gpsradiooccultationdataassimilationinthearemregional...

Research ArticleGPS Radio Occultation Data Assimilation in the AREM RegionalNumerical Weather Prediction Model for Flood Forecasts

Wei Cheng 12 Youping Xu12 Zhiwu Deng2 and Chunli Gu2

1State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid DynamicsInstitute of Atmospheric Physics Chinese Academy of Sciences Beijing 100029 China2Institute of Applied Meteorology Beijing 100029 China

Correspondence should be addressed to Wei Cheng chengwmailiapaccn

Received 6 August 2018 Accepted 2 October 2018 Published 11 November 2018

Academic Editor Anthony R Lupo

Copyright copy 2018 Wei Cheng et al ampis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Based on the Backward Four-Dimensional Variational Data Assimilation (Backward-4DVar) system with the Advanced RegionalEta-coordinate Model (AREM) which is capable of assimilating radio occultation data a heavy rainfall case study is performedusing GPS radio occultation (GPS RO) data and routine GTS data on July 5 2007 ampe case study results indicate that the use ofradio occultation data after quality control can improve the quality of the analysis to be similar to that of the observations andthus have a positive effect when improving 24-hour rainfall forecasts Batch tests for 119 days from May to August during theflood season in 2009 show that only the use of GPS RO data can make positive improvements in both 24-hour and 48-hourregional rainfall forecasts and obtain a better B score for 24-hour forecasts and better TS score for 48-hour forecasts When usingradio occultation refractivity data and conventional radiosonde data the results indicate that radio occultation refractivity datacan achieve a better performance for 48-hour forecasts of light rain and heavy rain

1 Introduction

After decades of development the accuracy of numericalweather predictions has greatly improved ampe main contri-butions lie in the following two aspects first the perfection ofthe dynamic framework of the numerical model itself and therefinement of various physical processes second the highdevelopment of exploration technologies and the applicationof data assimilation technologiesampe assimilation applicationof unconventional observation data such as satellite remotesensing data plays an important role in improving the ac-curacy of numerical weather predictions especially the ac-curacy of forecasting in the Southern Hemisphere ampereforeusing various observational methods to obtain more detailedinformation on the atmospheric state developing and im-proving advanced assimilation methods to effectively utilizeall atmospheric observation information to improve thequality of the initial conditions is a critical way to improve theaccuracy of numerical weather predictions at this stage

Among the various new observation methods the globalpositioning system (GPS) and small satellite technology for

occultation detection are relied upon as new methods forobtaining atmospheric information Compared with con-ventional observations and other satellite data occultationdata have the advantage of high vertical resolutions uniformglobal coverage and weak influences from aerosols cloudsand precipitation [1] In theory the assimilation of data canimprove the vertical distribution of the physical quantityfield especially for temperature and humidity near theobservations which allows the analysis quality of the initialvalue to be improved to some extent

In the study of assimilation applications to GPS RO datarefractivity data are simpler and more feasible in the ap-plication of routine assimilations because of their simpleobservation operators and economic time-saving featuresPreviously Zou et al [2] and Kuo et al [3] performed anobservational system test in the 4DVar assimilation systemKursinski et al [4] and Poli et al [5] performed assimilationexperiments on local refractivity data under an one-dimensional variational assimilation framework Huanget al [6] used GPS RO refractivity data to test typhoonpredictions based on the WRF 3DVar assimilation system

HindawiAdvances in MeteorologyVolume 2018 Article ID 1376235 9 pageshttpsdoiorg10115520181376235

ampe results showed that the use of GPS RO refractivitydata has a positive effect on the simulation of typhoonprecipitation

ampe key point of this article is introducing the qualitycontrol scheme for the GPS RO refractivity data in theAREM-B4DVar system to evaluate the role of GPS ROrefractivity data assimilation in regional numerical weatherpredictions and through actual rainstorm cases and batchexperiments during the flood season to explore an effectivemethod for improving the forecasting ability of regionalnumerical forecast models by using GPS RO refractivitydata Furthermore the focus of this article is to providea basic theoretical basis and technical support for the de-velopment of occultation data assimilation methods in re-gional numerical forecast models and weather forecasts forshort periods of time

2 AREM-B4DVarDataAssimilationSystemandObservational Operators of GPS RO Data

21 Introduction of the AREM-B4DVar Data AssimilationSystem Four-dimensional variational data assimilation(4DVar) is one of the most promising methods for providingoptimal analyses of numerical weather predictions It is thepreferred development plan formost of the numerical weatherprediction centres in the world Based on the principle theoryof the method due to the dynamic and physical constraints ofnumerical models all observations are best fitted in the as-similation time window by the variational method and theinitial value of the optimal analysis at the beginning of theassimilation time window is obtained Under the constraint ofthe model the evolution trajectory of the analysis field withinthe assimilation time window is consistent with the actualobservation trend which allows the accuracy level of theforecast to be better improved at a future time

