<scp>SARIMA</scp> ‐Random Forest Framework for Forecasting Anthracnose Severity in Bottle Gourd Under Variable Transplanting Dates

ABSTRACT Anthracnose, caused by Colletotrichum lagenarium , is an economically important disease affecting bottle gourd. This study aimed to evaluate the influence of transplanting time and weather parameters on anthracnose progression and to develop a forecasting framework using statistical and machine‐learning models. Field experiments were conducted during the monsoon seasons of 2023 and 2024, with four transplanting dates: 1 June, 15 June, 1 July and 15 July. Disease severity was assessed weekly on leaves and fruits along with concurrent recording of weather data. Correlation and regression analyses revealed minimum temperature as the most influential weather variables, particularly during early transplanting dates. The regression models yielded the highest explanatory power for 1 June fruits ( R 2 = 0.675), while later transplanting dates showed reduced disease pressure and lower model accuracy. To capture seasonal trends and short‐term predictability, Seasonal Autoregressive Integrated Moving Average (SARIMA) models with configuration (1,1,1) (1,1,1) [15] were applied. These models effectively forecasted disease progression, especially for July transplanting with lower mean squared errors (MSE &lt; 200). Time series decomposition showed strong seasonal and trend components in early sowings, while cross‐correlation analysis confirmed a 1–3‐week lag between weather triggers and disease expression. This study emphasises the importance of transplanting time in disease development and demonstrates the potential of combining SARIMA and random forest models for developing weather‐based early warning systems. These findings contribute to climate‐resilient crop protection strategies and can aid in timely decision‐making for anthracnose management in bottle gourd and related cucurbits.