QUANTITATIVE ESTIMATION OF DROUGHT RISK IN UKRAINE USING SATELLITE

DATA

Sergii Skakun1, Nataliia Kussul

1,2, Olga Kussul

2, Andrii Shelestov

1,2,3

1Space Research Institute NAS Ukraine and SSA Ukraine;

2National Technical University of Ukraine

“Kyiv Polytechnic Institute”; 3

National University of Life and Environmental Sciences of Ukraine

ABSTRACT

In this paper, we focus on quantitative drought risk

assessment using satellite data. Methods of the extreme

value theory (EVT) are applied for a time-series of

vegetation health index (VHI) derived from NOAA satellites

in order to provide drought hazard mapping. For this, a

Poisson-GP (Generalized Pareto) model is applied for

modelling VHI extreme values. The model allows estimation

and mapping of return periods of different categories of

drought severity. An approach to economical risk

assessment due to droughts is presented. The derived

drought hazard map is integrated with high resolution crop

map to provide final estimates of risk. The proposed

approach is implemented for quantitative assessment of

drought risk for the Kyiv region in Ukraine.

Index Terms— risk, drought, agriculture, satellite data,

Ukraine.

1. INTRODUCTION

Over last decades there has been an upward global trend in

natural disaster occurrence. Hydrological and

meteorological disasters, such as floods and droughts, are

the main contributors to this pattern [1]. Droughts are one of

the most dangerous and complex natural hazards. Unlike

other extreme phenomena, droughts usually develop slowly,

with no a clear onset, and do not have a direct structural

impact.

There are several types of droughts [2]: meteorological

(lack of precipitation during an extensive period of type),

hydrological (lack of surface and ground waters), and

agricultural (lack of soil moisture that leads to crop

damages). In past years droughts have had a significant

impact on food security in many regions [1]. In recent years

a risk-oriented approach for managing the risks of disasters

has been adopted [1], [3]-[6]. To enable drought risk

assessment, corresponding drought hazard and drought risk

maps should be developed. Drought risk is a function of two

arguments: hazard probability and vulnerability. In other

words, risk is a mathematical expectation of vulnerability

(consequences) function [3]. A traditional approach to assess

drought hazard is to analyze rainfall, temperature and soil

moisture measurements from meteorological stations using,

for example, extreme value theory (EVT) [7]. A number of

ground-based indicators have been developed to identify

droughts, e.g. PDSI (Palmer drought severity index) [8], SPI

(Standardized Precipitation Index) [9], CMI (Crop Moisture

Index) [10], SWSI (Surface Water Supply Index) [11]. The

accuracy of drought hazard estimates from in situ stations is

dependant on the density and uniformity of station location.

Providing a dense spatial coverage of ground measurements

is not always possible due resource constraints. For example,

Ukraine has a network of 180 meteorological stations with

density of a station per 3,225 sq. km which is too coarse.

Alternatively, Earth observation (EO) from space can

provide better spatial resolution with long-term archived

data. Other advantages of EO include human-independent

information, continuous and repetitiveness observations in

space and time, coverage of large areas, operational delivery

of information. EO data can be effectively used for both

drought hazard mapping [5], [12], [14], and drought

vulnerability assessment [15].

In this paper, a new approach to quantitative estimation

of drought risk in Ukraine based on satellite data is

presented. Drought hazard mapping is performed based on

the use of vegetation health index (VHI) derived from

NOAA satellites, and the EVT techniques. Drought

vulnerability is assessed by estimating the crop areas and

crop yield to quantify potential impact of a drought on crop

production. Finally, drought hazard and vulnerability maps

are integrated to derive a drought risk map.

2. STUDY AREA AND MATERIALS DESCRIPTION

Ukraine is one of the most developed agriculture countries

and one of the biggest crop producers in the world [13].

According to the 2011 statistics provided by the U.S.

Department of Agriculture (USDA) Foreign Agricultural

Service (FAS), Ukraine was the 8th largest exporter and

10th largest producer of wheat in the world. Since past

decade Ukraine experienced five droughts (2003, 2007,

2008, 2009 and 2010) that covered between 40% and 60%

of the country and up to 80% of the major grain crop area

[14].

