基于MODIS植被指数评估洞庭湖区东方田鼠在大暴发年份的危害

Abstract

It is very important to develop a rapid algorithm for rodent pest damage estimation,especially for an outbreak year. The Yangtze vole( Microtus fortis) has been identified as a primary rodent pest in the Dongting Lake region. In July2007,there was a sudden population explosion in that region,representing the most serious Yangtze vole disaster on record.More than two billion voles crossed the lake levee,causing massive crop damage. The study area of Beizhouzi,to the west of East Dongting Lake,was selected because it had the highest population density during the outbreak year. A difference method of vegetation index was used in the assessment,in which the outbreak year 2007 was contrasted with 2004,which was considered a normal year. All vegetation-index imagery was from the Moderate Resolution Imaging Spectroradiometer( MODIS),which was masked by a land-use map of the study area to remove water bodies and urban and built-up lands.The Normalized Difference Vegetation Index( NDVI) in the outbreak year declined during the vegetation growth period,in contrast with the normal year. This NDVI change was a result of plant destruction by the Yangtze vole population. Regarding the local crop system,Julian days 193 and 209 were chosen for observations representative of before and after the Yangtzevole problem. Performances of the NDVI and enhanced vegetation index( EVI) were analyzed. ΔVI was defined as the ratio of a seasonally detrended vegetation index( VI) reduction before and after the Yangtze vole problem to seasonally detrended VI before the problem. Given the mean ΔVI,shift amplitude( the difference of ΔVI between 2004 and 2007),statistical dispersion( inter quartile range( IQR) of ΔVI in 2007),and percentage of detected damaged pixels,the damage threshold was defined as the sum of the mean ΔVI and mean absolute deviation. These were calculated to evaluate sensitivity of the vegetation index. The EVI was selected as the optimal vegetation index for detecting damaged area after sensitivity analysis.An ΔEVI of 0.07 was considered the damage threshold in the study area. The damage indicator was then rescaled to a range of 0 to 1,for constructing damage severity. The upper and lower limits for rescaling were determined by statistical analysis of the damage indicator,whose values were the 95% point of the ΔEVI histogram for the study area and the damage threshold.Only damage pixels greater than 0 after rescaling,and the 25% and 75% quantiles,were treated as thresholds for three damage levels( serious,moderate and light). As a result,the study area was separated into four groups,in which 80.30%of pixels were damaged. The damage pixels were aggregated according to the action of the Yangtze vole population during the outbreak year,and were dominated by moderate-damage pixels. There was a certain distance between the damage area and levee,which could been explained by the population migration. Pixels with serious damage were dispersed over the entire study area. We constructed a disaster region map for the Dongting Lake region in 2007,from investigation and the spatial distribution of damage pixels with the threshold 0.07. This map was used to validate the method of assessing outbreak-year rodent pest damage based on MODIS imagery. The new approach to evaluating rodent pest damage proved to be rapid and promising.