Produtividade da soja associada as variáveis meteorológicas e espectral no Estado do Paraná: modelagem de dados de painel espacial

Abstract

Climate change affects the productivity of several agricultural crops, including soybean, which has a significant impact on Brazil’s economy. Therefore, investigating the influence of meteorological and spectral variables on soybean productivity in soybean-growing regions in the state of Paraná, Brazil is fundamental for the development of strategies aimed at increasing the productivity of this crop. Thus, the general objective of the study was to analyze the effects on soybean productivity (t ha-1) in the state of Paraná, Brazil, of measures associated with the following meteorological variables: accumulated rainfall (mm); average, minimum and maximum air temperatures (°C); potential evapotranspiration (mm); global solar radiation (MJm-2 day-1); and the spectral variable Enhanced Vegetation Index (EVI) during the phenological phases of soybean, at a spatial scale of 9 × 9 km, considering the time series from 2010/2011 to 2019/2020. The dissertation is organized in the format of scientific papers: in the first article, spatial panel data modeling was used, which considers spatial and temporal information. Measures of explanatory variables were considered in overlapping intervals during the crop’s phenological phases, and the four municipalities with the highest soybean productivity values were analyzed. In the second article, in spatial panel data modeling, measurements of explanatory variables generated without overlap during the crop’s phenological phases were considered, and the mesoregion with the highest productivity values in the state during the considered time series was analyzed. Finally, a comparison was made between the results of Article 1 and Article 2 in terms of explanatory variables generated with and without overlap in the phenological cycle. The results showed a spatial dependence of soybean average productivity in the state, meaning that the productivity of a specific virtual station (EV) is correlated with the productivity of neighboring EVs. Furthermore, the spatial autoregressive (SAR) model with fixed effects was the best-estimated model. In both articles, results showed similarity in measures associated with the precipitation variable, in the interval containing the flowering and grain-filling phase, potential evapotranspiration in measures associated with the interval close to the harvest date, and EVI with higher values in intervals close to the date of maximum vegetative development (DMDV) and near the harvest date had an impact on soybean productivity. Finally, the average temperature variable in intervals from the seeding date up to 16 days before DMDV and close to the harvest date had a negative impact on soybean productivity. Therefore, the use of spatial panel data modeling with information on meteorological variables and EVI in relation to soybean productivity may be viable for understanding soybean productivity variation in the state of Paraná.