Assessment of Soybean yield using changes meteorological and satellite-based drought indices in the west of Golestan province

In the most parts of the country, the number of meteorological stations is inadequate and don't cover the appropriate statistical period. The conventional method of yield assessment in Iran is based on those experiments which require numerous field measurements, that makes it expensive, difficult and sometimes impossible. On the other hand; the final data is available much later than when field managers need them. Remote sensing data can be used as an alternative or complementary data for meteorological data to estimate the yield, drought and crop vegetation. To evaluate the soybean yield and its relationship with drought risk potential in western parts of Golestan province (AQ-Qala, Aliabad Katoul, Gorgan, Bandar-e-Gaz, Bandar-e-Torkaman and Kordkouy), the meteorological (SPI) and Landsat satellite imagery-based vegetation indices (including NDVI, VCI and DSI) were used.

In this study Landsat satellite images from 2000 to 2016 were used. After appropriate pre-processing and processing, the vegetation indices were prepared. Also the meteorological drought index (SPI) was calculated using the weather stations data of the study area. To calculate the relationship between the yield and all aforementioned studied indices (NDVI, DSI, VCI and SPI indices), the averaged-DN average for each index in each city was calculated. Then, the yield of each city was regressed against the meteorological (SPI) and satellite-imagery based indicators (NDVI, DSI, VCI). The linear function with the highest signifsignificant determination coefficient was selected and the soybean yield map was provided for the study area. The final drought risk model was adopted using drought frequency maps for SPI, DSI and VCI indices. The yield change in all drought risk categories was assessed through investigating the compatibility of the Boolean-classified yield maps and drought risk maps of the studied region. The

Coefficient of determination for yield prediction in different years ranged from 0.13 to 0.52, also the most of the predicted values were put in confidence level of 15% range of discrepancy which proved the accuracy of used equations in predicting yield maps. It can be concluded that yield can be predicted in a precise and accurate manner at the peak of soybean vegetation growth by applying meteorological and satellite-imagery indicators. The results showed that the yield of 43% to 50% of the soybean fields was higher than reported mean yield. Results showed that soybean fields in the AQ-Qala, Aliabad Katoul, Gorgan, Bandar-e-Gaz, Bandar-e-Torkaman and Kordkouy counties were not classified as very severe drought risk areas. The compliance of the yield and drought risk maps indicated that the possibility of yield reduction in facing with drought is higher in those areas with higher drought risk.

Totally, the results showed that Bandar-e-Torkaman, Gorgan and the central and southern parts of Aliabad Katoul are facing a lower risk of drought. The intersected map of yield and drought risk can be used as a predictive tool to provide strategies to manage drought risks as well as coping with drought effects on the yield.