Some Issues Of Remote Sensing Based Crop Phenology And Condition Monitoring

Crop phenology and condition monitoring is one of the most essential research themes in agricultural remote sensing. To fulfill its objective, which is to help farmer to make decisions on field management and early crop yield estimation, phenological and conditional information are the most valuable reference for crop-related decisions. Timely and accurately crop information is proven to be beneficial to water and nutrition management, plant diseases and insect pest detection, crop yield, food price and food security for both developmented and developing countries. Modern precision agriculture needs combination of geographical information system (GIS), gobal position system (GPS), remote sensing (RS) and agricultural regionalization. This is not only the requirement of modern precision agriculture, but also the needs of changeover from’large agricultural country’to a’powerful agricultural country’.There are several liminations in existing related research works bcause of the absence of satellite remote sensors apporiate vidalition materials:firlstly, most research focused on regional scale applications, which often employed 1000m spatial resolution remote sensing datum and degraded the purity of single pixel. Secondly, normalized differential vegetation index (NDVI) time series composited by Maximum Value Composite (MVC) is still the main data resource for current studies. However, phenology is a time-sensative concept and the MVC process has not been quantitatively analysised in crop phenological researches. Thirdly, remote sensing derived metrics and field obversitions have not been coupled together for specific crop types, which also hinderes the development and application of remote sensing in phenology studies. Lastly, crop condition research is dominated by NDVI that is sensitive to canopy soil background and saturate at high vegetation density. New vegetation index should be developed and validated in crop condition assessment.Considering the requirement of modern agriculture, several works have been conducted in this manuscript:Firstly, this dissertation investigated multi-temporal data reconstruction techniques and made sone improvements to double logistic function fitting method. Crop phenology and condition evaluation researches are mainly based on time series remote sensing vegetation index and these data are usually containing considerable noise which should be eliminated. The chosen of denoising techniques is finally determinng the caululation of phenological metrics. Since there exists plenty of reconstruction methods, evaluting and analysising these techniques for phenology related researches should be conducted. The manuscript made a comparison over these existing methods and found that double logistic function has the best performance. Two improvements have done to double logistic function fitting technique to make it work better.Secondly, the thesis evaluated the MVC process by using the composited date in MODIS Vegetation Index products vesion 5, from which the manuscript describe the distribution pattern of temporal error in a composite period. Also the acticle presented a noval technique for deriving phenological metrics from smoothed time series curve. A root mean square error matrix based method that coupling remote sensing metrics with ground obversations were described in the manuscript.Thirdly, the dissertation illustrated a quantitative technique to assess the effect of three different temporal resolution datums, which are daily,8 day composite and 16 day composite MODIS products, on remote sensing based crop phenology detection studies. The evalution indicator included two parts:matching error and estimation error. While the matching error described the relationship between remote sensing metrics and NASS statistical information, the estimation error showed the difference of different datasets on phenological calculation.At last, the manuscript presented a new vegetation index (GRNDVI) for crop condition monitoring. The comparison experiment showed that the GRNDVI has inherted the merits of both SR and NDVI. And it could be considered as an alternative of NDVI in crop condition monitoring researches.