Study On Maize-yield Estimation Of Ganzhou District By Remote Sensing In The Middle Reaches Of Heihe River Basin

Food supply and demand situation is directly related to the development of nationaleconomies and people’s living. Timely and accurate understanding the food productionsituation in the country or region, the advance estimates of crop production and dynamicmonitoring for the crop have great significance. Yield estimation method by Remote sensingis a new method accompanied by high-tech’s developing, which replacing gradually theconventional yield estimation methods because of the short detection period, covering a widearea, the strong Current potential, data-rich, easy way to get the data, and low cost advantages.In order to use the corn canopy reflectance spectrum time series information Monitoring byRemote sensing for corn production estimates, this paper Considered normalized differencevegetation index (NDVI) as a key parameter of Regional maize production estimate model,The wavelet transform fusion between the moderate resolution imaging spectroradiometer(MODIS) data and high spatial resolution Landsat8TM data was adopted for obtaining NDVITime-series information with a spatial resolution of30m. And the standard growing curves ofmain fall crops were then constructed with the NDVI time series. take full advantage ofMODIS data’s long time series and TM data’s high spatial resolution, by indicating cropsdifference in phenology, combining timing analysis and hierarchical classification werecarried out with the NDVI time series for effectively distinguish the spatial distribution ofmaze sown and other crops in Ganzhou district. the use of ground statistics data and fieldsurvey sample data were conducted for classification accuracy assessment. According to thecorrelation analysis of spring maize NDVI values and yield in study area, selected the bestproduction estimate period, and the regression analysis was used to build the spring maizefitting model between NDVI and yield. The2011a spring corn acreage and yield data in studyarea were carried out to the optimal proposed combined model’s accuracy verification anderror analysis. The results showed that:(1) Use the wavelet transform fusion method to the MODIS data with250m spatialresolution and the TM data with30m spatial resolution of long sequence spring corn growingperiod of the study area in2013, the tests prove that application of Symlets when the waveletdecomposition of images was3, and the fusion rule select high frequency and low frequencycompletely replace can obtained the best timing data with30m spatial resolution of the studyarea. (2)Timing analysis and hierarchical classification were used to mapping maze sown areaof fusion imaging at different growth stages of spring maize in study area. Cartographyaccuracy was97.6%, user accuracy was89.71%, the overall classification accuracy achieved91.28%, Kappa coefficient was0.8038, much higher than the single MODIS dataclassification and conventional supervised classification to prove the image fusion thatresolution difference multiples huge can obtained more accurate crop acreage. Comparativethe spring maize area between Real statistics and Monitored by remote sensing in study area,the classification accuracy reached95.3%, obtained a better classification results.(3) According to the correlation analysis of spring maize NDVI values and yield in studyarea, the jointing stage, heading stage and flowering stage with higher correlation coefficientswhich was0.7204,0.8133and0.762were selected to be the best spring corn yield estimationperiod, using regression analysis to constructed fitting model of spring corn sample pointsNDVI values and pixel yield with a single Growth period or combination of Growth phase.the combined model which correlation coefficient R reached0.8158, goodness of fit reached0.6655, RMSE reached300.2982(the jointing stage+heading stage+flowering period) wasselected to be the optimal spring corn yield estimation model of study area, The2011a springcorn acreage and yield data in study area were carried out to the optimal proposed combinedmodel’s accuracy verification and error analysis. the average relative error range between-0.83%and4.66%, the model’s relative error was only1.7%. The results show that theoptimal yield estimation model of study area had the ability to accomplish spring corn yieldestimation in study area.