Estimating maize and cotton yield in southeastern Turkey with integrated use of satellite images, meteorological data and digital photographs

Alganci U., OZDOGAN M., Sertel E., ORMECI C.

FIELD CROPS RESEARCH, vol.157, pp.8-19, 2014 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 157
  • Publication Date: 2014
  • Doi Number: 10.1016/j.fcr.2013.12.006
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.8-19
  • Keywords: Yield estimation, Crop cover fraction, Light use efficiency, Digital photographs, Spectral vegetation index, Satellite images, RADIATION-USE EFFICIENCY, REMOTE ESTIMATION, VEGETATION INDEXES, CORN YIELD, CROP, GROWTH, TEMPERATURE, MODEL, LEAF, CLASSIFICATION
  • Istanbul Technical University Affiliated: Yes


This study focuses on yield estimates of planted areas of cotton and maize in southeastern Turkey. It integrates multi-temporal satellite images, daily digital photographs of cultivated parcels, and daily meteorological data. Our research produced vegetation cover fraction (VF) estimates from digital photos and defined relationships between this information and the spectral vegetation index (VI) obtained from satellite images. Meteorological parameters limiting the light use efficiency of crops (LUE), such as temperature and vapor pressure deficit, were also calculated and incorporated into the yield estimation process. Results showed that the use of digital photo-based VF rather than the fraction of photosynthetically active radiation (fAPAR) in the LUE model provided the most accurate yield estimates. It produced less than 5 percent relative error in cotton and maize test parcels. In general, the VF-SVI relationship showed high linear correlation, with a range of 0.825-0.980 R-2 in all test parcels. Crop specific regression equations derived from these relationships enabled yield estimates at the parcel level across the study area. When compared to statistical yield information at four districts, the remote sensing-based method proved to be reliable, with relative errors below 10 percent in most cases. Moreover, greenness index (Cl) was also used in gross primary production (GPP) approximation, and yield estimates using this method also provided reasonable accuracy. Results also provided valuable information about the effects of region-specific meteorological conditions and crop management activities on yields. Finally, the higher yield estimation errors that result from the use of generic SVI-fAPAR equations in the literature indicate the need for local calibration of this relationship. (C) 2013 Elsevier B.V. All rights reserved.