
9999
7 - 34
2,81
6483
9170
0002
8 35
9,52
7548
2700
0001
1 32
9,47
4170
6849
9999
0 0,
0489
0409
1604
500
17,1
8263
5284
3000
02
259,
7077
4174
1000
007
- 0,95
3519
3772
1100
0
Sock
enpl
an
500,
0000
0000
0000
000
0,55
1103
1912
0400
0 57
6,64
8072
4079
9996
2 - 39
9,78
9837
1360
0001
7 0,
1517
4346
0933
000
49,8
9824
9109
7999
98
654,
2108
4713
4999
995
- 76,6
4807
2408
3999
97
452,
2705
9992
4000
010
339,
9483
9584
2000
025
0,04
7066
8948
8420
0 19
,695
9311
5889
9999
26
4,67
2953
3500
0000
0 - 0,
1694
7392
2075
000
Bred
äng
1100
,000
0000
0000
0000
0,
4591
9746
2487
000
1069
,660
1391
5999
9972
- 26
6,68
6094
0280
0001
4 0,
2368
8281
1407
000
67,2
2863
2306
5000
03
466,
9798
6317
5999
981
30,3
3986
0840
6999
98
438,
5181
8342
6999
997
399,
5801
4991
3000
014
0,07
6043
9270
6010
0 26
,149
0664
2549
9999
32
3,58
0326
1639
9998
5 0,
0691
8723
5529
400
Skog
skyr
kogå
rde
n 40
0,00
0000
0000
0000
0 0,
5979
8818
1753
000
740,
5866
0755
7999
969
- 122,
1926
1039
9000
003
0,15
2518
2304
6100
0 46
,017
8819
4209
9997
39
6,35
3821
9480
0001
8 - 34
0,58
6607
5580
0002
6 45
5,98
3534
8350
0000
0 32
5,35
4797
7759
9999
8 0,
0425
0062
9301
500
12,6
9565
2083
1000
01
254,
6750
4404
4999
993
- 0,74
6927
4250
9100
0
Mid
som
mar
kran
sen
900,
0000
0000
0000
000
0,75
9469
0408
3600
0 89
5,37
4784
7280
0000
9 - 17
92,3
5029
5349
9998
97
0,16
1764
1929
6100
0 60
,030
8490
4840
0000
20
75,1
0774
7869
9999
12
4,62
5215
2716
2000
0 37
8,33
1517
4219
9999
0 35
0,10
5691
7959
9997
4 0,
0576
0268
7137
600
31,2
8259
5881
9999
99
301,
6725
1084
2999
998
0,01
2225
2972
8190
0
85
Tens
ta
900,
0000
0000
0000
000
0,65
7087
4295
7100
0 69
6,91
5464
3509
9996
9 - 23
,631
8725
2890
0001
0,
1147
7816
2636
000
53,2
4488
5924
3000
00
305,
3408
5215
6999
972
203,
0845
3564
9000
003
450,
8460
6403
9999
987
282,
9489
5691
4000
007
0,05
7255
9493
7260
0 11
,527
7649
1310
0000
22
2,11
5710
5489
9999
9 0,
4504
5205
4143
000
Stor
a m
osse
n
500,
0000
0000
0000
000
0,67
5341
4092
0000
0 60
8,23
1827
7279
9998
5 - 97
4,06
8031
5880
0005
3 0,
1795
7992
0097
000
38,1
8074
5238
9999
97
1213
,542
2344
7999
9934
- 10
8,23
1827
7279
9999
9 43
7,46
2944
1029
9998
4 32
8,36
5425
9730
0001
5 0,
0432
4415
7140
900
24,8
0176
2267
0999
99
317,
4114
2095
8999
997
- 0,24
7407
9900
6500
0
Stad
ion
700,
0000
0000
0000
000
0,62
7413
356 5
8000
0 11
43,1
2470
2029
9999
80
- 1159
,758
7190
5999
9976
0,
1422
9083
6972
000
11,1
5197
2760
3000
00
1662
,575
2274
1999
9919