Developing an operational 4DVar system is a very largeproject and takes hard work It requires not only a corre-sponding tangent and adjoint model but also rigorouscorrectness and accuracy tests Given the ldquoon-offrdquo problemin the process of complex physics [7 8] these problemshave become a bottleneck that plagues the development of4DVar and limits the wide application of this method inoperational numerical models Many scholars have pro-posed numerous effective solutions to this problem Amongthem Wang and Zhao [9] proposed the concept of the 3Dmapping variation method (3DVM) by placing the initialvalue of the assimilation at the end of the assimilationwindow ampe use of mapping observations subtly avoidsthe use of an adjoint model Wang et al [10] proposeda four-dimensional variational assimilation method (calledDRP-4DVar) to reduce dimensionality projections usinghistorical sample fitting and dimensionality reduction pro-jection techniques Selecting perturbed samples that dependon the analysis time solves the problem that the backgrounderror covariance matrix is not explicitly developed in the4DVar Recently Wang et al [11] combined the advantagesof the 3DVM and DRP-4DVar proposed a backwardmapping four-dimensional variational assimilation method(Backward-4DVar referred to as B-4DVar) and established

the AREM mode of the B-4DVar system (which is termedthe AREM-B4DVar system) ampis method not only avoidstangent-linear and adjoint models but also reduces thecomputational cost of the assimilation window and becausethe initial value being generated at the end of the assimi-lation window it can also reduce the prediction error ac-cumulation throughout the assimilation window whichplays an important role in short-term and nowcastingforecasts is verified in the observational system experiments

ampe B-4DVar problem comes down to the minimizationof the cost function in the m-dimensional sample space(wherem represents the number of samples) [11] thereforethe classic 4DVar method which is defined for high di-mensions of the control variable space is implemented onthe m-dimensional sample space

xa xb + xaprime xb + Pxαa

1113957J αa( 1113857 minαisinEm

1113957J(α)

1113957J(α) 12αΤBminus1α α +

12Pyαminus 1113957yobsprime1113872 1113873

TPyαminus 1113957yobsprime1113872 1113873

⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩

1113957yobsprime Rminus1yobsprime

O RTR

(1)

ampe solution to the minimization problem mentionedabove can be expressed as

αa Bminus1α + PTy Py1113872 1113873minus1

1113957yobsprime (2)

Because the dimension of the matrix ism which is a smallnumber (under 100) calculating an inverse matrix is relativelyeasy in the above expression To slow down the un-derestimation of the B matrix and the false teleconnectionbetween the lattice variable and observed variable the Bmatrixis expanded to localize the above optimized solution [10 11]

ampe AREM-B4DVar system is based on AREM (version240) and the Backward-4DVar method (Wang et al [11])Cheng et al [12] used this system to establish the local andnonlocal operators of GPS RO refractivity data ampe as-similation experiments for different observational operatorswere carried out and the positive contribution of nonlocalobservation operators when forecasting heavy rain wasverified ampe coordinate plane of the assimilation systemadopts the mode surface for the AREM and the assimilationcontrol variables include the forecasted temperature zonalwind meridional wind specific humidity ground pressureand geopotential height ampe assimilated observation datainclude conventional ground and upper-air observationdataampe assimilated large-scale background field uses globalmedium-term numerical forecast products

22 -e Observation Operator of the GPS RO RefractivityData In GPS RO data refractivity data are usually given asan atmospheric observation product and the local obser-vation operator that links the data with the control variable is

2 Advances in Meteorology

N 776p

T+ 373 times 105

pq

T2(0622 + 037q) (3)

where p represents the air pressure (hPa) Trepresents the airtemperature (K) and q represents the specific humiditySince the refractivity data are obtained under the assumptionof spherical symmetry gradient information for the ele-ments on the ray is not considered therefore the calculationaccuracy is lower especially in the vicinity of bad weatherprocesses and non-local operators are considered ampeseoperators can theoretically partially compensate for theinadequacy of local observation operators

3 Case Description and Experimental Design

ampe study was performed with a sample rainstorm from July 4to July 5 2007 in the Yangtze-Huaihe River Basin in Chinaampe reason why the rainstorm occurred was because ofa large-scale circulation pattern at 500 hPa where the cir-culation to the southwas higher than that over East Asia therewas a weak ridge over the Hetao region and the Yangtze-Huaihe River Basin was influenced by a shear line at 700 hPaSuch a circulation pattern was extremely optimal for theformation of convection As the tropical system aroundHainan moved northward a torsion in the subtropic highoccurred which strengthened warm air propagation north ofthe subtropical high in the west To summarize the rainstormwas caused by the interaction between two air currents

