Predicting Wheat Yields in Pakistan Using the LSTM Model Based on SPEI and Water Availability Data

This study examines the impact of drought on the agricultural productivity in Pakistan, focusing on wheat yield prediction. Using the Long Short-term Memory (LSTM) model, the research integrates the Standardized Precipitation Evapotranspiration Index (SPEI) and water availability data to assess the effects of seasonal and annual droughts. Drought trends from 1990 to 2020 reveal severe dry periods, particularly those in 2001, which led to crop failures despite a steady increase in wheat yields. Water availability has remained stagnant, highlighting the challenge of maintaining growth with limited resources. A correlation analysis shows a strong negative relationship between drought severity, water availability (–0.95), and wheat yield (–0.98). The LSTM model outperforms other ones, including ARIMA, Linear Regression, and Random Forest, with an R² score of 0.8, explaining 80% of the variance in wheat yield. The findings align with the Sustainable Development Goals, particularly Zero Hunger and Climate Action, by promoting sustainable agricultural practices and improving food security.