Remote Sensing Monitoring Habitat Factors Of Rice Planthopper And Its Application

Rice planthopper is the main destructive and long-distance migratory insect pest of rice in Asia and is becoming a threat to the grain yield and food security of China. Rice planthopper had outbroken several times in middle and lower reaches of Yangtze River, especially in years of1991,1997,2005-2007. The long-range forecasting of rice planthopper outbreak is a challenge for the entomologists since its sudden occur-rence. With the development and improvement of sensors, remote sensing technique is becoming an important means for acquiring large-scale insect pest information. The habitat factors of planthopper inverted by using remote sensed data could be applied to investigate the population dynamics and occurrence rules of planthopper so as to make reliable predictions.In this study, we chose Yangtze River Delta region as the study region and used multi-sensors remote sensed data, such as MODIS, Landsat, GDEM, TRMM etc. to obtain the main habitat foctors of rice planthopper, including tempo-spatial distribu-tion of paddy rice, vegetation index, air temperature and precipitation. According to the relationship of the remote sensed habitat foctors and insect survey data, remote sensing models were built for planthoppers occurrence forecasting. The main results obtained were listed as follows:(1) Mornitoring tempo-spatial distribution of rice planting area using MODIS imagesTo devise efficient rice planthopper monitoring and management plans, it is ne-cessary to get the accurate and updated information about the spatial distribution of paddy rice.The Normalized Weighted Difference Water Index (NWDWI) was pro-posed for identification of flooding and transplanting stage of paddy rice and good at enhanceing the characteristic of water shed. Annul spatial distribution of single, early and late rice planting regions in the period of2000-2012was mapped using MODIS data according to the analysis of time series MODIS-EVI of different land cover type. The agricultural census data and Landsat imagery were applied for accuracy valida-tion and spatial matching test. The result indicate that the accuracy of total rice plant- ing area derived from MODIS in years2000-2010were greater than85%except for the year of2007and2010. The annual decision coefficient (R2) for early rice between MODIS-derived area and census data ranged from0.388to0.678, while for the sum area of single and late rice, annual R2ranged from0.545to0.742during the years2000-2010. In the pixel level, the user’s and producer’s accuracy were73.70%and77.33%, while in the3X3moving window level, the user’s and producer’s accuracy were96.77%and99.96%, respectively. The fragmentation of rice field, terrain com-plexity and cloud cover were the main causes for the underestimatin of rice areas.(2) All-weather air temperature estimation in Yangtze River Delta regionThe optimal models for estimation of daily maximum, average, minimum air temperature were generated based on the application of MOD-IS/Terra and MODIS/Aqua LST, EVI, cos(SZA), latitude, longitude and elevation. The model developed by merging the day and night pass LST data of both MOD-IS/Terra and MODIS/Aqua has the highest accuracy with the lowest RMSE of Tavg (1.424℃), followed by Tmax(1.605℃) and the highest RMSE of Tmin (1.992℃). RMSE of these models were all less than3℃, which can meet the need for applying in ecological models. The ten-day composite air temperature was generated using the average value of the good days, while the missing data caused by cloud and cloud shadow were interpolated by inverse distance squared (IDS) method. Compared with actural meterological data, mean ten-day Tavg has a good correlation with actural me-terological data with RMSE<2℃, MAE<1.4℃, mean ten-day Tmax with RMSE<2.5℃, MAE<0.18℃, and mean ten-day Tmin with RMSE<2.54℃, MAE<2.0℃. RMSE of the ten-day composite air temperature were lower than2℃in the farmland.(3) Accuracy validation of TRMM satellite precipitation dataThe TRMM3B423-hour precipitation data during1998-2010were integrated to daily and monthly data and validated using rain gauge data. It indicated that at the rain gauge scale, the accuracy of TRMM daily precipitation during the rainy season is both higher than that in the whole year and dry season. At the climate zone scale, the accu-racy of daily TRMM precipitation data is higher and the MAE was less than3mm/day. The monthly TRMM precipitation and rain gauge data has a significant pearson coef- ficient higher than0.8. At the rain gauge scale, a fair capability was shown when us-ing daily TRMM data to detect rain events of different magnitudes, but the capability for identification rain event was as high as60%in the rainy season. While at the cli-mate zone scale, the forcasting accuracy of rainy event in the whole year and rainy season was higher than85%.(4) Monitoring rice planthopper habitat and early warningFirstly, the suitable and optimum EVI for the occurrence of rice planthoppers was about0.40-0.60and0.45-0.55, respectively. EVI=0.4is the key period and should be paied more attention to field survy and pest controling. Secondly, the re-mote sensed models for planthopper ten-day forcasting from July to September were developed using multivariate statistical methods based on the habitat factors derived from remote sensed data (e.g. EVI, mean temperature, maximum temperature, min-mum temperature, precipitation days and cumulative rainfall) during years of2000-2009, and are significant at the level of0.001. In2010, the estimated occurance degree of planthopper in July and August were consistent well with the actural level, but the accurancy for absolutedly and generally consistent in September were36.36%and54.55%, respectively and the difference between the predicted and actural degree was not higher than2. The models can forecast ten days in advance. Finally, the dy-namics distribution of planthopper hazard from late August to late September in2005-2007were mapped and analyzed based on the analysis of the response of NDVI corresponding to insect amount of planhopper.