Indian mustard (Brassica juncea L.) is an important oilseed crop contributing significantly to the country’s edible oilseed production. Out of the total global production, India contributes 13.7 mt of produce from an area of 10.7 Mha with a productivity of 1,444 kg/ha (ICAR-DRMR, 2024). The production and productivity are often affected by both biotic and abiotic stresses among which Sclerotinia rot (SR) is one of the important disease (Boomiraj et al., 2010). It is caused by Sclerotinia sclerotiorum (Lib) de Bary, which is a globally distributed and highly destructive soil borne fungal pathogen having a broad host range (Sharma et al., 2015). The pathogen has been reported to infect more than 500 host species across a wide range of phylogenetic lineages, encompassing 278 genera and 75 dicotyledonous families along with several economically important monocotyledonous plants (Boland and Hall, 1994; Sharma et al., 2015).

In mustard-growing regions, only a small proportion of farmers (less than 10%) opt for early sowing during first week of October. The majority, however, sow their crop in the third to fourth week of October, which is considered timely sowing, as it coincides with favorable soil temperature conditions for crop growth. A few farmers, whose fields are not ready due to the delay in harvesting of the previous crop, resort to late sowing in mid-November. During its cultivation, the disease poses a significant threat particularly in major mustard-growing states such as Rajasthan, Madhya Pradesh, Haryana, Uttar Pradesh, West Bengal, Bihar, and Gujarat (Aggarwal et al., 1997). In Rajasthan state, which accounts for 48% of the area and 47% of the production of the country, SR has been reported to cause up to 40% yield loss when the disease incidence reaches 60% during the crop maturity stage.

The pathogen S. sclerotiorum belongs to the ascomycetes group. The infection is initiated through two primary sources of inoculum, namely, airborne ascospore and soil-borne hyphae. The development of apothecia requires continuous soil moisture approximately for 10 days, whereas even slight moisture stress inhibits their formation. Additionally, infection by ascospores typically requires 2–3 days of uninterrupted leaf wetness, underscoring the crucial influence of environmental conditions on disease initiation and progression. Furthermore, in the context of B. juncea, petal infestation and soil moisture have been recognized as critical parameters for inclusion in the predictive disease model. Additionally, increased relative humidity (RH) and soil moisture during flowering significantly influences SR incidence (Sharma et al., 2009). The ascospore germination, subsequent mycelial growth, and lesion initiation and development are key factors necessary for the onset of epidemics (Abawi and Grogan, 1975).

Sharma et al. (2010) documented the infection of petals by ascospores during the full bloom stages and highlighted rainfall as a crucial factor in the development of carpogenic infection of S. sclerotiorum in B. juncea. Notably, ascospores do not germinate upon landing on leaf surfaces; however, when they come in contact with petals, they germinate, colonize the petal tissues for their nutrition, and subsequently infect the plant upon contact with the leaves (Jamaux and Spire, 1994). The amount and distribution of rainfall during these stages were also identified as critical factors in creating microclimate within the canopy that favored disease development. In a subsequent study, Hall and Mwiindilila (2000) provided quantitative insights into the magnitude and duration of key epidemiological factors such as canopy density, flowering, moisture, and apothecia germination within the field, all of which were found to be closely associated with the development and progression of the disease. Predictions of SR based on early bloom petal infestation were generally accurate under conditions of low disease risk and incidence. However, their reliability decreased when disease risk and incidence levels were moderate to high (Turkington et al., 1991a). Forecasts of SR based on petal infestation during early bloom were generally accurate under low disease risk and incidence but become less reliable under moderate- to high-risk conditions (Turkington et al., 1991a). Crop canopy density and rainfall, which influence ascospore production and release, were found to affect the correlation between petal infestation and Sclerotinia disease incidence (Turkington et al., 1991a; Hall and Mwiindilila, 2000).

Despite the recognized economic importance of SR in Indian mustard, there is a scarcity of quantitative, long-term field-based epidemiological studies conducted under Indian agro-climatic conditions, and most of the existing literature are limited to short-term observational trials. Secondly, weather-based forecasting models for SR remain underdeveloped and are rarely customized for mustard cropping systems in India. In addition, there is limited understanding of how different sowing windows influence disease development over multiple crop seasons. This study is premised on the hypothesis that the SR incidence is significantly influenced by specific weather variables, namely, temperature, RH, rainfall, and bright sunshine hours (BSSH), particularly during the flowering period. These agrometeorological parameters when monitored over time and across sowing windows can be used to develop statistically reliable models for forecasting disease incidence. In light of these, the present study aims to quantify long-term trends in SR incidence under three major sowing windows in Indian mustard cultivation, to examine the relationship between petal infestation by S. sclerotiorum and subsequent field-level disease incidence under natural conditions, to develop and validate predictive models using both weather-based weighted indices and machine learning techniques for accurate forecasting of SR incidence, and, lastly, to project future disease risk scenarios for the period 2025 to 2030, based on historical weather trends and model output, thereby assisting in strategic planning for disease avoidance through optimized sowing schedules.

