Artificial neural networks and satellite images as support for modelling biophysical properties of maize canopies

IRIGOYEN, ANDREA INÉS; RIVAS, RAÚL; BAYALA, MARTÍN; FERRELLI, FEDERICO

Goiania

Simposio; 9th International Symposium AgroEnviron2014; 2014

Leaf area index (LAI) is a key input for many crop and environmental models. The LAI patterns and other canopy properties measured in situ are time consuming and expensive. Artificial neural networks (ANNs) are data-driven approaches with recognized capability to approximate non-linear relationships. Remote sensing provides a feasible platform for spatially and temporally continuous bservations of biophysical parameters at the scales from local to global. The objective of this project is to evaluate the estimation of maize canopy properties in south-eastern of Buenos Aires, Argentine, using neural network models combined to remote sensing techniques. Periodical measurements of LAI on tagged plants at experimental sites exhibiting conditions of contrasting agronomic practices (maturity hybrids, planting date, plant density and intensification of crop production) are used to develop, evaluate and test the neural network models to approximate variations of leaf area at plot scale. Data from canopy structure properties as leaf area, height and leaf area density profile are obtained by non-destructive methods. Biophysical relationships attempting to correct crop management factors in addition to developmental model with degree-days basis will be tested. Canopy reflectance is associated to structural characteristics (leaf area index; leaf area profile) and biochemical properties (chlorophyll content). Moreover, empirical relationships between spectral indexes and leaf area are the basis to simulate temporal variations at field scale. Validation of functions will be carried over the crop production region. Combination of techniques is proposed to estimate biophysical properties in canopies, relevant topic for food production and security.