Experiments for predicting heavy rainfall which occurredfrom July 4 2007 to July 5 2007 in the Yangtze-Huaihe RiverBasin in China with data from the Constellation ObservingSystem for Meteorology Ionosphere and Climate (COSMIC)which included GPS RO refractivity data bending angles androutine GTS data are designedampe designed assimilation testtime windowwas the period from 1800 on July 3 2007 (UTCsame below) to 00UTC on July 4 2007ampe timewindow datainclude data from each occultation detection time (Figure 1)conventional ground data at 18 UTC and 00 UTC and high-altitude observation data

ampe forecast experiment scheme adopts the AREM240forecasting model for a limited-area regional numericalweather forecasting system Its horizontal resolution is30 km the top layer of the model is at 10 hPa and the modelarea (14degN-51degN 74degE-136degE) covers China and the sur-rounding areas A specific description of the experiment isgiven in Table 1 which includes the time variable boundaryconditions and explicit cloud physics processes using theparameterization process for cold cloud precipitationprocess parameterization [13]

4 A Case Study of the GPS RORefractivity DataQuality Control Scheme

As the GPS RO data cannot be regarded as uncorrelated andare very dense at the vertical direction (Chen et al [12] [14])the resolution the coverage and data density can differentlycontribute to the analysis and the forecast [15 16] An es-sential step before data assimilation is quality control Givenefficient quality control data with too many observation

errors data with observation operators that cannot simulatereasonable values and data that represent small-scale pro-cesses and cannot be resolved by the model resolution can beremoved and not reduce the positive effect of the dataassimilation

Finding outlier data is commonly used in quality controland we should find the statistical characteristics and dis-tributions by studying a large amount of observation dataand the differences between observation data and modelsimulationsampe standard deviation is intended to be used inthis paper as the main quality control method

S

1

nminus 1

1113970

1113944

n

i1xi minusx( 1113857

2 (4)

where S represents the standard deviation of observation xampe GPS RO refractivity data are derived from the

spherical symmetry assumptionampe closer the height is to theground the more nonuniformly distributed is the watervapor which causes data errors in the form of nonlineargrowth To improve the effect of GPS RO refractivity dataassimilation and forecasting this paper adopts a simplequality control scheme for refractivity data (1) exclude ob-servation data below 3 km with errors in O-B that are toohigh (2) set a high vertical resolution due to the GPS ROrefractivity data (there are 500m intervals in the troposphereand the upper troposphere to the stratosphere has an intervalof nearly 1 km) while keeping the vertical level of the patternrelatively small (usually only approximately 30 layers)therefore unnecessary observations between the two modellevels should be eliminated to match the resolution of themodel (3) use standard deviation between the observationdata and model simulations to get rid of the outlier data

To determine the influence of the quality control schemeon the assimilation effect four experiments (Table 2) weredesigned the control run test (CTRL) the GPS RO re-fractivity data local observational operator assimilationscheme (REF_NQC) the GTS conventional radiosondeobservation data assimilation scheme (STN) and the GPSRO refractivity data local operator plus quality control as-similation scheme (REF_QC)

By comparing the difference before and after qualitycontrol between the initial values of the assimilation andobservation data for GPS RO refractivity it can be clearly seenthat the assimilation analysis is closer to the observations inthe region when using observation data after quality controlHowever after performing the quality control near theground where some excluded data were removed and therewas no observation constraint the analysis values were farfrom the observation values (Figure 2) Based on the relativedeviation when comparing the initial value of the qualitycontrol with the actual observation the region with a rela-tively small difference was in the upper troposphere and thearea with the greatest error was still concentrated in the lowertroposphere which was also due to the large error in the low-level refractivity observations

ampe deviation caused by the quality control does notmean that causes poor prediction ability this can be seenfrom the differential forecast field for 24-hour cumulative

Advances in Meteorology 3

precipitation In the northwest part of the main rain-beltafter the quality control the increment in the data assimi-lation moves southward and strengthens (Figure 3) Fromthe assimilation increment after quality control (Figure 3)increments in the 700 hPa and 500 hPa quality control testsare mainly concentrated to the northeast of the low-pressure

system and south of the low-pressure system respectivelyWith the increase in the northeastern region of low pressureand the weakening of the southern region of low pressureairow to the south near the rain-belt is enhanced whichcauses an overall southward movement in the rain-belt andis consistent with the actual situation

2007070322

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

15N

20N

25N

30N

35N

40N

45N

50N 2007070323

20070703222007070321

20070704002007070319

20070703212007070322

2007070321

2007070400 2007070323

20070703192007070321

2007070400

2007070322

2007070321

Figure 1 e distribution of GPS RO data in the case study (assimilation time window from 18 UTC on July 3 2007 to 00 UTC on July 42007) [12]

Table 1 Specic description of the experiment

Items SchemeForecast model AREM240Spatial resolution 37 kmVertical resolution 32 levels model top at 10 hPaModel area 14degN-51degN 74degE-136degE