The present investigation establishes a robust framework for SR of Indian mustard through comprehensive analysis of weather–disease interactions across 14 crop seasons (2009–2023). The multivariate and non-linear relationship between climatic variables and disease manifestations demonstrates that SR is governed by weather parameters like T _max, T _min, RH, wind speed, and BSSH, consistent with the findings of Goswami et al. (2012) and Sharma et al. (2015). The identified optimal temperature range of 16–20 °C for maximum disease expression aligns with the physiological requirements for apothecia development and ascospore release reported by Sun and Yang (2000), who demonstrated that sclerotial development accelerates under low light intensity conditions at temperatures exceeding 20 °C. The critical role of RH in disease progression, with a morning RH threshold of 85% and an afternoon RH of 76.6%, corroborates the moisture-dependent nature of Sclerotinia pathogenesis. Chattopadhyay et al. (2007) similarly reported that enhanced petal infestation by ascospore under reduced BSSH supports our observation that BSSH below 5 h creates a favorable condition for disease development. The episodic influence of rainfall, despite showing no consistent linear correlation with disease severity, likely enhances microclimatic wetness essential for sclerotial germination and apothecial formation as documented by Gugel (1986) and Boland and Hall (1988a, b). These apothecia formed from sclerotinia when they receive the ideal temperature, RH, and BSSH. Apothecium development occurs at a faster pace when low light intensity greater than 20°C was available (Sun and Yang, 2000).

The CWI integration of temperature and humidity dynamics provides a holistic approach to disease risk assessment. Years with elevated CWI values corresponded with peak disease incidence, while reduced CWI values have subdued disease levels. Interpretability analyses using PDPs and SHAP summaries further confirmed these insights. Afternoon RH displayed a sharp non-linear increase in disease risk once values exceeded 75%–80%, while BSSH cutoffs below 4 h/day consistently predicted higher incidence. Maximum temperature exerted a threshold influence, with the 18–20°C band marking peak severity. These model-estimated thresholds aligned precisely with observed petal infestation peaks in the 29 October sowing window, thereby strengthening the epidemiological linkage of canopy microclimate, and petal colonization leads to epidemic development. The integration of weekly petal infestation data into an AUPPC metric provided an additional, epidemiologically integral measure of inoculum pressure. Across sowing windows, the AUPPC succinctly summarized epidemic potential, with the October 29 sowing showing threefold greater cumulative petal infestation pressure compared to 19 November. This integral measure correlates directly with field disease incidence and offers a valuable complementary index to weather-based predictors, enabling comparison of sowing windows on a single, interpretable scale. This integrated approach addresses the limitation of single-parameter models and provides a more robust foundation for predictive modeling as advocated by Turkington and Morrall (1993). However, it is important to note that point forecasts may obscure inherent uncertainty. By incorporating multiple stochastic realizations, our ensemble forecasts produced uncertainty bands (50%, 80%, and 95% intervals) around the prediction trajectories. While the central tendency suggested generally low baseline disease pressure (<2% mean incidence), the uncertainty bands reveal occasional possibilities for localized outbreaks (>8%). Assumptions in the weather generator, particularly regarding extremes of T _max and humidity persistence, may modestly bias outbreak probability estimates. Explicit communication of these uncertainty envelopes provides a more balanced risk outlook for farmers and policy planners than point predictions alone.

The comparative analysis of three showing windows reveals distinct disease risk profile that fundamentally alters the management strategy. The 29 October sowing window emerges as the highest-risk period, with a mean incidence of 5.89% and a maximum value reaching 30.13%, attributed to the convergence of optimal pathogenic conditions like moderate maximum temperature, exceptionally high RH, and consistently low BSSH. This risk profile aligns with the findings of Sharma et al. (2010), who documented apothecial presence under similar conditions. The intermediate risk profile of an 8 October sowing demonstrates partial disease escape through suboptimal temperature conditions that frequently exceed the critical 20 °C threshold. The moderately elevated BSSH value occasionally surpasses the critical 5-h period threshold providing intermittent unfavorable condition for pathogen development. This temporal disease escape mechanism represents a practical compromise between agronomic requirements and disease management objectives. The 19 November sowing consistently exhibited the lowest disease risk representing 6.4-fold reduction compared to the 29 October window. The reduced RH and elevated BSSH create consistently unfavorable conditions for pathogen establishment and proliferation. The higher average temperature further contributes to disease suppression through thermal stress on the pathogen, supporting the disease’s escape strategy documented in late-sown crops (Turkington, 1988). In this context, the prediction model developed using environmental and weather variables can provide reliable forecasts and timely warnings, enabling farmers to make informed decisions and stay prepared. The present study also reinforces that petal-based forecasts, while valuable, achieve maximum reliability when integrated with contemporaneous weather-based indices, as demonstrated in Canadian and European forecasting systems (Turkington et al., 1991a: Hall and Mwiindilila, 2000). Although there is reduction in the yield of seeds and vegetative growth (Patel et al., 2017; Pattam et al., 2017) when there is a delay in sowing, this yield reduction can be compensated if there is lesser incidence of SR in Indian mustard. Early sowing hastens the reproductive period (Tripathi and Singh, 2007). When the mustard crop is sown early, it reaches its flowering period earlier than the ascospores are released. Thus, it avoids the overlap between the petals that are available and the load of ascospores on B. juncea. In this case, there is a chance that ascospores were present, but due to the limited number of petals, the chances of the incidence lessened. A slight delay in sowing time reduces the chances of Sclerotinia incidence as the flowering occurs a bit later than the period of maximum ascospore release. Thus, there will be fewer chances of petal infestation and stem infection (Jurke, 2003). Percent Sclerotinia incidence peaked with maximum T _max and evening RH. This was similar to the results recorded by Singh et al. (2022). This study’s data also supports that specific weather combinations Temperature between 18–20°C, morning, RH≥ 94%, and BSSH< 4h are critical thresholds for epidemic development. Such findings align with multi-year epidemiological analyses in oilseed rape, which have consistently linked dense canopies, low irradiance, and high humidity during boll development to severe Sclerotinia outbreaks. Such findings align with multi-year epidemiological analyses in oilseed rape, which have consistently linked dense canopies, low irradiance, and high humidity during bloom to severe Sclerotinia outbreaks (Twengstrom et al., 1998; Clarkson et al., 2004).

