Drought Monitoring Methodology Based on AVHRR Images and SPOT Vegetation Maps ()
1. Introduction
In the past few decades, both the frequency and intensity of droughts have increased in a number of regions in the world [1,2]. This recurring trend has negatively impacted many of these areas in terms of the large annual losses in vegetation it causes. Because of the serious social, economic, and environmental ramifications, drought monitoring has become a high priority for many countries, and especially developing ones. Since the late 1980s, satellites have been used for detecting and monitoring droughts as well as assessing their impact on agriculture [3].
One of the most efficient monitoring methods involves the use of Remote Sensing Technology. With this technique, sensors operating in several spectral bands are mounted on satellites in order to rapidly obtain and distribute drought information over large geographic areas. While the satellite is in orbit, it is able to explore the earth’s surface where in just a matter of a few days it is able to identify, monitor, and assess drought conditions. Using this technology, one can not only investigate the effect of droughts on vegetation cover but also their effects on ground water, surface temperature, and precipitation. In this way, a better understanding of temporal and spatial characteristics of the drought for a specific region can be achieved.
By monitoring droughts over a long period of time (i.e., 10 years or more) early drought warning systems can be developed. These early warning systems are important because they are being relied upon more and more to ensure global food security [4]. Previous methods of drought monitoring have typically used vegetation indices for drought monitoring. Kogan (1997) developed a method that analyzed the relationship between the Vegetation Condition Index and the Temperature Condition Index (VCI-TCI) [3]. Eklundh in 1996 assessed the possibility of using NDVI data for crop and natural vegetation monitoring by measuring the cross-correlation between the time series of NDVI and meteorological indicators such as rainfall (for areas where rainfall is a limiting factor) [5]. Eklundh’s results showed that the correlation between NDVI and rainfall coefficients is quite high, between 0.7 and 0.9, and that NDVI lags behind rainfall by one to three months. Eklundh concluded that if rainfall is to be used as an indicator of seasonal vegetation development, then there will also be certain limitations in the ability of NDVI’s to monitor temporal vegetation variations.
Singh et al. (2003) used NDVI, VCI and TCI to monitor droughts as well as estimate vegetation health. In their research, they used both vegetation and temperature condition indices to monitor droughts in India [6]. Bhuiyan et al. (2006) developed a new SWI index to assess groundwater recharge-deficit [7]. The correlations of drought with respect to different indices are visually interpreted and necessitate certain disclaimers, namely that negative SPI anomalies do not always correspond to drought, and that a delay exists between hydrological and vegetative stress. There is also a delay between vegetative and hydrological stress. In order to identify a trend over 10 years, the annual rainfall was plotted against the cumulative annual NDVI values and it was found that the NDVI values were parallel to the rainfall, but with a time lag of one year. Thus, this 2006 study developed a hypothesis that there is a time lag between the rainfall and NDVI responses, and our study also rests on such a hypothesis. The results obtained with this method were compared with two other methods: 1) the relationship between NDVI and rainfall during the plant growing season, and 2) the relationship between NDVI and rainfall as well as NDVI and surface temperatures. The former characterizes the dynamics of vegetation development via its growing season’s parameters on a consistent spatial scale, while the latter is based on the relationship between the Global Vegetation Index (GVI) and the Temperature Condition Index (TCI) with rainfall [8]. Bajgiran et al. (2008), using AVHRR images they obtained for 1997-2001, compared precipitation data in the north-west region of Iran with VCI and NDVI indices [9]. According to their results, both NDVI and VCI indices can be used to monitor regional droughts. A good linear correlation between monthly precipitation levels and NDVI or VCI amounts was observed and can possibly be used to predict and manage drought risk.
The aim of this research was to develop drought maps specifically for the southwest region of Iran, with a particular focus on the province of Fars, an area that has been suffering from disastrous hydrological drought since 2001. Despite the existence of previous research on the development of drought maps, a special method for extracting exact drought estimates for this particular Iranian region has yet to be developed. Previous research has typically relied on only one or two meteorological or hydrological indices. In our study, however, we incorporated remote-sensing data, related data from the NOAA-AVHRR sensor, and SPOT vegetation data (to verify extracted vegetation indices) to draw our conclusions.
