Identifying landscape targets for effective conservation biological control in arable landscapes: insights from an ecological – economic simulation and machine learning analysis

Abstract: Conservation biological control (CBC) is promoted as an alternative to pesticidebased pest management, but its effectiveness depends strongly on the agroecological and economic context. We use the spatially-explicit, dynamic ecological – agronomic – economic model of Martinet and Roques (2022) to quantify CBC at a landscape for more than 400 k parameter combinations. Using clustering on profit gain and treatment frequency index reduction, we isolate “high-performing” landscapes. We then train Random Forests to identify the combinations of ecological and agronomic parameters that characterise them. The analysis highlights the key role of economic and agronomic drivers in the successful CBC, in addition to usual ecological drivers. These results will be used to identify empirical ‘agricultural region – crop – pest – enemy’ case studies where CBC can credibly support pesticide reduction without undermining profitability, and match them with effective policy tools.