Study Of Drought Remote Sensing Model In Eastern Agricultural Region Of Qinghai Province

Drought events usually expand slowly but will impact large areas. So, timely andaccurate grasping of the degree and extent of drought becomes the premise of effectivelyresponse to drought disasters. The eastern agricultural region of Qinghai province plays a veryimportant role in development of the Qinghai’s agriculture and the national economy.However, drought is a seriously restricted factor of agricultural sustainable development ineastern agricultural region. It is of great signifcance in enhancing the initiative in combatingthe drought and the ability of disaster prevention and mitigation to strengthen droughtmonitoring and prediction in the studied area.In this paper, combining the ground measured meteorological data, remote sensingdata with digital elevation model, ten day drought remote sensing monitoring modelof studied area is built. The drought situation in the studied area is monitored using this model,at the same time, the monitoring results and the accuracy of model are analyzed. The mainconclusions of this paper are as follows:(1) Using the remote sensing data and meteorological data from2003and2004July,combining tools such as RS and GIS, a ten day drought remote sensing monitoring modelcalled DI is built considering several parameters: the typical remote sensing drought indextemperature vegetation drought index(TVDI), the reference crop evapotranspiration(ET0) andunderlying surface conditions of vegetation coverage(VC).By using of remote sensing data and meteorological data, the third parameters used inbuilding the DI are inversely calculated and estimated. First of all, a ternary linear inversionmodel of ET0is established by using remote sensing images and meteorological data in2003and the evaluation of the model is performed. The evaluation shows that the model workswell in inversion process. Then ET0on July2004is gotten by inversion using this model. Theresult shows the following fact: the value of ET0increases with the position changed fromnorthwest to southeast; along the zonal area in valley with low altitude, the value is high;along the ridge with high altitude, the value is lower. According to pixel statistic data, the ET0of every10-day mostly have a value between20mm and50mm; the percentage of data withsuch values are respectively92.32%,92.29%and93.26%for each10-day. Among the statistic values of ET0, the most frequent are between30mm and40mm which have apercentage of42.53%,50.89%and56.19%; while the least frequent values are between50mm and60mm which has a percentage1.38%,0.01%and0.01%. The average value of ET0for each ten day are35.86mm,32.62mm and33.88mm.Secondly, basing on dimidiate pixel model, the normalized difference vegetation index(NDVI) datasets deduced from MODIS product on July2004are used to calculate the VC foreach ten day. Pixel statistics show that: most the value of vegetation coverage for each ten dayare between0.7and1.0; the percentage of data with such values are60.82%,56.98%and64.59%for each10-day. Among the statistic values of VC, the most frequent are between0.8and0.9, the percentage of data with such values are respectively29.14%,28.76%and32.80%; while the least frequent values are between0and0.1which has a percentage1.90%,2.58%and1.98%. The average value of VC for each ten day are0.6911,0.6725and0.7246.And then, calculating typical remote sensing drought index TVDI, result shows that:most the value of TVDI for each ten day are between0.2and0.8; the percentage of data withsuch values are81.33%,85.94%and82.93%for each10-day. Among the statistic values ofTVDI the most frequent are between0.2and0.4which have a percentage of33.50%and39.15%on fist ten day and last ten day. The second ten day has the most frequent between0.4and0.6and a percentage of35.12%; while the least frequent values are between0and0.2which has a percentage8.85%,5.83%and7.91%. The average value of TVDI for each tenday are0.4926、0.5103and0.4745.(2) DI, the drought remote sensing monitoring model, is used in monitoring the droughton early August2004. The monitoring results show: in spatial distribution, the values of DI inthe valley of the Yellow River and Huangshui River Valley, as well as the region alongDatong River are higher than that of the rest of studied area. This fact suggests that theseplaces are prone to drought. DI drought classification gray scale image is generated in GIS fora visible manner to show the monitoring result. Pixel statistics show that: most the value of DIclassification between1and3; the percentage of data with such values are99.75%; the mostfrequent are1which has a percentage of44.26%while the least frequent is classification4which has a percentage of0.25%. The average value of DI is0.5300, showing thatdrought monitoring as a result of light drought.Extracting the DI values of area with different land cover types, the results show thatthere are11land cover types in the non-drought level. Among them are water, evergreenconiferous forest, deciduous coniferous forest, deciduous broad-leaved forest, mixed forest,open shrub lands, grassland with plenty of tree, wetland, crop, snow and bare land. There are4land cover types in light-drought level, including evergreen broad-leaved forest, dense shrubs, savanna and grasslands; only land cover type of city and urban district are in themiddle-drought level.(3) The DI model is validated and its accuracy is evaluated in this paper. The TVDI andDI classification value of14meteorological stations are extracted from gray scaleclassification image using the tools of ENVI4.7and ArcGIS9.3. And then, the extractedvalues are compared with the drought standard-M index-to calculate the accuracy of themodel. The results is shown as following: in early August2004,3stations-DatongHuangyuan and Huangzhong-have the same drought classification gotten separately byTVDI and M index; the monitoring precision is21.43%which is very low. On the other hand,the number of stations is8using DI to replace the TVDI; the stations are Datong, Haidong,Huangnan, Huangyuan, Huangzhong, Ledu, Menyuan and Xining. The monitoring precisionis57.14%which is significantly higher than using TVDI.The study results show that the drought monitoring of DI model have higher accuracythan that of TVDI with a value of35.71%.