ClimaCare: Explainable IoT-ML Framework for Microclimate Forecasting

The increased need for precise and localized weather predictions has prompted the development of sophisticated forecasting algorithms. However, typical weather prediction techniques frequently lack the resolution and flexibility required for microclimate forecasting. Internet of Things (IoT) and Artificial Intelligence (AI) integration strongly reinforces the observation and forecasting of microclimates for agriculture and environmental applications. This study proposes an ESP32based weather mapping system with numerous low-cost sensor systems to monitor temperature, humidity, rainfall, and sunlight in real time. Dominant microclimatic parameters with machine learning algorithm-based ensemble modeling such as XGBoost and random forest were modeled with high precision. Explainable AI (XAI) tools, such as SHAP, are integrated to enhance the robustness and transparency of forecasts. They identified relative feature significance-relative sunshine, relative humidity, and windspeed-on rainfall and temperature forecasts and enabled domain experts and farmers to interpret and believe in the model forecasting better. The system was evaluated with realistic sensor datasets and was observed to possess robust predictive power with interpretable descriptions of model behaviour. The solution not only enhances microclimate prediction precision but also stimulates adoption with human-interpretable descriptions of model behavior. The integration of IoT, ML, and XAI offers an accessible, scalable, and trustworthy climate-resistant and smart agriculture decision-support systems.