Parameterization of explicit cloud physics Parameterization process for cold cloud precipitation(Xu [13])

Parameterization of land surface processes ZENG (1998)Parameterization of boundary layer processes CCM3 (1993)Equilibrium radiation of the land surface MM5 (1982)Boundary conditions Time variable boundary conditions

Table 2 Descriptions of the four experiments with their quality control schemes

Experiment With or withoutGTS data

With or without GPSRO refractivity data

With or without GPSdata quality control

Control run test (CTRL) No No NoRefractivity data local observational operatorassimilation scheme (REF_NQC) No Yes No

Conventional observation data assimilation scheme(STN) Yes No No

GPS refractivity data local operator plus qualitycontrol assimilation scheme (REF_QC) No Yes Yes


5 A Batch Test for Occultation DataAssimilations and Forecasts during the FloodSeason in 2009

e batch assimilationprediction programmes during theood season are as follows

(1) Batch tests are performed for 119 days from May 42009 to August 30 2009

(2) e observation data used for assimilation are theCOSMIC occultation refractivity data and conven-tional observation data at intervals of 6 hours from18 UTC to 00 UTC

(3) Global midterm numerical forecast products areused as background eld data

(4) e B-4DVar batch assimilationprediction testschemes include the control run (CTRL) a GTSconventional radiosonde data assimilation testmethod (stn_b4dvar) the GPS RO refractivity dataassimilation test (gps_only) and the test that usesGPS RO refractivity data simultaneously with GTSconventional radiosonde data (stn_gps_b4dvar)

All of the observation operators for GPS RO refractivitydata assimilation are local observation operators

To evaluate the attribution of the assimilation to themodel forecast we usually use reat Score (TS) and BiasScore of the precipitation forecast accuracy

e formula for computing the TS iscorrect

(forecast + observedminus correct) (5)

For a perfect forecast correct forecast observed toyield a TS of 1 e worst possible forecast with correct 0yields a TS of zero

e basic formula for computing the Bias is

forecastobserved

(6)

is quantity gauges the accuracy of arealstation cov-erage of a specied precipitation threshold amount re-gardless of accuracy in location An ideal forecast wouldhave forecast observed to yield a Bias of 1

e TS score and B score results for the GPS ROrefractivity data assimilation test (gps_only) and controlrun (CTRL) are analysed (Figures 4 and 5) It can be seenthat there are certain improvements in the TS score for alllevels of light rain moderate rain and heavy rain in the24-hour forecast but the score for heavy rain is slightlyworse Excluding heavy rainfall the B score improvedsomewhat and was better than the reference test At 48hours the improvement in the TS score was more ob-vious and the B score was similar to the reference test interms of light rain and moderate rain In contrast the Bscore for heavy rain and rainstorms was larger than thereference test results

As a whole data assimilation using only radio occul-tation data canmake positive improvements to both 24-hourand 48-hour rainfall forecasts it can obtain a better B scorein the 24-hour forecast and the TS score is better in the 48-hour forecast

e results of the three experiments including the GPSRO refractivity data and GTS conventional radiosonde data

120 125 130 140135 150145 160155 170165 180175Refractivity (N)

7600

7400

7200

7000

6800

6600

6400

6200

6000

5800

5600

5400

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(a)

195 200 205 210 215 220 225 230 235 240 245 250 255 260 265Refractivity (N)

2800

2900

3000

3100

3200

3300

3400

3500

3600

3700

3800

3900

4000

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(b)

Figure 2 Comparison of dicopyerent initial values and occultation refractivity observations in the middle-upper troposphere (a) and middle-lower troposphere (b) represents actual occultation observations (OBS) bull stands for the reference control test (CTRL) represents thenonquality control scheme (REF_NQC) and represents the quality control scheme (REF_QC)


assimilation test (stn_gps_b4dvar) the GTS conventionalradiosonde data assimilation test (stn_b4dvar) and thecontrol run test (CTRL) (Figures 6 and 7) were comparedWe know that stn_b4dvar has a certain improvement in theTS scores in both the 24-hour and 48-hour forecasts for alllevels of light rain heavy rain moderate rain and heavyrain compared with the CTRL In addition the B scoreresults are better than the control run test results at alllevels (except that it is larger for heavy rain in the 48-hourforecast) When comparing the ecopyect of GPS RO re-fractivity data when using GTS conventional radiosondedata in the stn_gps_b4dvar and stn_b4dvar tests the TSscore did not improve in the 24-hour forecast and therewas only a slight improvement in the B score However inthe 48-hour forecast the TS score was slightly improved forlight rain and heavy rain moderate rain and heavy rain

were comparable in the reference test and the corre-sponding B score was larger

Overall when using GPS RO refractivity data and GTSconventional radiosonde data the results indicate that theuse of GPS RO refractivity data can achieve a better per-formance for light rain and heavy rain at 48 hours but theyhave a less positive ecopyect on the 24-hour forecast

