Study On Winter Wheat Yield Estimation Model Based On Satellite Remote Sensing

The Huai River basin is located at the transition zone of the south and north climates,where winter wheat suffers from various meteorological disasters(such as cold and frost damage,drought,cold spell in late spring,and dry and hot wind)frequently throughout its whole production period,resulting in yield decline in a large area or even total crop failure.The fluctuation in grain output would inevitably destabilize national economy,and therefore,it is necessary to acquire accurate reliable information of agricultural crops including winter wheat,such as seeded area and yield,while remote sensing is a reliable means with low cost for the acquisition of such information in a certain region.There are difficulties in application of conventional remote sensing techniques in the Huai River basin due to the fact that the weather in the growth periods of agricultural crops in this area is mainly cloudy,rainy or snowy.However,radar works in microwave band(1 mm-1 m),and radar wave is less affected by cloud and could passes through cloud and fog due to its longer wavelength,thereby acquiring information of crops from crop canopy to stem at different levels.On the basis of these characteristics,synthetic aperture radar(SAR)becomes an important technical means for the classification,growth monitoring and yield estimation of agricultural crops.There have been studies on identification,growth monitoring and yield estimation of wheat by SAR at aboard.SAR is also applied to identification and yield estimation of agricultural crops in China.However,no studies on application of SAR in yield estimation of winter wheat have been reported in China.Especially,China is a great agricultural country where food production is a priority among priorities in agricultural production,while the application of microwave remote sensing in quantitatively recognition and yield estimation of agricultural crops could fill the blank in the field of serving agriculture with SAR satellite.The physical characteristics of winter wheat at different growth stages reflected on radar images are all different,and correspondingly,there are remarkable differences in radar backscattering coefficient.On the basis of selectively using radar images corresponding to different growth stages,a yield estimation model could be obtained by analyzing radar images corresponding to growth stages at which wheat has stronger sensitivity.In this paper,in order to acquire a broadly-adapted and localized yield estimation model of winter wheat,different satellite-borne polarimetric SAR images of different time phases(different growing period of winter wheat)were investigated,combined with artificially-measured yields,yield estimation models were established according to the relationship between yields of winter wheat and backscattering coefficients at various time phases,and an optimal yield estimation model was selected,realizing large-area yield estimation of winter wheat.A pilot study was conducted to establish a winter wheat yield estimation model based on spaceborneSAR in Guoyang County,Anhui Province.Sampling for yield detection was performed in experimental field about one week before harvest of winter wheat,combining with synchronous SAR images,radar backscattering coefficients were extracted,and a unitary yield estimation model was established,with a yield estimation accuracy exceeding 80%.On the basis of the pilot study,spaceborneSAR images(RADARSAT-2)in April and May in 2013 and 2014 in the Huai River basin and yield information of winter wheat in experimental field before harvest(at the end of May)in this area were selected.The satellite data and experimental field information were pre-processed,and yield estimation models were established using radar backscattering coefficients obtained from different polarizations(HH,HV)of RADARS AT-2 and the artificially-estimated yields.An optimal yield estimation model was selected and used for the yield estimation of large areas of winter wheat,and both the planting area and yield estimation accuracy of winter wheat calculated in 2013 was close to 90%.However,due to lodging of winter wheat in a large area in 2014 and inadequate completeness of manual samples and experience of sampling proportion,the yield estimation accuracy decreased compared with 2013.In this study,1 image of each of 2 important growth stages of winter wheat,i.e.,regreening tillering stage and booting stage was selected,as the images of the 2 stages could reflect the growth trend of winter wheat in this area.Furthermore,milk-ripe stage is an important growth stage for winter wheat when the filling of winter wheat has been completed basically and wheat is about to ripen,weight of fresh ear at this time could directly reflect yield,and due to great effect of ear change on geometric structure of the crop at this stage,radar sensing shows high sensitivity to ear and could thus be used for direct estimation of yield of winter wheat.The yield estimation model established with polarization ratio VV/VH realized a yield estimation accuracy close to 80%.Moreover,in this study,optimal remote sensing satellite,i.e.,small satellite for environment and disaster monitoring and forecasting(HJ-1),was also used for yield estimation,so as to perform mutual verification with the yield estimation result by radar satellite.The yield estimation model established on the basis of NDVI(normalized differential vegetation index)extracted from HJ-1 and yield of winter wheat in experimental field at milk-ripe stage achieved the yield estimation accuracy close to 70%,and was compared with the radar yield estimation model obtained in 2014.Spaceborne SAR is playing a more and more important role as a new research direction in yield estimation and growth monitoring of winter wheat as well as yield estimation of other agricultural crops.