- 44
3,12
4702
0339
9999
5 45
3,14
8249
9260
0000
5 30
3,65
9489
2690
0000
5 0,
0321
8459
4652
900
16,1
5412
1335
5000
01
292,
6607
6637
9999
984
- 0,97
7880
2016
9300
0
Skär
holm
en
800,
0000
0000
0000
000
0,22
7901
5211
0800
0 12
82,1
8722
9849
9999
94
155,
3812
8857
8000
010
0,22
2021
7974
0200
0 48
,276
7527
2039
9998
10
6,44
9029
9689
9999
4 - 48
2,18
7229
8459
9998
0 34
6,77
9366
8300
0001
5 38
1,48
6338
7930
0001
6 0,
1037
9144
3127
000
26,2
8677
2779
7000
01
277,
8819
8156
9999
994
- 1,39
0472
6635
1000
0
Vår
berg
1100
,000
0000
0000
0000
0,
1838
5106
6737
000
516,
8056
5169
0999
980
286,
6942
4470
4000
027
0,17
6210
4362
1100
0 48
,338
9476
3930
0002
24
,535
4621
3039
9999
58
3,19
4348
3090
0002
0 39
4,46
0731
1920
0002
6 37
3,45
9934
3130
0001
9 0,
1079
4602
0406
000
26,8
0658
2163
8000
02
277,
7871
7704
3999
975
1,47
8459
8369
2000
0
Häs
selb
y str
and
700,
0000
0000
0000
000
0,67
0841
9894
4400
0 70
1,94
2654
3289
9999
3 - 0,
7379
2501
2073
000
0,07
6731
1265
0610
0 72
,859
2613
1710
0007
29
6,68
3962
1980
0001
7 - 1,
9426
5432
8530
000
446,
4990
4573
2000
013
347,
3732
6271
2999
974
0,08
3202
0862
2550
0 24
,716
3846
3420
0001
29
4,69
8825
5929
9998
1 - 0,
0043
5085
8858
710
Häg
erste
nsås
en 10
00,0
0000
0000
0000
00
0,60
4943
0850
2900
0 79
9,30
5372
4799
9996
0 - 93
8,16
0146
3660
0004
9 0,
1876
2104
2078
000
70,8
1764
7676
5000
02
1177
,622
9206
9999
9895
20
0,69
4627
5200
0001
2 43
3,81
6594
7190
0002
2 37
4,89
4866
3990
0002
3 0,
0594
2948
9224
800
31,1
7930
4113
4999
99
320,
5579
9966
0999
997
0,46
2625
5195
4700
0
86
Väs
tra sk
ogen
1300
,000
0000
0000
0000
0,
6381
1952
2340
000
607,
1011
8954
5000
011
- 903,
6359
0501
0999
977
0,20
3308
2860
6200
0 20
,182
3890
6699
9999
10
92,5
1578
3430
0000
56
692,
8988
1045
4999
989
438,
5735
5603
2999
988
290,
2462
1521
0999
992
0,03
4709
1295
1000
0 17
,071
8468
3859
9999
26
8,55
8305
3949
9999
1 1,
5798
9190
3930
000
Mar
iato
rget
1900
,000
0000
0000
0000
0,
7298
4181
2448
000
1749
,952
0005
3999
9972
- 18
20,1
8460
3630
0000
83
0,17
1143
3191
6100
0 16
,000
8551
3839
9999
21
83,7
9746
3719
9999
96
150,
0479
9946
2999
996
412,
3982
4164
8000
010
314,
9778
2482
9999
975
0,03
5709
9262
1360
0 18
,194
3051
9879
9999
32
0,59
5529
0379
9999
5 0,
3638
4248
1149
000
Blac
kebe
rg
500,
0000
0000
0000
000
0,71
3217
9592
6100
0 56
0,16
9099
9219
9998
7 - 15
,741
5758
0410
0000
0,
0259
4173
4532
300
73,9
7872
0022
2999
96
397,
8089
4015
5000
016
- 60,1
6909
