Influência ambiental na produtividade de clones de Eucalyptus e mapeamento das suas adaptabilidades

Abstract

The continental character of Brazil contains a wide range of edaphoclimatic conditions. As plant production has a quantitative distribution, environmental heterogeneity represents difficulties for plant breeding due to the interaction of genotypes by environments (GxE). Thus, the knowledge of the environmental influences on phenotype expression is important for genetic breeding. This knowledge can indicate strategies to increase the predictive capacity of the effects of the GxA interaction and, therefore, improve the evaluation and efficiency of cultivar recommendation. The objective of the study was to quantify the impact of the environmental variability in eucalyptus production in Brazil and map the adaptability of commercial clones based on geographic information system (GIS). A dataset of 13,483 forest inventory data points of six commercial clones was provided by Suzano S.A. and encompassed the four main plantation units of the company (Mato Grosso do Sul, São Paulo, Bahia e Maranhão). The 13,483 was reduced for 2,262 mean productivity values. Partial Least Square (PLS) regressions were used to model the productivity of the six Eucalyptus clones using the environmental covariables. Soil data were obtained from the SoilGrids database. Climatic data from three databases (INMET, NASA Power, and WorldClim) were compared by their predictive capabilities for each clone using the leave-one-out method. The metrics evaluated were coefficient of determination (R²); mean predictive error (RMSE); and concordance index. WorldClim was considered the most suitable bank for three clones considering R² and all of them considering RMSE and therefore selected for predictions. Productivity maps were plotted for each clone with a spatial resolution of ~5 km2. The similarity analyses showed that Maranhão unit presented the most distinct environments and, therefore, the lowest proportion of adequate predictions. Consequently, the clone with the highest proportion in the region (CLZ003) was harmed and removed from the last version of the winning genotype map. Within each forest unit, the most recommended genotypes were clone CLZ005 in Bahia (29.99%) and São Paulo (57.35%), CLZ001 in Mato Grosso do Sul (50.18%) and CLZ004 in Maranhão (77.29%). The covariates that most affected the performance of the clones were annual rainfall (BIO12), rainfall of the driest month (BIO14), rainfall of the driest quarter (BIO17), maximum temperature of the hottest month (BIO5), and mean temperature of the wettest quarter (BIO8). The methodology allowed the identification of important edaphoclimatic factor evolved in GxE interaction of different genotypes and represents a resource for potential advance in the recommendation of cultivars.