Comparison of geostatistical- and remote sensing data-based methods in wheat yield predication in some of growing stages (A case study: Nemooneh filed, Golestan province)

Food security has been most important concern of the mankind on the earth. On the other hand, agricultural productions have always been face by risk probability in the case of weather and changes in international markets, however this risk probability never undeleted completely, but could be minimized that by pre-harvesting yield estimation. In this study, different methods of vegetation maps provision were involved toprovide a suitable pre-harvesting map for wheat yield.

For comparison of remote sensing and geostatistics-based methods capabilities in wheat yield predication in wheat fields, a survey was conducted in 2011-12 growing season. 101 plant samples were taken from 2500 hectare wheat fields in tillering, booting, seed filling and maturity stages (three times for leaf area index and dry weight and one sampling for yield) and related measurements were done. Ordinary, Universal and Disjunctive Kriging methods were applied and semivariograms were provided, then proper models were fitted. Different statistical indices were used to test the accuracy. Also, three +ETM images acquired by Landsat satellite were used which were matched by sampling dates. Four images for previous years also were used as needed. Eight plant indices were provided from aforementioned images and were compared with plant variables which were recorded or measured simultaneously, then related relations were determined and maps were provided. By fitting the logistic model between yield and plant variables, yield prediction maps were evaluated by remote sensing and, the obtained maps were compared using different statistical indices.

Evaluation results of interpolation methods revealed that spherical, exponential, Gaussian and circular models were superior models in this study. Also, results on the survey indices derived from satellite images showed a significant relationship between the variables and indices derived from satellite images in the end of tillering stage. Assessment of generated yield maps, demonstrated pronounced superiority of remote sensing techniques compared with geostatistical-based analysis methods. The results demonstrated the capability of satellite images in regional scale to predict wheat yield (with 715 kg.ha-1 biass in tillering stage).

According to the acceptable accuracy of the remote sensing compared with the Geostatistics- based method along with ease of and low cost of this method, use of the remote sensing and satellite images – derived vegetation indices could be a new horizon in regional yield estimation. Since satellite images provide an actual representation from the crop status, could involve significantly to the growth modeling.