Estimation of geographical variations in virtual water content and crop yield under climate change: comparison of three data mining approaches

One of the most crucial issues in water and food security is to assess the impacts of climate change on virtual water content (VWC) and crop yield of agricultural products. The objective of this study is to efficiently predict the VWC patterns and yields of different crops under various climate change conditions using three data mining approaches including artificial neural network (ANN), genetic programming (GP), and support vector machine (SVM). The study region included the eastern provinces of Iran containing North Khorasan, Khorasan Razavi, South Khorasan, Sistan-Baluchestan (in a range of latitude from 25 to 40°N). Specifically, VWC and crop yields were estimated for both baseline period (1985–2005) and several climate change conditions including four time horizons (2030, 2050, 2070, and 2090) under RCPs 2.6, 4.5, and 8.5 based on the second generation Canadian Earth System Model (CanESM2). The data mining models were evaluated with the RMSE and NSE goodness-of-fit criteria. The results showed that the SVM model achieved the highest NSE and lowest RMSE values. It was also found that under the climate change conditions, VWC increased from 6 to 42%, while crop yield decreased from 8 to 53% for all products in the southern regions. An opposite trend was observed in the northern regions for wheat and barley with an increase from 12 to 72% for VWC and from 4 to 27% for the yield.