Título:	Estimating Arabica coffee productivity using Planetscope images and artificial neural networks.
Autoria:	FAGUNDES, R. B. KAYSER, L. P. BENEDETTI, A. C. P. AMARAL, L. de P. BOLFE, E. L. LEMOS, O. L.
Afiliação:	ROZYMARIO BITTENCOURT FAGUNDES, UNIVERSIDADE FEDERAL DE SANTA MARIA; LUIZ PATRIC KAYSER, UNIVERSIDADE FEDERAL DE SANTA MARIA; ANA CAROLINA PAIM BENEDETTI, UNIVERSIDADE FEDERAL DE SANTA MARIA; LÚCIO DE PAULA AMARAL, UNIVERSIDADE FEDERAL DE SANTA MARIA; EDSON LUIS BOLFE, CNPTIA; ODAIR LACERDA LEMOS, UNIVERSIDADE ESTADUAL DO SUDOESTE DA BAHIA.
Ano de publicação:	2026
Referência:	Coffee Science, v. 21, e212401, 2026.
Conteúdo:	Research on orbital remote sensing and artificial neural networks (ANN) for estimating coffee productivity has advanced in Brazil, but applications remain scarce in states such as Bahia. This study aimed to develop and validate a model to estimate Arabica coffee (Coffea arabica L.) yield using PlanetScope images and ANN, based on two commercial areas with contrasting production systems (rainfed and irrigated). Georeferenced productivity samples (2 per hectare) were collected during the 2024 harvest, and PlanetScope images were analyzed from August 2023 to May 2024. In R software, vegetation indices related to vigor, nutritional status, and water stress were calculated, and their minimum, mean, and maximum values were extracted. Most variables followed a normal distribution (Shapiro–Wilk test, p < 0.05), enabling Pearson’s correlation analysis. The best correlation was observed in Santa Vera, where the CCCI index of 08/20/2023 showed a moderate positive correlation with yield (r = 0.66; p = 0.001). After ANN training in SPSS software, the model achieved very high predictive performance in both areas, with R² and adjusted R² above 0.99, low errors (MAE = 0.031–0.072; RMSE = 0.042–0.087; relative RMSE = 0.029–0.043), and almost no bias (BIAS = 0.001; relative BIAS = 0). Among the vegetation indices, CCCI was the most relevant, followed by NDRE, NDVI, NDWI, and EVI. Overall, the findings demonstrate that combining PlanetScope imagery with ANN provides a highly accurate approach for estimating Arabica coffee productivity under different production systems in Bahia.
Thesagro:	Agricultura de Precisão Cafeicultura Coffea Arábica Sensoriamento Remoto
NAL Thesaurus:	Precision agriculture Artificial intelligence Remote sensing
Palavras-chave:	Inteligência artificial Coffee Farming
ISSN:	1984-3909
Digital Object Identifier:	https://doi.org/10.25186/.v21i.2401
Tipo do material:	Artigo de periódico
Acesso:	openAccess
Aparece nas coleções:	Artigo em periódico indexado (CNPTIA)