Integrating Remote Sensing And Weather Data For Accurate Rice Yield Prediction In Punjab, India

Abstract

Rice cultivation, crucial for global food security, faces challenges due to limited land availability, climate change, and unpredictable weather patterns. Accurate prediction of rice yield before harvest is essential for informed decision-making at international, national, and regional levels. Traditional methods like Crop Cutting Experiments (CCE) are labor-intensive and time-consuming. This study explores the use of Remote Sensing (RS) and Geographic Information System (GIS) technologies along with weather parameters to predict rice yield in Bathinda, Ludhiana, and Gurdaspur districts of Punjab, India, for the years 2019 and 2020. Spectral indices (NDVI, LSWI) from Sentinel-2 MSI satellite data, along with weather parameters (temperature, rainfall) and crop length period, were utilized as independent variables, while CCE yield served as the dependent variable. Stepwise regression analysis was employed for yield prediction.

Results indicate varying influences of independent variables on rice yield across districts and years. Bathinda consistently exhibited higher model accuracy, followed by Ludhiana and Gurdaspur. The models showed improved performance in predicting crop yield in 2020 compared to 2019. Statistical parameters such as RMSE, NRMSE, MAE, and MAPE were used to assess model accuracy, with Bathinda consistently demonstrating superior performance. Scatter plots of predicted versus observed yields illustrate the model's capability to approximate actual yields, with around 55% and 49% of data points aligning closely for Bathinda in 2019 and 2020, respectively. Similarly, Gurdaspur and Ludhiana districts showed trends aligning with observed values, albeit with slightly lower percentages.

In summary, the study demonstrates the potential of RS and GIS technologies, coupled with weather parameters, in accurately predicting rice yield, aiding stakeholders in decision-making processes related to agriculture, market dynamics, and resource allocation.