Título:	Mapping anti-hail net systems in apple orchards using multisensor time series and machine learning.
Autoria:	FURUYA, D. E. G. BOLFE, E. L. PARREIRAS, T. C. SOARES, V. B. SILVEIRA, F. da BARBEDO, J. G. A. SANTOS, T. T. GEBLER, L.
Afiliação:	DANIELLE ELIS GARCIA FURUYA; EDSON LUIS BOLFE, CNPTIA; TAYA CRISTO PARREIRAS, UNIVERSIDADE ESTADUAL DE CAMPINAS; VICTÓRIA BEATRIZ SOARES, UNIVERSIDADE ESTADUAL DE CAMPINAS; FRANCO DA SILVEIRA; JAYME GARCIA ARNAL BARBEDO, CNPTIA; THIAGO TEIXEIRA SANTOS, CNPTIA; LUCIANO GEBLER, CNPUV.
Ano de publicação:	2026
Referência:	Remote Sensing, v. 18, n. 10, 1465, May 2026.
Conteúdo:	Apple orchards are increasingly adopting anti-hail nets to mitigate climate risks; however, these structures alter canopy reflectance and pose challenges for remote sensing. This study presents an operational framework to map apple orchards under different netting conditions in Vacaria, Brazil. Multisensor surface reflectance data from Sentinel-2 and Harmonized Landsat and Sentinel-2 were used to generate dense spectral index time series combined with field observations. Five spectral indices, Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil-Adjusted Vegetation Index (SAVI), Normalized Difference Water Index (NDWI), and Bare Soil Index (BSI), were evaluated individually and in combination within a hierarchical classification framework. Random Forest (RF) and one-dimensional convolutional neural networks (1DCNN) were applied as complementary machine learning approaches. RF showed more stable performance across hierarchical levels, while indices contributed differently depending on scale: BSI and NDVI were more effective at broader levels, whereas EVI and SAVI were critical for discriminating net colors. To our knowledge, this is the first study applying multisensor time series and machine learning to map anti-hail net systems in apple orchards.
Thesagro:	Produção Agrícola Uso da Terra Sensoriamento Remoto
NAL Thesaurus:	Land use Land cover Remote sensing
Palavras-chave:	Floresta aleatória Cobertura da terra Pomar de maçã Aprendizado de máquina Machine learning Random forest 1DCNN Harmonized Landsat and Sentinel-2 (HLS) Agricultural production
ISSN:	2072-4292
Digital Object Identifier:	https://doi.org/10.3390/rs18101465
Tipo do material:	Artigo de periódico
Acesso:	openAccess
Aparece nas coleções:	Artigo em periódico indexado (CNPTIA)