Beyond Pollination: Computer Vision Reveals how Flower Visitors, Climate and Agroforestry Management Drive Cocoa Yields in China and Brazil

Abstract

Cocoa (Theobroma cacao L.) is a multi-billion-dollar crop that is strongly affected by climate change. As a pollination-limited crop, improving little-understood pollination services in agroforestry systems may offer a scalable solution to increase yield sustainably in a changing climate. Here, we use embedded computer vision devices and structural equation models to quantify the interactions of flower visitors and temperature across shade tree-diversity and canopy cover gradients in cocoa systems on fruit set, as a yield precursor. In China and Brazil we show that flower visits are done by nectar and pollen foragers (23.1%), herbivores (6.1%), predators (2.3%), and visitors combining the three functions (63%). Forager visits were driven by increased shade-tree diversity and canopy cover management, with stronger effects in China than in Brazil. Foraging midges in China and multifunctional ants in Brazil enhanced fruit set, showing that diverse pollinators across continents affect cocoa yields. In China, higher canopy cover reduced aphids foraging and feeding on flower tissue, while in Brazil temperature increase reduced flowering. Overall, new technologies can guide implementation of agroforestry management strategies to enhance pollination while reducing pest pressures to ensure sustainable cocoa production under climate change.