Document Type : Case Study
Author
Departman of Natural Resources, Agricultural Education & Extension Institute, Agricultural Research, Education & Extension Organization, Tehran, Iran
Abstract
Drought can lead to substantial crop losses and may exacerbate food insecurity and socio-economic tensions. Agricultural production and rangelands in Markazi Province, Iran, are also adversely affected by drought events. In this study, the performance of remote sensing–based indices under drought and wet conditions was evaluated using 15-year MODIS-derived time series of the Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST) for August. August was selected because it coincides with peak vegetation cover and maximum thermal stress, enabling a clearer assessment of index performance. Accordingly, the Vegetation Condition Index (VCI) and Temperature Condition Index (TCI) were computed from the 15-year NDVI and LST time series, respectively, and the Vegetation Health Index (VHI) was then derived by assigning equal weights to VCI and TCI. The Vegetation Temperature Condition Index (VTCI) was also calculated by integrating NDVI and LST using an alternative algorithm that accounts for minimum and maximum vegetation thermal conditions. Drought severity maps based on VHI and VTCI were generated for August, and temporal variations in drought severity classes were analyzed over the 15-year period. The results indicated good agreement between VHI and VTCI in representing drought severity across the study years. For validation, the 3-month Standardized Precipitation Index (SPI) ending in August was calculated using data from 36 meteorological stations. Comparison with SPI showed that VHI better captured agricultural drought severity (R2=0.67), likely due to the balanced contribution of vegetation condition and land surface temperature in the VHI formulation. Overall, the findings suggest that remote sensing–based agricultural drought indices can support decision-makers and planners by mitigating the spatial limitations of point-based measurements.
Extended Abstract
Introduction
Agricultural drought disrupts plant growth by reducing soil moisture. It occurs when the water demand for evapotranspiration exceeds the available water in the soil. As a natural and recurring phenomenon driven by rainfall variability, agricultural drought leads to delayed planting, reduced growth, and decreased crop yield, ultimately affecting farmers' livelihoods. Monitoring agricultural drought using satellite imagery is an effective method for mitigating its impacts at local and regional scales. In this context, remote sensing has been widely employed for agricultural drought monitoring, leading to the development of indices such as the Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST). NDVI indicates vegetation health and density, but alone it is insufficient for drought detection as it does not reveal initial water stress. LST serves as a complementary measure, reflecting the surface energy balance and plant water stress. The relationship between these two indices—often visualized as a triangular scatter plot—can be negative (when water is the limiting factor) or positive (when sunlight is the limiting factor), depending on the season and regional moisture conditions. To effectively monitor agricultural drought, various composite indices have been developed. Two important examples are the Vegetation Health Index (VHI) and the Vegetation Temperature Condition Index (VTCI). VHI is derived from a simplified, equally weighted combination of NDVI and LST, while VTCI employs a more complex algorithm that targets thermal conditions within specific NDVI ranges. Field studies (e.g., in Bangladesh and Markazi Province, Iran) have confirmed the efficacy of these indices, particularly VHI, in zoning and classifying drought severity from mild to very severe. This research compares the effectiveness of these two indices in monitoring agricultural drought.
Material and Methods
This study was conducted in Markazi Province, Iran. The province covers an area of approximately 29,127 km², has a semi-arid climate, and receives an average annual rainfall of about 290 mm, ranging from 520 mm in Shazand to 177 mm in Delijan. About 75% of the province is mountainous. The data used in this study include meteorological data extracted from monthly rainfall statistics of 36 synoptic and rain gauge stations over a 15‑year period (1999–2014). The primary meteorological index used was the Standardized Precipitation Index (SPI) on a 3‑month scale. Remote sensing data were obtained from the MODIS sensor onboard the Terra satellite. NDVI was derived from the MOD13Q1 product with a spatial resolution of 250 m and a temporal resolution of 16 days for the months April to September. LST was extracted from the MOD11A2 product with a resolution of 1 km and an 8‑day interval for the same period. The remote sensing time‑series indices studied included two composite indices based on NDVI and LST for agricultural drought monitoring: (1) the Vegetation Health Index (VHI), obtained from a simple combination of NDVI and LST, which itself comprises the Vegetation Condition Index (VCI)—measuring long‑term NDVI changes—and the Temperature Condition Index (TCI)—evaluating long‑term LST changes—with VHI calculated as the equally weighted average of VCI and TCI; and (2) the Vegetation Temperature Condition Index (VTCI), which has a more complex algorithm defined based on the relationship between LST and NDVI on a specific date within the study area. In this method, for each NDVI value, a maximum (warm/dry edge) and minimum (cold/wet edge) range of LST is determined, and the temperature position of each pixel within this range is measured. Lower VTCI values indicate more severe drought. The aim of this research is to compare the efficiency of VHI and VTCI in monitoring agricultural drought in Markazi Province.
