INTEGRATING GIS, BIG DATA ANALYTICS, AND CLIMATE MODELING FOR PREDICTIVE AGRICULTURAL PLANNING AND RISK MANAGEMENT

Abstract

This research develops and simulates a single framework that unites Geographic Information Systems (GIS), Big Data Analytics, and Climate Modelling to plan agriculture and climate risk management.  A mixed-methods experimental design was employed that is integrative of spatial data layers of soil, crop distribution and topography with the climatic data (premeditation, temperature and evapotranspiration), and the socio-economic indicators (surveys among farmers).  Big data pipelines helped to deal with much geospatial data. Both linear and nonlinear relationships between yield changes and climate changes were found using machine learning models, including Random Forests, Gradient Boosting and Long Short-term Memory (LSTM) networks.  Regression study indicated that yields are most accurately predicted by temperature variances, rainfall, and moisture content in the soil. The predictive model was rather close to the actual values indicated by the RMSE and R 2 values.  Further quantification of yield failure probabilities in a variety of Representative Concentration Pathway (RCP) scenarios was afforded by climate downscaling and Monte Carlo simulations, leading to risk assessment of different agricultural zoning.  The introduction of qualitative farmer opinions ensured that the findings were contextually valid, and thus, were more reliable and applicable.  The findings indicate that the system will be able to produce zonal risk maps, seasonal production forecasts, and early warning alerts, which are beneficial in climate-intelligent production planning.  Overall, the study demonstrates how predictive systems based on data could enhance how we respond to climate hazards, how we make food more resilient, and how we apply to sustainable approaches to farming in a rapidly evolving world.