Yield prediction models for some wheat varieties with satellite-based drought indices and machine learning algorithms

In recent years, frequent drought events in Konya, one of T & uuml;rkiye's most important cereal production centres, have led to increased pressure on water and soil resources, resulting in yield losses, particularly in wheat production. Alternative yield prediction models, especially those that play a crucial role in agricultural import-export planning in the region, are important for economic contributions and the development of early warning systems. In this context, the aim of this study is to develop models that can be used in the yield prediction of wheat varieties widely grown in the Konya Alt & imath;nova region. Agricultural drought indices obtained from Normalized Difference Vegetation Index (NDVI) and land surface temperature (LST) products of the Terra Moderate Resolution Imaging Spectroradiometer (MODIS) satellite were used to obtain model inputs. These indices are the Vegetation Condition Index (VCI), Temperature Condition Index (TCI), Vegetation Health Index (VHI) and Vegetation Supply Water Index (VSWI). In obtaining the input parameters for the models, the growth periods of the varieties in the region were also considered. Using various machine learning algorithms, 21 yield prediction models for Bayraktar-2000, 12 for K & imath;z & imath;ltan-91 and 8 for Bezostaya-1 were presented as alternatives, with model performances (coefficient of determination, R2) ranging between 0.74 and 0.97, 0.73 and 0.96, and 0.69 and 0.87, respectively. & Agrave; Konya, l'un des plus importants centres de production c & eacute;r & eacute;ali & egrave;re de Turquie, les & eacute;pisodes de s & eacute;cheresse, fr & eacute;quents ces derni & egrave;res ann & eacute;es, ont accru la pression sur les ressources en eau et en sol et entra & icirc;n & eacute; des pertes de rendement des cultures, en particulier du bl & eacute;. Les mod & egrave;les alternatifs de pr & eacute;vision des rendements, qui jouent un r & ocirc;le important notamment dans la planification des importations et exportations agricoles dans la r & eacute;gion, sont importants en termes de contribution & eacute;conomique et de d & eacute;veloppement de syst & egrave;mes d'alerte pr & eacute;coce. Dans ce contexte, l'objectif de cette & eacute;tude & eacute;tait de d & eacute;velopper des mod & egrave;les qui peuvent & ecirc;tre utilis & eacute;s dans la pr & eacute;diction du rendement des vari & eacute;t & eacute;s de bl & eacute; communes dans la r & eacute;gion d'Alt & imath;nova de la province de Konya en utilisant les indices de s & eacute;cheresse agricole Vegetation Condition Index (VCI), Temperature Condition Index (TCI), Vegetation Health Index (VHI), and Vegetation Supply Water Index (VSWI) obtenus & agrave; partir des produits Normalized Difference Vegetation Index (NDVI) et Land Surface Temperature (LST) du satellite Terra MODIS (Moderate Resolution Imaging Spectroradiometer). Lors de l'obtention des param & egrave;tres d'entr & eacute;e du mod & egrave;le, les p & eacute;riodes de croissance des vari & eacute;t & eacute;s dans la r & eacute;gion ont & eacute;galement & eacute;t & eacute; prises en consid & eacute;ration. & Agrave; l'issue de l'& eacute;tude, 21 mod & egrave;les de pr & eacute;vision du rendement pour Bayraktar-2000, 12 pour K & imath;z & imath;ltan-91 et 8 pour Bezostaya-1 ont & eacute;t & eacute; pr & eacute;sent & eacute;s comme des alternatives avec des performances de mod & egrave;le (R2) comprises entre 0,736-0,973, 0,733-0,956 et 0,686-0,874, respectivement.