Forecasting and classifying potato yields for precision agriculture based on time series analysis of multispectral satellite imagery

Organised by Laboratory of Geo-information Science and Remote Sensing

Tue 26 May 2015 09:00 to 09:30

Venue Gaia, gebouwnummer 101
Room 2

by Nikolaos Tziolas (Greece)


Since, March 2012 the National Satellite Dataportal of the Netherlands provides Disaster Monitoring Constellation (DMC) images with a time resolution of 2 days and sufficient spatial resolution of 22m, making it an ideal data source for application in precision agriculture sector. For a farm spanning between Belgium and the Netherlands as a research area data from two growing seasons (2013 and 2014) were analyzed, in order to determine if potato yield potential could be estimated utilizing an in-season estimation of normalized difference vegetative index (NDVI) and meteorological data. This research provides important insight into availability of cloud free satellite images during critical periods of agricultural growing season emphasizing that is a key for agricultural monitoring and yield prediction. In this research, inclusion of information related to crop phenology showing significantly improved model performance. Several methods used to provide recommendations for the estimation of yield at the field scale. Linear regression models developed using parameters of NDVI time series profiles were evaluated as a stand-alone yield predictor. Additionally, multivariate regression models were developed introducing bio-climatic variables (solar radiation, temperature and precipitation) in conjunction with NDVI. Beyond the regression models, decision trees were used to analyze a qualitative relationship between yield NDVI and meteorological variables. In general the results were significant and promising. The resulting yield maps provide a unique opportunity to inform agricultural management decisions. Future satellite missions should permit estimation of potato yields using image resolutions that facilitate extraction of information in more frequent times. This analysis has also described cloud cover frequency throughout the agricultural growing season, providing insight into how yield forecasting approach could be impacted by cloud cover.