Results and discussion
This research aimed to monitor agricultural drought in Markazi Province from 2000 to 2014 and compared two major remote sensing indices—the Vegetation Health Index (VHI) and the Vegetation Temperature Condition Index (VTCI)—with the meteorological Standardized Precipitation Index (SPI). The characteristics of the indices are as follows:
SPI was calculated as a ground‑based meteorological drought indicator for August at a 3‑month scale, using precipitation data from 36 synoptic and rain gauge stations. The extracted SPI values indicated mild to moderate drought conditions in most stations during August. VHI is a composite index derived from VCI and TCI, obtained by normalizing NDVI and LST time series, respectively. This index was calculated using equal weighting of the two sub‑indices and classified into five drought intensity classes. The final maps revealed that 2000, 2001, and 2010 were the driest years, while 2003, 2004, 2006, and 2007 had the wettest conditions. VTCI, with a more complex algorithm, operates based on the relationship between LST and NDVI on specific dates. By plotting a scatter diagram of these two variables, the "warm edge" (dry conditions) and "cold edge" (wet conditions) were determined, and the position of each pixel within this spectrum was assessed. This index was also divided into five drought classes. The results indicated that 2000, 2004, 2011, and 2013 experienced the most severe droughts, while 2006, 2010, and 2014 had the least stressful conditions.
Comparison of the indices revealed that VTCI is more sensitive to drought and wetness variations than VHI. This difference stems from their mathematical logic: VHI is derived from a global averaging of NDVI and LST across the entire region, whereas VTCI analyzes temperature variations among pixels with similar NDVI values, making it more detail‑oriented. Regression analysis between the satellite‑based indices and SPI showed that in August, TCI and VHI had a significant correlation with SPI. In contrast, VTCI did not show a significant relationship with SPI during this month. This finding indicates that, in this specific study, VHI aligns better with the standard meteorological drought index (SPI). For validation, a linear regression model based on TCI and NDVI was developed to predict SPI. Evaluating this model for the year 2015 (outside the study period) using Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) confirmed the model’s acceptable accuracy (low error values). This research demonstrated the effectiveness of MODIS satellite imagery and its derived indices (e.g., NDVI and LST) in the spatiotemporal monitoring of agricultural drought in Markazi Province. Although VTCI exhibited higher analytical sensitivity, in this particular case, the composite VHI index was identified as a more suitable tool for monitoring agricultural drought and estimating conditions consistent with precipitation data, due to its significant correlation with the ground‑based SPI and its relatively simpler calculation. This approach can address the gap caused by the scarcity of ground stations and enable extensive, timely drought monitoring.
Conclusion
Satellite imagery, particularly products from the MODIS sensor such as NDVI and LST, has high capability for monitoring dynamic phenomena like agricultural drought due to its temporal and spatial continuity. These products require no complex preprocessing and enable the creation of long‑term time series, which form the basis for developing composite indices such as VHI and VTCI. This study compared two key composite indices for monitoring agricultural drought in Markazi Province. VHI operates based on the simplified average of vegetation density (NDVI) and land surface temperature (LST). VTCI uses a more complex algorithm and determines drought intensity by identifying the warm edge (dry conditions) and cold edge (wet conditions) in the LST–NDVI relationship. The research findings indicated that in the study area, the performance of VHI in monitoring agricultural drought was superior to that of VTCI. This superiority was demonstrated by a more significant correlation of VHI results with the ground‑based SPI. Specifically, VHI maps identified 2000, 2001, and 2010 as the driest years, which aligned better with SPI results that identified 2000 and 2008 as the driest. Furthermore, VHI classified over 56% of the province’s area as "no drought," while VTCI placed about 44% of the area in the "moderate drought" category. In conclusion, composite remote sensing indices such as VHI can serve as suitable alternatives to meteorological indices for regional drought monitoring, especially in areas with sparse weather stations, and can assist planners in drought risk management. To improve accuracy, future studies are recommended to incorporate more ground data, consider physiographic factors, and utilize other sensors (e.g., Sentinel and Landsat).
Keywords
Subjects
