Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2013, Vol. 10, No. 1, pp. 215-227
Comparative analysis of regression and biophysical models for the problem of forecasting the yield of winter wheat
F. Kogan
1, N.N. Kussul
2, T.I. Adamenko
3, S.V. Skakun
2, A.N. Kravchenko
2, А.А. Krivobok
4, A.Yu. Shelestov
5, A.V. Kolotii
2, O.M. Kussul
6, А.Н. Lavrenyuk
61 National Oceanic and Atmospheric Administration, 5200 Auth Rd, Camp Springs MD 20746, USA
2 Space Research Institute NASU-NS AU, Kyiv -03187, Glushkov Ave, 40, 4/1
3 Ukrainian Hydrometeorological Center, Kyiv -01034, Zolotovoritskaya st., 6-B
4 Ukrainian Hydrometeorological Institute, 03650, Kyiv, Nayki Prospekt, 37
5 National University of Life and Environmental Sciences of Ukraine, Kyiv -03187, Heroyiv Oborony st., 15
6 National Technical University of Ukraine “KPI”, Kyiv -03056, Peremogu ave, 37
This article aims at estimation of the relative efficiency of the use of satellite data to predict winter wheat yield in Ukraine at the level of administrative regions (oblast), as well as to compare the effectiveness of prediction based on empirical and biophysical models. As an empirical model we use a linear regression model that uses as predictor 16-day NDVI composite derived from MODIS data at 250 m resolution. Non-linear regression model is considered as well as a predictor that uses meteorological parameters. These two empirical approaches are compared with the biophysical approach, implemented in CGMS (Crop Growth Monitoring System) system, based on WOFOST crop growth model and a dapted for Ukraine. It is shown that the highest accuracy of the forecast yield of winter wheat provides an approach based on biophysical models of growth. The disadvantage of this approach is the difficulty of setting up a model (a large amount of input data) and the lack of lead time. Regression models (linear model based on satellite data and nonlinear, based on weather) show roughly the same accuracy, which is given sufficient lead time allows the use of these predictions in practice.
Keywords: прогноз урожайности, регрессионная модель, биофизическая модель, маска культуры, информационная технология, MODIS, NDVI, кластеризация, CGMS, WOFOST, forecast yield, the regression model, biophysical model, the mask of culture, information technology, MODIS, NDVI, clustering, CGMS, WOFOST
Full text