6 Summary

According to the analyses from the experiments above itis obvious that the use of GPS RO refractivity data canimprove the prediction accuracy of heavy rain-belts andregional rainfall intensity based on the AREM-B4DVar dataassimilation system By comparing various test schemes thefollowing conclusions are obtained

80E 90E 100E 110E 120E 130E

50N

45N

40N

35N

30N

25N

20N

15N

ndash50

ndash25

ndash10

1

10

25

50

(a)

0

0

0

0

0

0

0

0

0

10

15

5

ndash5

ndash5

ndash5 ndash10

ndash10

ndash15

ndash15

5

5

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

6

(b)

ndash5

ndash5

ndash10

ndash10

ndash15

ndash15

ndash15ndash20ndash25

ndash30ndash35

0

0

0 0 0

0

0

510

15

20

105

5

5

0

0

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

9

(c)

Figure 3 24-hour cumulative precipitation forecast dicopyerence between the quality control plan (REF_QC) and nonquality control plan(REF_NQC) (Figure 3a unit mm) and the dicopyerence in the initial values of the data assimilation analysis (Figure 3b 700 hPa Figure 3c500 hPa) e contour represents geopotential height increments (unit GPM) e arrow vectors represent wind speed increments in ms


0

01

02

03

04

05

06

07

08

ge01mm ge10mm ge25mm ge50mm

TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)

Figure 5 Comparison of the (a) TS score and (b) B score for 48-hour cumulative precipitation across several schemes fromMay 4 2009 toAugust 30 2009

0

01

02

03

04

05

06

07

08


TS sc

ore

gps_stn_b4dvarstn_b4dvarCTRL

(a)

0

05

1

15

2

25

3


B sc

ore


(b)

Figure 6 Comparison of the 24-hour cumulative precipitation (a) TS scores and (b) B scores for several schemes fromMay 4 2009 to August30 2009

0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)



(1) Both tests show that this new method can makea positive improvement to regional rainfall forecastsby using GPS RO refractivity data

(2) Only the use of GPS RO refractivity data can makepositive improvements to both 24-hour and 48-hourrainfall forecasts and obtain better B scores in 24-hour forecasts and TS scores in 48-hour forecasts

(3) When using GPS RO refractivity data and GTSconventional radiosonde data the results indicatethat the use of GPS RO refractivity data can achievebetter performances in 48-hour forecasts of light rainand heavy rain but there is a less positive ecopyect onthe performance in the 24-hour forecasts

Data Availability

e gure of GPS RO observation location of the heavyrainfall case on Jul 3 2007 was used to support this study andis available at DOI 101360012012-17 ese prior studiesare cited at [12] within the text as references

Conflicts of Interest

e authors declare that there are no conicts of interestsregarding the publication of this paper

Acknowledgments

e authors are grateful for the GPS RO refractivity datafrom the COSMIC Data Analysis and Archive Center isresearch was nancially supported by the National KeyResearch and Development Program of China (Project No2017YFB1002702)

References

[1] R A Anthes C Rocken and Y H Kuo ldquoApplications ofCOSMIC to meteorology and climaterdquo Terrestrial Atmo-spheric and Oceanic Sciences vol 11 no 1 pp 115ndash1562000

[2] X Zou Y H Kuo and Y R Guo ldquoAssimilation of atmo-spheric radio refractivity using a nonhydrostatic adjointmodelrdquo Monthly Weather Review vol 123 no 7 pp 2229ndash2249 1995

[3] Y H Kuo X Zou andW Huang ldquoe impact of GPS data onthe prediction of an extratropical cyclone an observing systemsimulation experimentrdquo Dynamics of Atmospheres andOceans vol 27 no 1ndash4 pp 413ndash439 1997

[4] E R Kursinski S B Healy and L J Romans ldquoInitial results ofcombining GPS occultations with ECMWF global analyseswithin a 1DVar frameworkrdquo Earth Planets and Space vol 52no 11 pp 885ndash892 2000

[5] P Poli J Joiner and E R Kursinski ldquo1DVAR analysis oftemperature and humidity using GPS radio occultation re-fractivity datardquo Journal of Geophysical Research vol 107no D20 p 4448 2002

[6] C Y Huang Y H Kuo S H Chen and F VandenbergheldquoImprovements in typhoon forecasts with assimilated GPSoccultation refractivityrdquo Weather and Forecasting vol 20no 6 pp 931ndash953 2005

[7] X Zou ldquoTangent linear and adjoint of ldquoon-ocopyrdquo processes andtheir feasibility for use in 4-dimensional variational data as-similationrdquo Tellus A vol 49 no 1 pp 3ndash31 1997