5091978-1-4799-5775-0/14/$31.00 ©2014 IEEE IGARSS 2014

Vegetation health index (VHI) derived from polar-

orbiting NOAA satellites at 16 km spatial resolution from

1981 to 2012 is used as a main variable [14]. This spatial

resolution enables analysis of droughts at 256 sq. km scale

which is 12 times better than meteorological stations. VHI

data are provided as weekly composites. The reason for

selecting this vegetation index instead of others (e.g. NDVI,

VCI) is that it incorporates moisture and thermal conditions

of vegetation canopy and directly relates to the classification

of the droughts [16]: abnormally dry conditions (36 <=

VHI<= 40); moderate (26 <= VHI<= 35); severe (16 <=

VHI<= 25); extreme (6 <= VHI<= 15); exceptional (0 <=

VHI<= 5).

3. METHODOLOGY

For quantitative drought risk assessment it is necessary to

estimate drought probability (hazard) and potential losses. In

this paper, we focus on agricultural droughts that result in

crop damages, and consequently lead to decrease of crop

production and economical losses. Let z be a parameter that

characterizes a drought. In general, z can be represented

using one of the drought indicators that were previously

developed (e.g. PDSI, SPI, VHI). To estimate potential

losses due to a drought, the following information is

required: damage rate that depends on the parameter z (we

assume that lower z values lead to the larger damage rate);

crop area (geographical distribution of crops in the given

region); expected (potential) crop yield; and cost of crops.

Therefore, drought risk in the area (x, y)∈А can be estimated

as follows:

∑∫ ∫

∫

=

==

k A

z

z

kkkkxy

A

A

dzdxdyvyxsyxyzyxdzp

dxdyyxrR

2

1

),(),(),,()(

),(

, (1)

where pxy(z) is the probability density function (pdf) of

drought occurrence with level z at point (x, y); dk(x, y, z) is

the damage rate of crop k due to the drought with parameter

z (dk(x, y, z)∈[0; 1]); yk(x, y) is the crop yield (t/ha); sk(x, y)

is crop area (ha); vk is the crop cost (USD/t).

In this paper, a Poisson-GP (Generalized Pareto) model

[7] is applied to the time-series of satellite-derived VHI

values to estimate drought pdf function pxy(z). In particular,

for VHI datasets is applied. This model consists of a Poisson

process for modeling the occurrence of exceedance of high

threshold, and a generalized Pareto (GP) distribution for the

excess over the threshold:

( )[ ] γ−

σγ+−=γσ/1

/11),;( xxF . (2)

In order to relax assumption of independence of block

minima, we used cluster minima below a threshold value of

40 that corresponds to vegetation stress. To estimate

parameters of the GP distribution, we applied the maximum

likelihood (ML) method [17]-[19]. For estimating crop area

and crop yield, satellite images were used as well. In

particular, we used Landsat-5 satellite images and ground

area frame surveys for crop area estimation [20]-[23], and

MODIS-derived NDVI values for crop yield forecasting

[24]-[26].

4. RESULTS

The proposed methodology was applied to the VHI datasets

(1981-2012) for the territory of Ukraine. For each pixel of

the images, VHI cluster minima were estimated. The number

of clusters was different for each pixel.

Fig. 1. Example of VHI time-series with cluster minima

(top), and distribution of the number of clusters over

Ukraine (bottom)

Fig. 1 shows an example of VHI time-series with cluster

minima, and distribution of the number of clusters over

Ukraine. For each pixel, GP parameters were estimated.

Based on these parameters it is possible to estimate

return period (and correspondingly probability) of a drought

with specified VHI level. Fig. 2 shows return period for the

exceptional droughts (VHI < 5), and the VHI values for the

return period of 20 years (corresponds to the probability of

0.05).

Using the obtained estimates for drought probability,

crop maps for the region of Kyivska oblast in Ukraine and

Eq. (1), we estimated economical risk of exceptional

droughts. This case study region is selected because of

availability of a high resolution crop map. This map was

produced in 2010 at 30 m resolution using time-series of

Landsat-5 images. In particular, we consider winter wheat

that is a major crop in the region accounting for more than

40% of production of all crops. Therefore, a winter wheat

5092

mask is extracted from the crop map, and is further used for

drought risk mapping and quantification (Fig. 3).

