Bayesian estimation of crop yield function: drought based wheat prediction model for tigem farms

Yildirak K., Kalaylioglu Z., Mermer A.

ENVIRONMENTAL AND ECOLOGICAL STATISTICS, vol.22, no.4, pp.693-704, 2015 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 22 Issue: 4
  • Publication Date: 2015
  • Doi Number: 10.1007/s10651-015-0327-6
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.693-704
  • Keywords: Drought, Markov Chain Monte Carlo, Normalized difference vegetation index, Prediction model, Spatiotemporal model, Water
  • Hacettepe University Affiliated: Yes


Drought is one of the biggest threats that affects agriculture. Based on recent climatic observations and future projections, drought continues to increase its harmful impact on agricultural productivity especially in the arid areas of Turkey. Wheat farming in these arid and semi-arid areas such as Central Anatolia depends heavily on precipitation, thus monitoring for drought is needed. The timing of precipitation is also as important as its quantity. This study makes use of a crop and location specific model developed by FAO to simulate water related variables such as evapotranspiration, water deficiency and water satisfaction index to estimate the crop yield function for rain-fed wheat production in the arid regions of Turkey. A spatio-temporal yield model is estimated by Bayesian method known as Markov Chain Monte Carlo. By standardizing the simulated variables over normalized difference vegetation index (NDVI), impact of drought related variables on wheat yield is studied and two variables are found. Use of NDVI, as a numeraire, comes in handy in many ways. For actual evapotranspiration estimate, it strengthens separation between evaporation and transpiration and, for water deficiency, it better represents the drought properties of farms for the terrain chosen.