| In recent years, with the climate and environmental changes increasing, drought, floods, hail, wind and so a variety of agricultural pests and diseases are becoming increasingly frequent disasters, huge losses were caused every year due to agricultural disasters, especially in our production, food security, natural ecological environment and economic development.Hail damage has become one of the frequent occurrence of natural disasters, especially in Heilongjiang Province. Heilongjiang Province is the country’s major agricultural province, is also an important grain production bases, but the area is also the frequent occurrence of natural disasters.Hail of the province is one of the major natural disasters in recent years, the incidence of Heilongjiang Province, and hail damage occurs worsening trend, the provincial agricultural production had a significant impact.Therefore, the study of scientific and reasonable and effective hail monitoring methods,analysis temporal characteristics of Heilongjiang Province, evolution and trends, not only in favor of disaster warning, disaster mitigation and disaster damage assessment work smoothly, but also plays an important practical significance in guiding agricultural production and sustainable development.Traditional monitoring methods are mainly field investigation, measurement and samplingbased, while in data acquisition, data representation, there are many deficiencies and cost a lot, and time-consuming prediction lag,not achieve real-time or near real-time disaster monitoring, and more difficult to get a wide range of disaster monitoring.With its remote sensing technology to monitor a wide range of fast, low-cost, readily available, and easy to carry out a long-term dynamic monitoring features.It has gradually replaced the traditional field survey method as an effective means to obtain information on spatial and temporal distribution of hail damage. MODIS(Moderate-resolution Imaging Spectroradiometer)data with high spectral resolution and high temporal resolution and moderate spatial resolution, wide coverage, and highlight has advantages that can be obtained for free,has become the best of a wide range of fast dynamic agricultural disaster monitoring data source.In this study, continuous multi-temporal product data, which is about the spatial resolution of 250 m and synthetic MODIS vegetation index of 16 d, has been used for hail monitoring. First, I discussed the theoretical basis and principles of remote sensing hail; then, I extracted feature extraction of arable land range based on a different type of time series. Then, I proposed model that normalized difference vegetation index values were normalized(RNDVI_M), determined the threshold of disaster based on the spatial variation of RNDVI_M value, extracted disaster range, and verified extraction results based on the same period or quasi-period of HJ-1A / 1B CCD imaging and ground data. finally, I extract disaster caused by hail and analyze Characteristics and Genesis of the temporal pattern of hail on the basis of differences in landscape indices. The main contents and conclusions are as follows:(1)Extract ranges arable land. Based on product data of Heilongjiang Province in 2012, which is about the spatial resolution of 250 m, I used the maximum likelihood method to extract arable range. I used confused matrix method to evaluate accuracy of the overall classification and precision of results of arable extraction. Results of the accuracy of the overall classification was 92.00%. Kappa coefficient was 0.69. Accuracy of classification is higher. User accuracy of arable extracted was 96.60%, and mapping precision was 95.30%. Accuracy of arable extraction was high.(2)Extract range of agricultural disaster. I used product data of Heilongjiang Province from 2006 to 2012, which is about the spatial resolution of 250 m and MODIS vegetation index, to present the mode that NDVI Intermediate values were normalized(RNDVI_M), to set the threshold to extract the scope of the disaster, and to test the disaster area from MODIS data accuracy based on MODIS data with higher resolution close to the date of HJ-1A / 1B CCD imaging. In 22 test results, the average relative error was 17.11%, verified accuracy was 82.89%, and accuracy of the extraction of the scope of the disaster based on MODIS data achieved the ideal results. The results showed that the threshold value to determine the spatial distribution of disaster RNDVI_M is objective and accurate method, and the method that uses RNDVI_M threshold to monitor regionalscale agricultural disaster is feasible.(3)Extract hail range. Based on the difference in landscape indices between hail and other agricultural disasters, I set plaque area, perimeter, and the threshold of shape index to exclude other disasters and extract hail. In 22 disasters, I extracted out of 8 hail in the hail of 10 and 2 hail in non hail types of 12. Correctly distinguish rate is 81.82 percent, and precision that extracts hail is high.(4) Analysis of spatial and temporal pattern characteristics of hail. According to the extracted hail range, respectively, hail distribution of Heilongjiang Province in 2012 were analyzed in the time and space. The results show that hail in 2012 in Heilongjiang province occurred mainly in June and July from the time of analysis, and in the space mainly in seven cities- Jiamusi, Shuangyashan, Jixi, Mudanjiang, Heihe, Qiqihar and Suihua.The results showed that the use of continuous multi-temporal product data, which is about the spatial resolution of 250 m and synthetic MODIS vegetation index of 16 d, about spatial and temporal disaster feature extraction on regional-scale, is with high accuracy based on RNDVI_M models. Hail range extraction with a high precision, depending on the scope of landscape indices, can be used for monitoring and evaluation of hail, which can provide technical support for early warning and forecasting hail and statistical data for meteorological department.. |
PDF Full Text Request