Spatial Analysis and Visualization of Triadic Crop Variety Trials in Central America

Organised by Laboratory of Geo-information Science and Remote Sensing

Wed 8 May 2019 09:30 to 10:00

Venue Gaia, gebouwnummer 101
Room 2

By Meng Zhang

A rapid increase in population, urbanization, and climate change has put huge pressures on all aspects of society, and in particular agriculture. These pressures lead to a massive demand of agriculture products which is increasingly difficult to meet due to climate change, or more specifically the challenges climate change introduces to farmers. Seeds for Needs is a global initiative from the Bioversity International Research Center which is aiming to help farmers adapt better to climate change through the use of agriculture biodiversity. The method employed by this Seeds for Needs initiative is called ‘Tricot’ (Triadic Comparisons of Technologies) and consists of a farm-specific comparisons between three randomly-assigned crop varieties (sampled from a large set) which are then also compared to local varieties. Numerous environmental datasets were linked in a spatially explicit way for each observation (e.g. elevation, temperature, and water balance), this was possible because all the farms were geo-located. These datasets were used to produce valuable information about the interactions between crop variety performance and crop growing environment. The method employed here included the Hargreaves method which calculated the necessary evapotranspiration, and ultimately the water balance. The Plackett-Luce statistical model was then used to predict crop varieties performance under different environmental situations using model based recursive partitioning of environmental covariates. As the data for these covariates were spatially explicit and continuous, all model results were able to be visualized by an interactive webpage which allowed farmers to assess appropriate crops for their particular farms.

Keywords: Tricot; environmental covariates; water balance; Hargreaves; Plackett-Luce model; recursive partitioning; interactive webpage