Food Web Modelling for Agroecological System Resiliency and Design

Abstract:

Agricultural models are key tools for predicting crop yields and planning best management practices worldwide, especially in the context of agricultural resiliency towards climate change. The vast majority of crop models developed to date, however, were designed for cropping systems typical of larger-scale industrialised agriculture in countries such as Canada, the USA or Europe. These mostly consider only abiotic impacts on crop growth and yield (soil nutrient and water dynamics, weather, human management choices and plant genotype), while processes central to small-scale traditional agriculture, such as agroecosystem food web dynamics and multicroping, remain technically difficult to implement in a crop modelling environment. Given the very substantial contribution of Indigenous, traditional, and small-scale agriculture to human nutrition worldwide, the development of modelling tools to better represent these very complex systems would be beneficial and could contribute to a better understanding of and respect for these communities' technologies and approaches to crop management, in addition to facilitating opportunities for climate adaptation and technology transfer between communities.

As existing crop models, at most, only consider pest pressures as exogenous variables, we present a new model, Tiko'n ("Tee-ko-'n", from Kaqchikel), that standardises and greatly simplifies the construction, as well as calibration and validation, of full agroecological food web models that include the dynamic trophic relationships between predators and herbivores that ultimately determine pest pressures on the crop. As Tiko'n is first and foremost an agricultural food web model (and not a crop model in itself), it also includes an API for dynamic runtime coupling with existing crop models such as PCSE or DSSAT. An example is given of the application of Tiko'n to biological pest control scenarios from coconut crops in ශ්රී ලංකාව-இலங்கை (Sri Lanka). Results suggest that model-optimised biocontrol strategies may generate better pest control with lower risk of failure than either fixed-date or economic action threshold approaches.