In this study, analyses of monthly drought dynamics were calculated to identify drought configurations within hydrological, meteorological and vegetative domains. Making both quantitative and visual comparisons of drought dynamics in meteorological, hydrological, and vegetative domains in the province of Fars allowed us to generate more useful and reliable results. To verify the vegetation indices extracted from the NOAA-AVHRR images, SPOT-VEG images from the same time period (1998-2007) were used. The Standardized Water Level Index (SWI) and the Standard Precipitation Index (SPI) were used to monitor and analyze hydrological and meteorological drought, respectively. Normalized Difference Vegetation Index (NDVI), Vegetation Condition Index (VCI), Temperature Condition Index (TCI) and Vegetation Health Index (VHI) were used to assess vegetative drought. The development of VHI was done using VCI and TCI indices because they are more effective at monitoring vegetative drought than are other indices [6,10,11]. Given that the province of Fars is the most important agricultural region of Iran, the effects of its ground water consumption on NDVI were also investigated. The hydrological and meteorological stations of Fars are well distributed (geographically) throughout the province, and temporal analysis was carried out over a span of 10 years. This specific time frame was used so that the long-term effects of precipitation and changing ground water levels could be studied. Compared with traditional in-situ measurements, the results obtained from remote-sensing methods are capable of providing more reliable drought maps. Chapter 3 explains our implementation methodology while Chapter 4 provides a comprehensive presentation and analysis of our results. Finally, concluding remarks are given in Chapter 5.
2. Materials and Methods
2.1. Study Area
Located in the southern part of the country, Fars is one of the 30 provinces that comprise present day Iran (Figure 1). As of 2006, the province was home to 4.34 million people, 62% of which are registered urban dwellers, 38.1% villagers, and 0.7% nomadic tribes. Three distinct climatic regions exist with the province’s 122, 400 km² territory: a mountainous area in the north and northwest characterized by moderately cold winters and mild summers; a central region with relatively rainy, mild winters and hot, dry summers; and a southern region with relatively rainy, mild winters and hot, dry summers. Shiraz is the capital and center of Fars. The province consists of the following counties: Estahban, Abadeh, Eqleed, Bovanat, Jahrum, Darab, Sepidan, Shiraz, Fasa, Firouzabad, Kazeroon, Lar, Lamerd, Marvdasht, Mamasani, Khonj and Nayriz. Agriculture is the most important activity in Fars, and its major products include cereal (wheat and barley), citrus fruits, dates, sugar, beets and cotton. From an agricultural point of view, Fars is one of the most strategic provinces in Iran, as it is responsible for producing 37% of the country’s wheat.
2.2. Satellite Data
2.2.1. Advanced Very High Resolution Radiometer (AVHRR) Images
The image data taken by the Local Area Coverage Advanced Very High Resolution Radiometer (LAC AVHRR) aboard the National Oceanic and Atmospheric Administration (NOAA) 14-16-17 satellite were preprocessed using ENVI. For temporal analysis, a 10-year period was chosen in order to study the long-term effects of precipitation and groundwater levels on the vegetation coverage. This research was limited to ten years due to the inability to access any data records prior to 1998.
Initial examination of AVHRR data collected from the NOAA satellite database [12] revealed that a number of the images had severe cloud contamination and/or missing passes. Of the 254 images collected during 1998 - 2007 (April - September), 190 raw AVHRR images were selected.
2.2.2. SPOT Vegetation Maps
In order to study the vegetation cover in Fars, 10-day composite NDVI data (derived from the sensor VEGETATION on board the SPOT satellite platforms) was acquired from the “Vlaamse Instelling Voor Technologish Onderzoek” [13]. The SPOT-VGT S10 (10-day composite) NDVI composites have a spatial resolution of

Figure 1. Study area: Map of Iran (left) and the various counties within the province of Fars (right).
1 km2 and were derived from primary SPOT-VGT products; the composites were corrected for reflectance, scattering, water vapor, ozone, and other gas absorption using the procedures described by Achard et al. and Duchemin et al. [14,15].
The maximum value compositing (MVC) procedure as described by Holben (1986) was used to merge NDVI values over the course of ten days [16]. The resulting surface reflectance value for each pixel thus corresponds to the date with the maximum NDVI-value in a 10-day period. Maximum value compositing for the synthesis of daily NDVI-values was found to be a reliable procedure for detecting changes in vegetation cover [17,18].
Typical NDVI-values range between 0.1 and 0.7 for vegetated areas, with a higher (composite) NDVI value equating to denser, greener vegetation. The temporal evolution of NDVI-values is considered to be an effective way to analyze the impact of 1) natural seasonal variations, 2) extreme climatic events, and 3) human activities on ecosystems [19]. The temporal evolution of NDVI values for the period 1998 - 2005 were analyzed in four selected counties in Fars. For each county average, decadal NDVI-values on farmland were calculated by spatial aggregation of the 1 km2 pixels.
2.3. Meteorological and Hydrological Data
Figure 2 shows the precipitation data collected from 122 synoptic and rainfall stations in the study area over a ten-year period from 1998 to 2007. Groundwater-level data from 378 observer wells within Fars was also collected.