Time Series Forecasting Of Crop Yield Under Climate Variability Using Iot Data And LSTM Networks

Abstract

Climate variability poses a significant challenge to agricultural productivity and food security. Accurate crop yield forecasting under changing weather patterns is crucial for proactive farm management and policy planning. This study integrates Internet of Things (IoT) based environmental sensing with Long Short-Term Memory (LSTM) deep neural networks to predict crop yields. We leverage real-world datasets combining IoT sensor data (temperature, humidity, soil moisture, rainfall, etc.) and historical crop yields from agricultural fields, as well as global datasets of climate indices and yields. The LSTM model is developed to capture temporal dependencies in multi-variant time series climate data and forecast end-of-season crop yields. We present the architecture of the LSTM network and actual Python code snippets used in model development. Experiments are conducted on two levels: (1) a local farm-level IoT dataset with high-frequency sensor readings and seasonal yield observations, and (2) a global historical dataset (e.g., FAO and World Bank data) of annual crop yields with climate variables across multiple countries. The LSTM-based approach is evaluated against baseline models (including linear regression and classical time-series models), demonstrating improved prediction accuracy. Results show that the LSTM achieves a lower mean absolute percentage error (MAPE) than baselines (e.g., ~12% vs 18% on the local dataset), indicating its superior ability to learn complex climate–yield relationships. We include tables summarizing the datasets and model performance metrics, and figures such as the LSTM network architecture and predicted vs. actual yield plots. This research highlights the potential of IoT-driven data combined with deep learning to enhance crop yield forecasting under climate variability. The findings can help farmers and decision-makers to mitigate climate risks, optimize resource use, and improve sustainability in agriculture.