[8] M Mu and J F Wang ldquoA method for adjoint variational dataassimilation with physical ldquoOnndashOcopyrdquo processesrdquo Journal of theAtmospheric Sciences vol 60 pp 2010ndash2018 2003

[9] B Wang and Y Zhao ldquoA new data assimilation approachrdquoActa Meteorological Sinica vol 20 no 3 pp 275ndash2822006

[10] B Wang J J Liu S D Wang et al ldquoAn economical ap-proach to four-dimensional variational data assimilationrdquoAdvances in Atmospheric Sciences vol 27 no 4 pp 715ndash7272010

[11] B Wang W Cheng Y Xu R Cheng Y Pu and B ZhangldquoA four-dimensional variational data assimilation ap-proach with analysis at the end of assimilation windowPart I methodology and preliminary testsrdquo Journal of theMeteorological Society of Japan vol 89 no 6 pp 611ndash6232011

[12] W Cheng B Wang and Y Xu ldquoAssimilation of GPS radiooccultation data with the local and non-local operators usingBackward-4DVar approach (in Chinese)rdquo Scientia SinicaMathematica vol 42 no 5 pp 377-378 2012

0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)

Figure 7 Comparison of the 48-hour cumulative precipitation (a)TS scores and (b) B scores for several schemes fromMay 4 2009 toAugust 30 2009


[13] Y Xu -e Improvment and Evaluation of the AREM WaterBearing Numerical Forecasting Model Institute of Atmo-spheric Physics Chinese Academy of Science Beijing China2009

[14] S-Y Chen C-Y Huang Y-H Kuo et al ldquoObservationalerror estimation of FORMOSAT-3COSMIC GPS radio oc-cultation datardquo Monthly Weather Review vol 139 no 3pp 853ndash865 2011

[15] J-H Ha J-H Kang and S-J Choi ldquoampe impact of verticalresolution in the assimilation of GPS radio occultation datardquoWeather and Forecasting vol 33 no 8 pp 1033ndash1044 2018

[16] R L Cucurull and T R Peevey ldquoAssessment of radio oc-cultation observations from the COSMIC-2 mission witha simplified observing system simulation experiment con-figurationrdquo Monthly Weather Review vol 145 no 9pp 3581ndash3597 2017


Hindawiwwwhindawicom Volume 2018

Journal of

ChemistryArchaeaHindawiwwwhindawicom Volume 2018

Marine BiologyJournal of


BiodiversityInternational Journal of


EcologyInternational Journal of


Hindawiwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2018

Forestry ResearchInternational Journal of



International Journal of

Geophysics

Environmental and Public Health

Journal of




Microbiology


Public Health Advances in

AgricultureAdvances in


Agronomy




MeteorologyAdvances in

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018Hindawiwwwhindawicom Volume 2018

ChemistryAdvances in

ScienticaHindawiwwwhindawicom Volume 2018


Geological ResearchJournal of

Analytical ChemistryInternational Journal of


Submit your manuscripts atwwwhindawicom

ampe results showed that the use of GPS RO refractivitydata has a positive effect on the simulation of typhoonprecipitation

ampe key point of this article is introducing the qualitycontrol scheme for the GPS RO refractivity data in theAREM-B4DVar system to evaluate the role of GPS ROrefractivity data assimilation in regional numerical weatherpredictions and through actual rainstorm cases and batchexperiments during the flood season to explore an effectivemethod for improving the forecasting ability of regionalnumerical forecast models by using GPS RO refractivitydata Furthermore the focus of this article is to providea basic theoretical basis and technical support for the de-velopment of occultation data assimilation methods in re-gional numerical forecast models and weather forecasts forshort periods of time

2 AREM-B4DVarDataAssimilationSystemandObservational Operators of GPS RO Data

21 Introduction of the AREM-B4DVar Data AssimilationSystem Four-dimensional variational data assimilation(4DVar) is one of the most promising methods for providingoptimal analyses of numerical weather predictions It is thepreferred development plan formost of the numerical weatherprediction centres in the world Based on the principle theoryof the method due to the dynamic and physical constraints ofnumerical models all observations are best fitted in the as-similation time window by the variational method and theinitial value of the optimal analysis at the beginning of theassimilation time window is obtained Under the constraint ofthe model the evolution trajectory of the analysis field withinthe assimilation time window is consistent with the actualobservation trend which allows the accuracy level of theforecast to be better improved at a future time

