Computational models to improve surveillance for cassava brown streak disease and minimize yield loss

Cassava brown streak disease (CBSD) is a rapidly spreading viral disease that affects a major food security crop in sub-Saharan Africa. Currently, there are several proposed management interventions to minimize loss in infected fields. Field-scale data comparing the effectiveness of these interventions individually and in combination are limited and expensive to collect. Using a stochastic epidemiological model for the spread and management of CBSD in individual fields, we simulate the effectiveness of a range of management interventions. Specifically we compare the removal of diseased plants by roguing, preferential selection of planting material, deployment of virus-free ‘clean seed’ and pesticide on crop yield and disease status of individual fields with varying levels of whitefly density crops under low and high disease pressure. We examine management interventions for sustainable production of planting material in clean seed systems and how to improve survey protocols to identify the presence of CBSD in a field or quantify the within-field prevalence of CBSD. We also propose guidelines for practical, actionable recommendations for the deployment of management strategies in regions of sub-Saharan Africa under different disease and whitefly pressure.
Author summary Cassava is the second largest source of calories in sub-Saharan Africa and is particularly important for the poorest farmers in the region. Cassava brown streak disease is a viral disease that causes cassava tubers to rot, rendering the roots inedible. Recently, the disease has begun to spread towards major cassava growing regions in West Africa from East Africa, where it continues to cause significant yield losses. Improved approaches for disease control are needed to enable small-holder farmers to prepare for and minimize the impact of the disease when their fields become infected. Using a combination of computational methods and mathematical models enables us to screen a much larger range of potential treatments for their likely effectiveness in managing disease and reducing crop loss than would be possible in conventional field trials, which are expensive and logistically difficult to conduct. Our results indicate that regularly planting part of the field with virus-free cassava greatly improves the yield. Removing visibly infected plants and replanting using visibly uninfected plants also improves yield, even when some of these plants may be infected but not yet showing symptoms. We also show how the survey protocol can be optimized to improve estimates of disease severity leading to more effective tailored advice to farmers in regions with different disease pressures.