Sensitivity Of Hyperspectral Reflectance To Monitor Rice Pests And The Monitor Methods For Rice Planthoppers

Crop pests and diseases are one of the key factors caused the final yield loss of crops. Early warning to pests and diseases is a powerful method to prevent the spread of insects and diseases and to protect the crops yields. Hyperspectral remote sensing technology is real-time, rapid, non-destructive, and accurate to monitor the crops and some other objects. Therefore, the hyperspectral method will be used widely in crop pests and disease monitoring.In this paper, Nilaparvata lugens (Stal)(BPH) and Sogatella furcifera (Guenee)(WBPH) were used to study the sensitivity of hyperspectral reflectance to monitor rice damage, and the regression models to minotor the BPH and WBPH number which damaged rice plants were established. The following were the main results.The results showed that the spectral reflectance at the range of the green and near-infrared regions significatnly decreased with the increase of the amount of rice planthoppers.535-567nm and710-1000nm were the sensitive wave bands for detecting the number of planthoppers.The sensitivity of reflectance to monitor the number of rice planthoppers was depended on the age of rice seedlings. The reflectance from the15and25days rice seedlings could detect the damage of four second instar WBPHs damaged for four days. The reflectance from rice plants at tillering stage could detect the damage of four second instar BPHs for eight days, or eight second instar BPHs for six days. The spectrum from the combined leaves damaged by rice leaf folder (RLF) could detect10%leaf-roll rate. The reflectance from rice leaves was very sensitive to the damage of rice planthoppers and RLF.The compensation effect of rice at tillering stage was found from the reflectance when the rice was damaged by less than eight second instar BPHs for two days, and the spectral indices were higher from the damaged rice than that from the healthy one. The spectral reflectance and indices also showed that the rice at tillering stage damaged for8days by four second instar BPHs or6-8days by six second instar BPHs could recover from damage and became similar as the control rice. However, they did not recover if they were damaged by six or eight BPHs for eight days.The spectral reflectance from leaves appeared a lag effect for exhibiting the rice planthoppers damage. The reflectance and spectral indices were no significant different among the rice damaged by0-8second instar BPHs damaged for four days, but after4-6days removed the BPHs, the RVI and NDVI from the rice damaged by8BPHs were significant less than the control rice.The reflectance from the fourth leaf from top (4LFT) and all leaves of rice damaged by WBPH at the tillering and booting stages was measured. The results showed that the sensitive wavebands from4LFT to detect the WBPH number were similar as the wavebands from all leaves of rice plant. The reflectance from the4LFT could be used to monitor the WBPH damage. The vegetation indices (DVI, RVI, and NDVI) from all wavebands400-1000nm were calculated, and their correlations with the number of WBPHs were analysized. The results showed that the vegetation indices had significant relation with the WBPH number after damaged three or four weeks, and the regression models to monitor the number of WBPHs based on the DVI, RVI, and NDVI were precise, and the errors for tillering and booting stage rice were15-20%and17-23%, respectively. At tillering stage, the model Y=-1381+2134.2DVI500-680+1422RVI862-873-1358.5DVI450-740+1293.8DVI670-740(R2=0.8173, RMSE=3.198) based on the4LFT of rice infested by WBPH for2weeks was the best; at booting stage, the model Y=2913.4-2487.9RVI956-989+545.66NDVI404-973+111.52DVI670-760(R2=0.8313, RMSE=4.522) based on all leaves of rice infested by WBPH for4weeks was the best.The sensitive wavebands to detect the BPH damage from the4LTF at400-700nm was the similar as from the all leaves of rice at tillering stage, and the sensitive wavebands at734-1000nm was similar as from all leaves of rice at booting stage. Therefore, the reflectance from4LFT also could be used to monitor the BPH damage in rice. The vegetation indices DVI, RVI, NDVI in some sensitive wavebands were significantly related to the number of BPHs. The regression models for detecting the number of BPHs were established at tillering and booting stages of rice, and their predicted errors were10-25%and20-30%, respectively. At tillering stage, the model Y=18.931+220.45DVI550-680+186.97DVI450-760-370.13DVI680-760(R2=0.8979, RMSE=2.344) based on the4LFT of rice infested by BPH for3weeks was the best; at booting stage, the model Y=55.645-112.32NDVI550-680-90.864NDVI450-760+101.15NDVI680-760(R2=0.7370, RMSE=5.645) based on all leaves of rice infested by BPH for4weeks was the best.At the same time, damage degree of rice by BPH was difined as the weeks multiplying the number of BPHs, and degree regression model was established based on the vegetation indices to detect the damage degrees at tillering and booting stages. The predicted errors were15%and17%for tilling and booting stage of rice, respectively. The regression model Y=78.517-73.298NDVI680-760-1.5338RVI450-760(R2=0.7941, RMSE=10.02) at all leaves of rice at tillering stage was the best to detect the damage degree.