Estimating soil moisture and the relationship with crop yield using surface temperature and vegetation index

- We estimated soil moisture conditions and crop yield with satellite data.
- Ancillary data are not required for yield estimation with the developed methodology.
- Crop yield forecasting was achieved up to three months before harvest.
- The generalized model could be applicable in other regions.
- The method is useful for food security and decision making.

Soil moisture availability affects rainfed crop yield. Therefore, the development of methods for pre-harvest yield prediction is essential for the food security. A study was carried out to estimate regional crop yield using the Temperature Vegetation Dryness Index (TVDI). Triangular scatters from land surface temperature (LST) and enhanced vegetation index (EVI) space from MODIS (Moderate Resolution Imaging Spectroradiometer) were utilized to obtain TVDI and to estimate soil moisture availability. Then soybean and wheat crops yield was estimated on four agro-climatic zones of Argentine Pampas. TVDI showed a strong correlation with soil moisture measurements, with R2 values ranged from 0.61 to 0.83 and also it was in agreement with spatial pattern of soil moisture. Moreover, results showed that TVDI data can be used effectively to predict crop yield on the Argentine Pampas. Depending on the agro-climatic zone, R2 values ranged from 0.68 to 0.79 for soybean crop and 0.76 to 0.81 for wheat. The RMSE values were 366 and 380 kg ha−1 for soybean and they varied between 300 and 550 kg ha−1 in the case of wheat crop. When expressed as percentages of actual yield, the RMSE values ranged from 12% to 13% for soybean and 14% to 22% for wheat. The bias values indicated that the obtained models underestimated soybean and wheat yield. Accurate crop grain yield forecast using the developed regression models was achieved one to three months before harvest. In many cases the results were better than others obtained using only a vegetation index, showing the aptitude of surface temperature and vegetation index combination to reflect the crop water condition. Finally, the analysis of a wide range of soil moisture availability allowed us to develop a generalized model of crop yield and dryness index relationship which could be applicable in other regions and crops at regional scale.

131