INVESTIGADORES
PASTORE Juan Ignacio
artículos
Título:
Maize (Zea Mays L.) Yield Estimation Using High Spatial and Temporal Resolution Sentinel-2 Remote Sensing Data
Autor/es:
GAVILÁN, S; ACEÑOLAZA, P.G; PASTORE, J. I
Revista:
COMMUNICATIONS IN SOIL SCIENCE AND PLANT ANALYSIS
Editorial:
TAYLOR & FRANCIS INC
Referencias:
Año: 2023 vol. 54 p. 1 - 14
ISSN:
0010-3624
Resumen:
Maize (Zea mays L.) is one of the worlds most important annual cereal crops and its yield can be estimated for a wide variety of purposes. The objective of this work is to evaluate in which stage of crop the best fit between remote sensing data and real yield occurs to predict yield in corn seed crops. For this, polynomial regression models were used between spectral indices of vegetationand real yield in 10 days times windows covering the critical period for generation of performance. Subsequently, the predictive capacity of the best goodness of fit model was evaluated by comparing estimates with those made using a conventional field estimation method. This experiment was carried out in production fields located in Tandil and Loberia district inside ofthe Argentine Pampas Region in southeast of Buenos Aires province in summer (from january to march) of 2020. We found the highest level of adjustment between vegetal index and real yield (R2 = 0.91) in the time window of 110 to 120 days after sowing (DAS) corresponding to the end ofthe critical period. Then, the predictive performance was evaluated, satellite model shows an underestimation of 53 kg/ha (0.72% relative error) while the conventional method underestimated by 955 kg/ha (13% relative error). A close relationship between remote sensing data and grain yield at the end of the critical period of maize can be evidenced, and this information can be used to predict yield early in the southeast of Buenos Aires province. Using the methodology here developed it is recommended to analyze- time series of satellite vegetal index in maize crops in other regions and climates to make more robust the yield prediction system.