Fig. 2. Return period (in years) for the exceptional droughts

(top), and VHI values for the return period of 20 years

(bottom)

Fig. 3. Winter wheat crop mask for the Kyiv region derived

from Landsat-5 images in 2010.

Fig. 4. Risk of economical losses due to exceptional

droughts influencing winter wheat in the Kyiv region

accumulated by counties.

All obtained parameters are integrated in Eq. (1) to map

and estimate exceptional drought risk for the Kyiv region

(VHI < 6). The obtained risk values are accumulated for the

counties (Fig. 4) and the whole region. Total risk of

economical losses due to exceptional droughts influencing

winter wheat production for the Kyiv region is estimated at

approximately 19 million USD.

5. CONCLUSIONS

A new approach to quantitative estimation of drought risk in

Ukraine based on satellite data is presented. Vegetation

health index (VHI) derived from NOAA satellites from 1981

to 2012 is used to estimate drought probability density

function. For this, a Poisson-GP (Generalized Pareto) model

is applied. Drought vulnerability is assessed by estimating

the crop areas and crop yield to quantify potential impact of

drought on crop production. Drought hazard and

vulnerability maps are integrated to quantitatively map a

drought risk.

6. REFERENCES

[1] Guha-Sapir, D., F. Vos, R. Below, and S. Ponserre, Annual

Disaster Statistical Review 2011: The Numbers and Trends, Centre

for Research on the Epidemiology of Disasters (CRED), Brussels,

2012.

5093

[2] A.K. Mishra, and V.P. Singh, “A review of drought concepts,”

Journal of Hydrology, 391(1), pp. 202-216, 2010.

[3] N.N. Kussul, B.V. Sokolov, Y.I. Zyelyk, V.A. Zelentsov, S.V.

Skakun, and A.Yu. Shelestov, “Disaster Risk Assessment Based on

Heterogeneous Geospatial Information,” Journal of Automation

and Information Sciences, 42(12), pp. 32-45, 2010.

[4] N. Kussul, A. Shelestov, S. Skakun, O. Kravchenko, Y.

Gripich, L. Hluchy, P. Kopp, and E. Lupian, “The Data Fusion

Grid Infrastructure: Project Objectives and Achievements,”

Computing and Informatics, 29(2), pp. 319-334, 2010.

[5] F. Tonini, G.J. Lasinio, and H.H. Hochmair, “Mapping return

levels of absolute NDVI variations for the assessment of drought

risk in Ethiopia,” International Journal of Applied Earth

Observation and Geoinformation, 18, pp. 564-572, 2012.

[6] S. Skakun, N. Kussul, A. Shelestov, and O. Kussul, “Flood

Hazard and Flood Risk Assessment Using a Time Series of

Satellite Images: A Case Study in Namibia,” Risk Analysis, 2013,

doi: 10.1111/risa.12156.

[7] S. Beguería, and S.M. Vicente-Serrano, “Mapping the hazard

of extreme rainfall by peaks over threshold extreme value analysis

and spatial regression techniques,” Journal of Applied

Meteorology and Climatology, 45(1), pp. 108-124, 2006.

[8] W.C. Palmer, “Meteorological drought,” Weather Bureau

Research Paper, 45, 58 p., 1965.

[9] N.B. Guttman, “Comparing the palmer drought index and the

standardized precipitation index,” JAWRA J. American Water

Resources Assoc., 34(1), pp. 113-121, 1998.

[10] W.C. Palmer, “Keeping track of crop moisture conditions,

nationwide: the new crop moisture index,” Weatherwise, 21, pp.

156-161, 1968.

[11] Shafer, B.A., and L.E. Dezman, Development of a Surface

Water Supply Index (SWSI) to Assess the Severity of Drought

Conditions in Snowpack Runoff Areas, Colorado State University,

Reno, NV, 1982, pp. 164-175.