9921
6000
01
441,
7452
7032
3000
000
352,
3297
0375
0000
022
0,07
9821
5852
8990
0 25
,816
3676
7079
9998
30
3,00
2354
1240
0002
1 - 0,
1362
0768
3396
000
Råck
sta
1100
,000
0000
0000
0000
0,
6730
2292
9519
000
617,
2586
8971
0999
988
- 5,94
3655
5245
4000
0 0,
0493
8635
7728
300
71,1
7354
0639
6999
93
347,
6524
5800
9999
975
482,
7413
1028
9000
012
448,
8337
0145
4999
982
334,
7030
6251
4999
999
0,07
3959
8713
5640
0 24
,443
3155
5560
0002
28
7,84
0319
7650
0000
4 1,
0755
4604
0160
000
Asp
udde
n
1000
,000
0000
0000
0000
0,
7684
5792
6444
000
1035
,292
8544
7000
0066
- 16
74,4
9154
7739
9999
87
0,15
6571
7567
3100
0 64
,980
5852
2750
0001
19
50,5
5142
2909
9999
28
- 35,2
9285
4472
0000
02
445,
7714
6309
3000
023
344,
2991
1187
5999
984
0,05
5364
8607
7970
0 29
,795
6524
7209
9999
29
0,85
8305
4520
0002
5 - 0,
0791
7252
9859
000
Rops
ten
3800
,000
0000
0000
0000
0,
6224
5080
4688
000
2641
,425
2861
7000
0163
- 10
40,0
0865
7670
0000
48
0,11
6698
2336
8300
0 14
,575
3089
7620
0001
16
12,5
1059
4939
9999
19
1158
,574
7138
3000
0064
33
4,50
9035
8439
9998
2 30
7,06
6853
4040
0002
8 0,
0331
7570
8288
100
12,7
8810
7905
5000
00
309,
5797
3247
7999
983
3,46
3507
9764
4000
0
Nor
sbor
g
400,
0000
0000
0000
000
0,07
6203
6912
4250
0 82
1,09
5838
9940
0001
9 57
5,93
8708
7280
0005
1 0,
0730
7303
9918
000
35,1
6078
8169
2999
98
- 111,
2263
8545
4999
999
- 421,
0958
3899
4000
019
263,
6866
6031
3000
004
394,
9909
3525
0000
007
0,13
6934
3356
1500
0 32
,174
4273
2899
9997
29
9,54
6624
9489
9997
7 - 1,
5969
5541
1000
000
87
Hal
lund
a
300,
0000
0000
0000
000
0,07
1536
2799
1830
0 91
6,50
3330
0579
9996
0 57
8,17
7871
9390
0005
6 0,
0872
4887
6847
000
31,7
5572
0135
0000
01
- 110,
1951
7841
5000
001
- 616,
5033
3005
7999
960
410,
2494
8911
4000
028
397,
3459
9165
9999
981
0,13
5548
6215
5700
0 32
,854
7708
5250
0003
29
8,28
1518
9699
9997
9 - 1,
5027
5221
8870
000
Bark
arby
sta
tion
608,
0000
0000
0000
000
0,65
9954
5689
2000
0 67
6,52
6256
7959
9998
4 - 28
,110
6219
2140
0000
0,
1170
4238
7051
000
54,7
0834
7429
9000
02
306,
2784
0429
3999
984
- 68,5
2625
6795
6999
93
436,
1332
2700
1000
023
300,
3022
1471
3000
012
0,06
4554
2047
3590
0 12
,552
2520
3050
0000
23
2,08
9291
1610
0000
6 - 0,
1571
2230
2437
000
Järla
1053
,000
0000
0000
0000
0,
7460
7417
9500
000
1012
,610
5466
4000
0052
- 51
4,98
8528
3830
0001
6 0,
1437
6295
6183
000
38,8
1296
1652
6999
97
844,
6237
8142
7000
040
40,3
8945
3364
6000
01
421,
2427
6465
1000