Developing an operational 4DVar system is a very largeproject and takes hard work It requires not only a corre-sponding tangent and adjoint model but also rigorouscorrectness and accuracy tests Given the ldquoon-offrdquo problemin the process of complex physics [7 8] these problemshave become a bottleneck that plagues the development of4DVar and limits the wide application of this method inoperational numerical models Many scholars have pro-posed numerous effective solutions to this problem Amongthem Wang and Zhao [9] proposed the concept of the 3Dmapping variation method (3DVM) by placing the initialvalue of the assimilation at the end of the assimilationwindow ampe use of mapping observations subtly avoidsthe use of an adjoint model Wang et al [10] proposeda four-dimensional variational assimilation method (calledDRP-4DVar) to reduce dimensionality projections usinghistorical sample fitting and dimensionality reduction pro-jection techniques Selecting perturbed samples that dependon the analysis time solves the problem that the backgrounderror covariance matrix is not explicitly developed in the4DVar Recently Wang et al [11] combined the advantagesof the 3DVM and DRP-4DVar proposed a backwardmapping four-dimensional variational assimilation method(Backward-4DVar referred to as B-4DVar) and established

the AREM mode of the B-4DVar system (which is termedthe AREM-B4DVar system) ampis method not only avoidstangent-linear and adjoint models but also reduces thecomputational cost of the assimilation window and becausethe initial value being generated at the end of the assimi-lation window it can also reduce the prediction error ac-cumulation throughout the assimilation window whichplays an important role in short-term and nowcastingforecasts is verified in the observational system experiments

ampe B-4DVar problem comes down to the minimizationof the cost function in the m-dimensional sample space(wherem represents the number of samples) [11] thereforethe classic 4DVar method which is defined for high di-mensions of the control variable space is implemented onthe m-dimensional sample space

xa xb + xaprime xb + Pxαa

1113957J αa( 1113857 minαisinEm

1113957J(α)

1113957J(α) 12αΤBminus1α α +

12Pyαminus 1113957yobsprime1113872 1113873

TPyαminus 1113957yobsprime1113872 1113873

⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩

1113957yobsprime Rminus1yobsprime

O RTR

(1)

ampe solution to the minimization problem mentionedabove can be expressed as

αa Bminus1α + PTy Py1113872 1113873minus1

1113957yobsprime (2)

Because the dimension of the matrix ism which is a smallnumber (under 100) calculating an inverse matrix is relativelyeasy in the above expression To slow down the un-derestimation of the B matrix and the false teleconnectionbetween the lattice variable and observed variable the Bmatrixis expanded to localize the above optimized solution [10 11]

ampe AREM-B4DVar system is based on AREM (version240) and the Backward-4DVar method (Wang et al [11])Cheng et al [12] used this system to establish the local andnonlocal operators of GPS RO refractivity data ampe as-similation experiments for different observational operatorswere carried out and the positive contribution of nonlocalobservation operators when forecasting heavy rain wasverified ampe coordinate plane of the assimilation systemadopts the mode surface for the AREM and the assimilationcontrol variables include the forecasted temperature zonalwind meridional wind specific humidity ground pressureand geopotential height ampe assimilated observation datainclude conventional ground and upper-air observationdataampe assimilated large-scale background field uses globalmedium-term numerical forecast products

22 -e Observation Operator of the GPS RO RefractivityData In GPS RO data refractivity data are usually given asan atmospheric observation product and the local obser-vation operator that links the data with the control variable is


N 776p

T+ 373 times 105

pq

T2(0622 + 037q) (3)










S

1

nminus 1

1113970

1113944

n


2 (4)









2007070322

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

15N

20N

25N

30N

35N

40N

45N

50N 2007070323

20070703222007070321

20070704002007070319

20070703212007070322

2007070321

2007070400 2007070323

20070703192007070321

2007070400

2007070322

2007070321


























forecastobserved

(6)





120 125 130 140135 150145 160155 170165 180175Refractivity (N)

7600

7400

7200

7000

6800

6600

6400

6200

6000

5800

5600

5400

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(a)

195 200 205 210 215 220 225 230 235 240 245 250 255 260 265Refractivity (N)

2800

2900

3000

3100

3200

3300

3400

3500

3600

3700

3800

3900

4000

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(b)






6 Summary


80E 90E 100E 110E 120E 130E

50N

45N

40N

35N

30N

25N

20N

15N

ndash50

ndash25

ndash10

1

10

25

50

(a)

0

0

0

0

0

0

0

0

0

10

15

5

ndash5

ndash5

ndash5 ndash10

ndash10

ndash15

ndash15

5

5

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

6

(b)

ndash5

ndash5

ndash10

ndash10

ndash15

ndash15


ndash30ndash35

0

0

0 0 0

0

0

510

15

20

105

5

5

0

0

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

9

(c)



0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)


0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3


B sc

ore


(b)


0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)






Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy















N 776p

T+ 373 times 105

pq

T2(0622 + 037q) (3)










S

1

nminus 1

1113970

1113944

n


2 (4)









2007070322

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

15N

20N

25N

30N

35N

40N

45N

50N 2007070323

20070703222007070321

20070704002007070319

20070703212007070322

2007070321

2007070400 2007070323

20070703192007070321

2007070400

2007070322

2007070321


























forecastobserved

(6)





120 125 130 140135 150145 160155 170165 180175Refractivity (N)

