A weather-informed decision tree framework for predicting stripe rust outbreaks in wheat

Wheat stripe rust (caused by Puccinia striiformis ) is highly influenced by weather conditions making weather based predictive modelling an essential tool in combating this disease. Keeping this in mind, a weather-based decision tree approach has been applied to predict stripe rust severity in Ludhiana district of Punjab. The meteorological data for the period (from 2009–10 to 2022-23) were collected from the Department of Climate Change and Agricultural Meteorology, PAU, Ludhiana. Weekly observations of yellow rust symptoms on crop leaves were recorded using modified Mannar’s scale for yellow rust of wheat. Area under disease progression curve (AUDPC) was computed for rust severity among different years. The year 2012–13 recorded the highest rust severity (100%) and AUDPC (671.3) marking it as a major epidemic year. In contrast, lowest disease severity was recorded in 2021-22 (56.75%, AUDPC 222.85) and 2022-23 (58.42%, 244.95). Correlation analysis and Principal Component Analysis (PCA) were carried out between primary and derived variables and yellow rust severity and AUDPC. Using Classification and Regression Tree (CART) analysis, decision tree models were developed which conclude that maximum stripe rust severity (79%) can occur when minimum temperature ≥ 9.1 °C and sunshine hours ≥ 9.2 h, or when mean temperature ≥ 15 °C along with humid thermal index (HTI) ≥ 2.4. Yellow rust severity remained moderate (34-39%) when Tmin 9.1 °C, Tdew ≥ 6.2 °C and RHm ≥ 94%. Various accuracy metrices were used to compare developed decision tree models and all-weather variable model surpassed the performance than primary and derived weather variable models. Such models can be helpful in real time disease risk assessment and crop agro-advisories.