016
351,
6087
2346
1000
011
0,03
7561
9116
7640
0 12
,036
0045
4440
0001
30
9,51
4425
9080
0002
1 0,
0958
8165
4841
100
Ham
mar
by
Kan
al
1854
,000
0000
0000
0000
0,
7529
1687
4797
000
1846
,365
3150
0999
9904
- 16
67,6
7851
3930
0000
08
0,17
2239
5555
5300
0 21
,962
9854
2690
0001
20
07,3
6610
6279
9999
40
7,63
4684
9854
4000
0 36
7,18
0894
6310
0000
1 32
7,43
2279
3669
9997
8 0,
0364
1713
1075
800
14,6
5470
3902
9000
00
330,
3433
7656
5000
028
0,02
0792
7076
1920
0
Nac
ka
Cent
rum
29
40,0
0000
0000
0000
00
0,74
0397
2393
1300
0 29
94,7
6531
7299
9998
79
- 587,
7136
1317
6999
957
0,13
8815
8096
8100
0 36
,517
0370
6800
0000
94
9,62
5133
7370
0005
6 - 54
,765
3173
0140
0003
17
2,93
8164
0100
0000
8 34
9,17
5848
0710
0001
9 0,
0366
3979
6542
300
12,2
3007
9545
8000
01
308,
2649
6292
8999
978
- 0,31
6675
7182
5300
0
Bark
arby
stade
n 265,
0000
0000
0000
000
0,68
2715
8947
7100
0 30
8,16
3415
0439
9997
6 - 6,
7022
5400
6800
000
0,12
1513
1764
9900
0 52
,664
2461
8330
0001
28
2,17
8624
5719
9999
0 - 43
,163
4150
4390
0002
41
7,66
6190
6190
0002
0 29
4,23
1243
4369
9998
9 0,
0606
9785
5560
000
11,3
5294
9671
1999
99
227,
1133
9146
5999
996
- 0,10
3344
2878
8700
0
Sofia
2026
,000
0000
0000
0000
0,
7129
1852
0532
000
1600
,625
1899
1999
9912
- 14
87,5
4512
1870
0000
03
0,16
6216
2266
0400
0 18
,791
5366
1350
0000
18
60,8
2295
3630
0000
29
425,
3748
1007
9999
975
395,
1839
1242
8999
975
296,
1536
1077
5000
004
0,03
2809
7010
4740
0 15
,133
0872
3259
9999
31
0,92
0346
6260
0002
5 1,
0763
9708
1720
000
88
Sick
la
1118
,000
0000
0000
0000
0,
7408
2602
3516
000
1024
,188
9513
7000
0041
- 62
8,53
2741
1089
9997
1 0,
1530
7304
7833
000
36,3
7045
2151
5999
99
942,
7644
5682
8999
982
93,8
1104
8633
2000
02
417,
5338
9997
9000
012
355,
5828
3736
5999
978
0,03
8007
8605
7470
0 12
,145
5675
4700
0001
31
9,67
0356
1690
0000
2 0,
2246
7887
9099
000
89
Appendix IV: GWR Prediction by ArcGIS (ArcMap)
ArcMap proposes an option while calculating the GWR model to predict dependent variables at given location as well as their respective GWR equations. The tool is very useful and the results for 2016 are presented in the table below with each respective prediction done with the OLS equation.
The way these predictions were executed and calculated by ArcMap are unclear. It was for this reason that the GWR equations for the new stations were finally determined using the prediction from the OLS equation.
TRITA TRITA-ABE-MBT-19647
www.kth.se