Wheat yield variability will increase in the future due to the projected increase in extreme weather events and long-term climate change effects. Currently, regional agricultural statistics are used to monitor wheat yield. Remotely sensed vegetation indices have a higher spatio-temporal resolution and could give more insight into crop yield. In this paper, we (i) evaluate the possibility to use Normalized Difference Vegetation Index (NDVI) time series to estimate wheat yield in Latvia and (ii) determine which weather variables impact wheat yield changes using both ALARO-0 and REMO Regional Climate Models (RCM) output. The integral from NDVI series (aNDVI) for winter and spring wheat fields is used as a predictor to model regional wheat yield from 2014 to 2018. A correlation analysis between weather variables, wheat yield and aNDVI was used to elucidate which weather variables impact wheat yield changes in Latvia. Our results indicate that high temperatures in June for spring wheat and in July for winter wheat had a negative correlation with yield. A linear regression yield model explained 71% of the variability with a residual standard error of 0.55 Mg/ha. When RCM data were added as predictor variables to the wheat yield empirical model a random forest approach resulted in better results compared to a linear regression approach, the explained variance increased up to 97% and the residual standard error decreased to 0.17 Mg/ha. We conclude that NDVI time series and RCM output enabled regional crop yield and weather impact monitoring at higher spatio-temporal resolutions than regional statistics.
Vannoppen, A, Gobin, A, Kotova, L, Top, S , Cruz, LD , Viksna, A, Aniskevich, S, Bobylev, L, Buntemeyer, L, Caluwaerts, S, De Troch, R, Gnatiuk, N, Hamdi, R, Remedio, AR, Sakalli, A, De Vyver, HV, Van Schaeybroeck, B & Termonia, P 2020, ' Wheat yield estimation from NDVI and regional climate models in Latvia ', Remote Sensing , vol. 12, no. 14, 2206.
Vannoppen, A., Gobin, A., Kotova, L., Top, S. , Cruz, L. D. , Viksna, A., Aniskevich, S., Bobylev, L., Buntemeyer, L., Caluwaerts, S., De Troch, R., Gnatiuk, N., Hamdi, R., Remedio, A. R., Sakalli, A., De Vyver, H. V., Van Schaeybroeck, B., & Termonia, P. (2020). Wheat yield estimation from NDVI and regional climate models in Latvia . Remote Sensing , 12 (14), [2206].
@article{ac4382c5e684405f8f1efe80c511133f,
title = " Wheat yield estimation from NDVI and regional climate models in Latvia " ,
abstract = " Wheat yield variability will increase in the future due to the projected increase in extreme weather events and long-term climate change effects. Currently, regional agricultural statistics are used to monitor wheat yield. Remotely sensed vegetation indices have a higher spatio-temporal resolution and could give more insight into crop yield. In this paper, we (i) evaluate the possibility to use Normalized Difference Vegetation Index (NDVI) time series to estimate wheat yield in Latvia and (ii) determine which weather variables impact wheat yield changes using both ALARO-0 and REMO Regional Climate Models (RCM) output. The integral from NDVI series (aNDVI) for winter and spring wheat fields is used as a predictor to model regional wheat yield from 2014 to 2018. A correlation analysis between weather variables, wheat yield and aNDVI was used to elucidate which weather variables impact wheat yield changes in Latvia. Our results indicate that high temperatures in June for spring wheat and in July for winter wheat had a negative correlation with yield. A linear regression yield model explained 71% of the variability with a residual standard error of 0.55 Mg/ha. When RCM data were added as predictor variables to the wheat yield empirical model a random forest approach resulted in better results compared to a linear regression approach, the explained variance increased up to 97% and the residual standard error decreased to 0.17 Mg/ha. We conclude that NDVI time series and RCM output enabled regional crop yield and weather impact monitoring at higher spatio-temporal resolutions than regional statistics. " ,
keywords = " ALARO-0, Latvia, NDVI, PROBA-V, Regional climate model, REMO, Spring wheat, Weather impact, Winter wheat, Yield estimation " ,
author = " Astrid Vannoppen and Anne Gobin and Lola Kotova and Sara Top and Cruz, {Lesley De} and Andris Viksna and Svetlana Aniskevich and Leonid Bobylev and Lars Buntemeyer and Steven Caluwaerts and {De Troch}, Rozemien and Natalia Gnatiuk and Rafiq Hamdi and Remedio, {Armelle Reca} and Abdulla Sakalli and {De Vyver}, {Hans Van} and {Van Schaeybroeck}, Bert and Piet Termonia " ,
year = " 2020 " ,
month = jul,
day = " 1 " ,
doi = " 10.3390/rs12142206 " ,
language = " English " ,
volume = " 12 " ,
journal = " Remote Sensing " ,
issn = " 2072-4292 " ,
publisher = " MDPI AG " ,
number = " 14 " ,
}