Early prediction of durum wheat yield in Spain using radial basis functions interpolation models based on agroclimatic data

• Autor(a) principal: Rocha, H.
• Data de Publicação: 2019
• Outros Autores: Dias, J. M.
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10316/88872; https://doi.org/10.1016/j.compag.2019.01.018
• topic: Radial Basis Functions; Cross-validation; Variable Screening; Durum wheat

Durum wheat is an important crop in semi-arid Mediterranean regions as Andalusia, an autonomous community in the southern part of Spain. Accurate early predictions of durum wheat yield can provide precious information for within-season adjustment of crop managing as well as for economical and political stakeholders. In this study, an alternative methodology to mechanistic crop models is proposed for within-season early prediction of durum wheat yield in Spain based on estimates for its larger producer community, Andalusia. The proposed mathematical framework embeds the construction of Radial Basis Functions (RBF) interpolation models based on the sown area and a large number of climatic variables. Global warming and increasing occurrence of extreme weather events are only two of the factors that make crop yield forecast extremely difficult as they can lead to an increased interannual yield variability. Nevertheless, the RBF models proposed presented good quality yield predictions clearly outperforming multivariate linear models used as benchmark. Moreover, RBF models' predictions made four months prior to harvest are able to capture the trend of the yield series as well as near-harvest predictions.