[12] G.J. Huffman, R.F. Adler, D.T. Bolvin, and E.J. Nelkin, “The

TRMM multi-satellite precipitation analysis (TMPA),” in: Satellite

rainfall applications for surface hydrology, (Eds) M.

Gebremichael, and F. Hossain, Springer, Heidelberg, 2010, pp. 3-

22.

[13] A.Yu. Shelestov, A.N. Kravchenko, S.V Skakun., S.V.

Voloshin, and N.N. Kussul, “Geospatial information system for

agricultural monitoring,” Cybernetics and System Analysis, vol.

49, no. 1, pp. 124–132, 2013.

[14] F. Kogan, T. Adamenko, and W. Guo, “Global and regional

drought dynamics in the climate warming era,” Remote Sens. Lett.,

4(4), pp. 364-372, 2013.

[15] N. Silleos, K. Perakis, and G. Petsanis, “Assessment of crop

damage using space remote sensing and GIS,” International

Journal of Remote Sensing, 23(3), pp. 417-427, 2002.

[16] M. Svoboda, “An introduction to the Drought Monitor,”

Drought Network News, 12, pp. 15-20, 2000.

[17] G.M. Bakan, and N.N. Kussul, “Fuzzy ellipsoidal filtering

algorithm of static object state,” Problemy Upravleniya I

Informatiki (Avtomatika), no. 5, pp. 77-92, 1996.

[18] N.N. Kussul, “Neural networks learning using method of

fuzzy ellipsoidal estimates,” Journal of Automation and

Information Sciences, vol. 33, no. 3, pp. 52-57, 2001.

[19] A. Shelestov and N. Kussul, “Using the fuzzy-ellipsoid

method for robust estimation of the state of a grid system node,”

Cybern. Syst. Anal., vol. 44, no. 6, pp. 847–854, 2008.

[20] J. Gallego, A. Kravchenko, N. Kussul, S. Skakun, A.

Shelestov, and Y. Grypych, “Efficiency Assessment of Different

Approaches to Crop Classification Based on Satellite and Ground

Observations,” Journal of Automation and Information Sciences,

vol. 44, no. 5, pp. 67–80, 2012.

[21] F.J. Gallego, N. Kussul, S. Skakun, O. Kravchenko, A.

Shelestov, and O. Kussul, “Efficiency assessment of using satellite

data for crop area estimation in Ukraine,” International Journal of

Applied Earth Observation and Geoinformation, vol. 29, pp. 22-

30, 2014.

[22] N. Kussul, S. Skakun, A. Shelestov, O. Kravchenko, J.F.

Gallego, and O. Kussul, “Crop area estimation in Ukraine using

satellite data within the MARS project,” in: IGARSS 2012, 22-27

July 2012, Munich, Germany, pp. 3756-3759.

[23] S.V. Skakun, E.V. Nasuro, A.N. Lavrenyuk, and O.M.

Kussul, “Analysis of Applicability of Neural Networks for

Classification of Satellite Data,” Journal of Automation and

Information Sciences, vol. 39, no. 3, pp. 37-50, 2007.

[24] F. Kogan, N. Kussul, T. Adamenko, S. Skakun, O.

Kravchenko, O. Kryvobok, A. Shelestov, A. Kolotii, O. Kussul,

and A. Lavrenyuk, “Winter wheat yield forecasting: A comparative

analysis of results of regression and biophysical models,” Journal

of Automation and Information Sciences, 45(6), pp. 68-81, 2013.

[25] F. Kogan, N. Kussul, T. Adamenko, S. Skakun, O.

Kravchenko, O. Kryvobok, A. Shelestov, A. Kolotii, O. Kussul,

and A. Lavrenyuk, “Winter wheat yield forecasting in Ukraine

based on Earth observation, meteorological data and biophysical

models,” International Journal of Applied Earth Observation and

Geoinformation, vol. 23, pp. 192-203, 2013.

[26] O. Kussul, N. Kussul, S. Skakun, O. Kravchenko, A.

Shelestov, and A. Kolotii, “Assessment of relative efficiency of

using MODIS data to winter wheat yield forecasting in Ukraine,”

in: IGARSS 2013, 21-26 July 2013, Melbourne, Australia,

pp. 3235-3238.

5094