The random forest ensemble approach demonstrates superior predictive capability with training and test R ², maintaining robust generalization with minimal overfitting. The model’s ability to capture complex non-linear interaction between weather parameters and disease development addresses the limitation of traditional linear regression approach. The Ultra ensemble mid-model’s exceptional performance for mid-sowing conditions indicates that the 29 October window provides the most predictable disease–weather relationships, likely due to consistent environmental conditions favoring pathogen development. The sowing-specific model performance variations reflect the underlying biological complexity of host–pathogen–environment interactions. Early sowing shows high sensitivity to environmental variations requiring sophisticated ensemble methods, while late sowing benefits from natural escape mechanisms that reduce model complexity requirements. This differential predictive accuracy has practical implications for targeted advisory systems, where high-risk periods require more frequent monitoring and intervention.

The 2025–2030 projection indicates stable disease risk ranking across sowing windows with minimal year-to-year variation within each window. This seasonal predictability enables precision agriculture approaches, where fungicide application and monitoring efforts can be concentrated during high-risk periods, optimizing resource utilization and minimizing environment impacts. The identification of critical weather thresholds enables the development of real-time risk assessment protocols. Integration of these thresholds with operational agrometeorological advisories can form the farmer-friendly decision support system for major mustard production regions. The petal sampling methodology, validated through correlation with the disease incidence, provides an early warning system that complements weather-based prediction. The study demonstrated that optimal environmental conditions must align with adequate inoculum pressure for significant disease development, emphasizing the importance of integrated management approaches. Combining predictive model with field monitoring enables targeted fungicide applications, reducing unnecessary treatment and associated cost while maintaining effects to disease control. Future research should focus on and develop a weather-based advisory system that integrates real-time weather data with predictive models, providing farmers with actionable recommendations for sowing date optimization and intervention timing. The establishment of this comprehensive prediction system enhances the precision and effectiveness of SR management in Indian mustard, contributing to sustainable crop production and food security objectives in the Indian subcontinent.

The long 13-year field trial in the experimental field of ICAR-IIRMR, Bharatpur leveraged precision agriculture analytics to unravel how sowing time and weather synergistically govern SR in Indian mustard. Using a split-plot layout (4.8 × 5 m, four replications) of cv. DRMR IJ-31 sown on 8 October, 29 October, and 19 November, researchers coupled weekly disease surveys, laboratory-confirmed petal infestation tests, and meteorological records that included temperature, humidity, rainfall, and BSSH captured 100 m from the plots. They transformed raw weather variables into simple and correlation-weighted indices, then built stepwise multiple-regression and modified quadratic models to quantify disease–weather relations, validating them with an independent 2022–2023 dataset. A random forest ensemble extrapolated these relations to 2025–2030, revealing that SR risk peaks due to two major factors, which are maximum temperature and afternoon RH.

Conclusively, the study demonstrates that integrating long-term micro-meteorological monitoring with correlation-weighted indices and machine learning forecasts can pinpoint sowing windows and weather thresholds that tip the balance between epidemic and escape. BSSH, maximum temperature, and RH emerged as the most reliable, easy-to-measure predictors; when monitored in real time, they can trigger site-specific advisories on sowing date adjustment, canopy management, and timely fungicide application. By coupling such decision rules with routine petal sampling, growers can shift from calendar-based to data-driven management, minimizing losses to SR while preserving the yield gains of optimum planting schedules—an actionable template for precision plant-protection services across India’s oilseed belts.