Monitoring Grain Quality Of Weak-Gluten And Medium-Gluten Wheat By Remote Sensing Technology In The Middle And Lower Reaches Of Yangtze River

This study was done to develoop predictive models for monitoring grain quality of weak-gluten and medium-gluten wheat in the region along Yangtze River, coastal region, and Lixiahe region of Jiangsu Province, using remote sensing data, ground spectral data, and growth parameters of wheat plants at the growth stages of jointing, flowering, and grain filling. Remote sensing data were collected from an environmental disaster mitigation satellite HJ1A/1B using remote sensing technology. Ground spectral data were collected using ground spectra technology. The growth parameters of wheat plants, including biomass and nitrogen content in each part of wheat plants, grain quality indices, and grain yield, were also collected through field sampling. Predictive models were developed by coupling these data and meteorological data. The feasibility of predicting grain quality of weak-gluten and medium-gluten wheat by remote sensing technology was also discussed. The main results were as follows:1. The correlations between growth parameters at different growth stages, grain yield and quality, and remote sensing variables were analyzed. The growth parameters of wheat plants at jointing stage were not significantly correlated to grain quality and yield. The SPAD readings of green leaves, nitrogen content in green leaves, nitrogen content in leaf sheaths at flowering could be used to predict grain protein content, starch content, and sedimentation value. At grain filling period, SPAD readings of green leaves, nitrogen content in spikes and in plant at flowering could be used to predict grain protein content, wet gluten content, starch content, and sedimentation value. Some growth parameters were correlated significantly with remote sensing variables. Suitable remote sensing variables could be used to monitor plant growth and provide timely and reasonable information for decision-making. At jointing stage, SIPI, NRI, PSRI, B1and B3could be used to monitor SPAD, biomass, nitrogen content in green leaves, nitrogen content in leaf sheaths, and plant nitrogen content. At anthesis, PSRI, B3, DVI could be used to monitor SPAD, biomass, nitrogen content in green leaves respectively. At grain filling stage, PSRI could be used to monitor SPAD, nitrogen content green leaves and nitrogen content in leaf sheaths. It was possible to make use of remote sensing variables to monitor wheat growing conditions at different periods. Based on these parameters, grain quality and yield could be monitored and forecasted indirectly.2. Remote sensing variables at flowering and grain filling period were found correlated to grain quality and yield. DVI, NDVI, and NRI at anthesis, NDVI, NRI, and GNDVI at grain filling stage could be used to predict grain protein content. DVI, NDVI, B4at anthesis and NDVI, GNDVI, and DVI at grain filling stage could be used to predict starch content. Most remote sensing variables were also correlated to grain wet gluten content at the0.05or0.01probability level. DVI at anthesis and DVI at grain filling period had the highest correlation coefficient with wet gluten content in grain. B4and DVI at grain filling stage were negatively correlated to sedimentation value. NDVI at two stages was positively correlated to grain yield at the0.05probability level. A number of remote sensing variables were also correlated to grain yield.3. By analyzing the correlations between remote sensing variables and grain yield and quality data, three predictive models were developed. The predictive efficiency of these three models was compared. Remote sensing variables at different stages could be used to forecast grain quality and yield. Considering the prediction efficiency, we suggested prediction should be initiated at anthesis. The prediction efficiency of single-factor model based on single remote sensing variable at grain filling period was better than that at anthesis. The prediction efficiency of bi-variables model was better than that of the single-factor model. The prediction efficiency of multi-variable model based on remote sensing variables was better than that of single-factor model and the bi-variables model. Multi-variable model significantly optimized some statistical parameters, including Fitted R2, RMSE, and relative error RE.4. In order to optimize statistical parameters the predictive models, including RMSE and RE, meteorological factors were incorporated into the models. The temperature factor N△t was incorporated to optimize the multi-factor model for the prediction of grain protein content. The temperature factor S△t and the illumination factor S△s were incorporated the nto multi-factor model for starch content prediction. As a result, statistical parameters of the predictive models, including RMSE and RE, were reduced significantly. Similarly, the incorporation of Y△t significantly increased prediction efficiency of yield prediction model. As a result, the analog coefficient was76%,72%, and78%respectively.5. The zoning map for predicting the content of protein and starch at grain filling stage, developed from the graphic superposition of HJ-1A/1B images, region distribution map of wheat production, and the administrative map of Jiangsu Province, could provide wheat production department and grain purchasers with useful information on grain quality...