7600

7400

7200

7000

6800

6600

6400

6200

6000

5800

5600

5400

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(a)

195 200 205 210 215 220 225 230 235 240 245 250 255 260 265Refractivity (N)

2800

2900

3000

3100

3200

3300

3400

3500

3600

3700

3800

3900

4000

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(b)






6 Summary


80E 90E 100E 110E 120E 130E

50N

45N

40N

35N

30N

25N

20N

15N

ndash50

ndash25

ndash10

1

10

25

50

(a)

0

0

0

0

0

0

0

0

0

10

15

5

ndash5

ndash5

ndash5 ndash10

ndash10

ndash15

ndash15

5

5

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

6

(b)

ndash5

ndash5

ndash10

ndash10

ndash15

ndash15


ndash30ndash35

0

0

0 0 0

0

0

510

15

20

105

5

5

0

0

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

9

(c)



0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)


0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3


B sc

ore


(b)


0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)






Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy

















2007070322

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

15N

20N

25N

30N

35N

40N

45N

50N 2007070323

20070703222007070321

20070704002007070319

20070703212007070322

2007070321

2007070400 2007070323

20070703192007070321

2007070400

2007070322

2007070321


























forecastobserved

(6)





120 125 130 140135 150145 160155 170165 180175Refractivity (N)

7600

7400

7200

7000

6800

6600

6400

6200

6000

5800

5600

5400

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(a)

195 200 205 210 215 220 225 230 235 240 245 250 255 260 265Refractivity (N)

2800

2900

3000

3100

3200

3300

3400

3500

3600

3700

3800

3900

4000

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(b)






6 Summary


80E 90E 100E 110E 120E 130E

50N

45N

40N

35N

30N

25N

20N

15N

ndash50

ndash25

ndash10

1

10

25

50

(a)

0

0

0

0

0

0

0

0

0

10

15

5

ndash5

ndash5

ndash5 ndash10

ndash10

ndash15

ndash15

5

5

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

6

(b)

ndash5

ndash5

ndash10

ndash10

ndash15

ndash15


ndash30ndash35

0

0

0 0 0

0

0

510

15

20

105

5

5

0

0

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

9

(c)



0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)


0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3


B sc

ore


(b)


0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)






Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy



























forecastobserved

(6)





120 125 130 140135 150145 160155 170165 180175Refractivity (N)

7600

7400

7200

7000

6800

6600

6400

6200

6000

5800

5600

5400

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(a)

195 200 205 210 215 220 225 230 235 240 245 250 255 260 265Refractivity (N)

2800

2900

3000

3100

3200

3300

3400

3500

3600

3700

3800

3900

4000

Hei

ght (

m)

GPS REF

OBSCTRL

REF_NQCREF_QC

(b)






6 Summary


80E 90E 100E 110E 120E 130E

50N

45N

40N

35N

30N

25N

20N

15N

ndash50

ndash25

ndash10

1

10

25

50

(a)

0

0

0

0

0

0

0

0

0

10

15

5

ndash5

ndash5

ndash5 ndash10

ndash10

ndash15

ndash15

5

5

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

6

(b)

ndash5

ndash5

ndash10

ndash10

ndash15

ndash15


ndash30ndash35

0

0

0 0 0

0

0

510

15

20

105

5

5

0

0

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

9

(c)



0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)


0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3


B sc

ore


(b)


0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)






Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy


















6 Summary


80E 90E 100E 110E 120E 130E

50N

45N

40N

35N

30N

25N

20N

15N

ndash50

ndash25

ndash10

1

10

25

50

(a)

0

0

0

0

0

0

0

0

0

10

15

5

ndash5

ndash5

ndash5 ndash10

ndash10

ndash15

ndash15

5

5

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

6

(b)

ndash5

ndash5

ndash10

ndash10

ndash15

ndash15


ndash30ndash35

0

0

0 0 0

0

0

510

15

20

105

5

5

0

0

15N

20N

25N

30N

35N

40N

45N

50N

75E 80E 85E 90E 95E 100E 105E 110E 115E 120E 125E 130E 135E

9

(c)



0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)


0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3


B sc

ore


(b)


0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)






Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy















0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3

35

4


B sc

ore

gps_onlyCTRL

(b)


0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3


B sc

ore


(b)


0

01

02

03

04

05

06

07

08


TS sc

ore

gps_onlyCTRL

(a)

0

05

1

15

2

25

3


B sc

ore

gps_onlyCTRL

(b)






Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy


















Data Availability




Acknowledgments


References













0

01

02

03

04

05

06

07

08


TS sc

ore


(a)

0

05

1

15

2

25

3

35

4


B sc

ore


(b)









Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy





















Journal of










Volume 2018





Geophysics


Journal of




Microbiology





Agronomy















gpsradiooccultationdataassimilationinthearemregional...

Documents