Modeling risk of Sclerotinia sclerotiorum-induced disease development on canola and dry bean using machine learning algorithms

Abstract

Diseases caused by the fungus Sclerotinia sclerotiorum are managed mainly through fungicide applications in canola and dry bean. Accurate estimation of the risk of disease development on these crops could help farmers make spraying decisions. Five machine learning (ML) models were evaluated in classification and regression modes for predicting disease establishment under different air temperatures and leaf wetness duration conditions. Model algorithms were trained and tested using 20-fold cross validation. Correspondence between predicted and observed values were measured using Cohen’s Kappa (classification) and Lin’s concordance coefficients (regression). The artificial neural network (ANN) algorithms had average accuracies ≥ 89% (classification) and R2 ≥ 88% (regression) on canola and dry bean and their correspondence agreements were ≥ 0.83, which is considered substantial to almost perfect. In contrast, logistic regression algorithms had accuracies of 88% for dry bean and 78% for canola; other models were similarly inconsistent. Implementation of ANN models in disease warning systems could help farmers with spraying decisions. At the same time, these models provide insights on temperature and leaf wetness requirements for development of S. sclerotiorum diseases in these crops. Results of this study show the potential of ML models as tools for epidemiological studies on other pathosystems.

Mitigating against Sclerotinia Diseases in Legume Crops: A Comprehensive Review

“…These various stages embrace biochemical, metabolic, physiological, and morphological transformations, and are influenced by both endogenous and exogenous challenges-for example, ultraviolet radiation, temperature, toxic metals, microbial rivals, and competitors [35,36]. A number of factors, such as temperature, humidity, and wet incubation, are implicated in the germination of sclerotia, as well as the development of ascospore [37][38][39].…”

“…For instance, sclerotia will yield apothecia when conditions (moisture and light) are met. Again, S. sclerotiorum ascospore will germinate and cause infection under leaf wetness [39]. The spores released by the apothecia infect the flowers, and the infection is promoted by the plant's canopy [57].…”

“…Notwithstanding this, fungicide application is successful and effective in managing S. sclerotiorum [76,180], especially during a disease outbreak in the field. Moreover, the application of machine learning could aid to estimate the S. sclerotiorum disease threshold, to inform spraying decisions [39].…”

Application of image processing and transfer learning for the detection of rust disease

“…Numerous studies have verified the efficiency of machine learning analyses in plant pathology. These studies are either focused on training disease prediction models using environmental factors (i.e., temperature, humidity, and wetness duration) 9 – 12 or the detection of plant diseases using image processing and machine learning 13 . The high accuracy of conventional machine learning algorithms such as artificial neural network (ANN), random forest (RF), support vector regression (SVR), multi-layer perceptron (MLP), extreme learning machine (ELM), and logistic regression (LR) in distinguishing healthy and infected samples has been reported repeatedly 9 .…”

A Hybrid Model for Leaf Diseases Classification Based on the Modified Deep Transfer Learning and Ensemble Approach for Agricultural AIoT-Based Monitoring

“…Shahoveisi et al [ 38 ] used ML techniques to model the risk of Sclerotinia sclerotiorum -induced disease development in canola and dry beans. Using a broad genome correlation study [ 39 ], they previously examined the genomic regions associated with Leptosphaeria maculans resistance in rapeseed.…”