整合1982-1999年NDVI与降雨量时间序列模拟中国北方温带草原植被盖度

Simulating Vegetation Fractional Coverage for Temperate Grassland in Northern China Combining NDVI with Precipitation Time Series from 1982 to 1999

植被盖度是刻画陆地表面植被数量的一个重要参数,是指示生态系统变化的重要指标.因为中国北方温带草原对降雨量变化较敏感,因此,可以利用降雨量与遥感数据结合估测植被盖度.文章利用基于遥感数据和降雨量的线性分解模型来估测不同植被类型盖度,每8km×8km地区的覆盖类型分为3种类型:裸地、灌丛及草地和裸地混合类型,混合类型又分为:草地和裸地.利用一元线性回归拟合模型计算的湿季NDVI对降雨量的敏感性作为模型的状态变量,将通过0.1显著水平检验的NDVI和降雨量线性关系的斜率与湿季NDVI均值作散点图,然后根据条件限制优化算法求解三角形确定模型中的端元参数.1982—1999年平均占地率、灌丛和草地裸地混合类型的平均盖度研究表明,占地率的空间变化特征是自西北向东南逐渐减小.灌丛的植被盖度自西北向东南逐渐增加,没有表现出与降雨量相似的变化格局.大部分地区的灌丛盖度小于30%,少部分地区的灌丛盖度大于50%.草地和裸地混合类型盖度并没有表现出与降雨量相似的空间变化特征.单年植被盖度研究表明,1996年(干旱年)草地盖度由西北向东南逐渐增大,与湿季降雨量具有相似的空间变化特征.1998年(湿润年)湿季降雨量充沛,大部分地区的草地盖度较高.利用地面测量数据检验基于线性分解模型模拟的1999年草地盖度,平均误差为0.147.与Gutman模型相比,线性分解模型模拟结果较好.

Vegetation fractional coverage is an important parameter depicting vegetation amount on the land surface and a significant index indicating ecosystem change. Since temperate grassland in northern China is more sensitive to rainfall, vegetation fractional coverage can be estimated combining remotely sensed data and rainfall time series. A linear unmixing model based on NDVI normalized difference vegetation index and rainfall was used to estimate fractional coverage. Three types of endmembers were selected for each pixel covered 8× 8km^2 ： bare soil, shrub and mixed type which was divided into bare soil and grass in advance. Endnumber parameters for this analysis were determined by condition constrained optimization algorithm, taking wet season precipitation and sensitivity of NDVI to wet season precipitation as state variables. Scatter plot between mean NDVI in wet season and the slope of precipitation and NDVI linear relationship that pass significance testing of 0. 1. Study of average fractional coverage from 1982 to 1999 show that, the fractional coverage of bare soil indicates a clear decrease from Northwest to Southeast, and that of shrubs shows a general increase having no obvious similarity to mean wet season precipitation variability. The fractional coverage of shrubs in most areas was less than 30% ,while a little area was more than 50%. The similarity did not hold for the space variable pattern between fractional coverage of mixing type of grassland and bare boil and wet season precipitation. Results of fractional coverage for a single years showed that the fractional coverage of grasses increased gradually from Northwest to Southeast in 1996 （dry year）, which was similar to space variable characteristic of the mean wet season precipitation. Higher fractional coverage of grasses appeared in most areas in 1998（wet year） because of abundant rainfall. Furthermore, the fractional coverage of grassland by linear unmixing model in 1999 was tested with field measure data with an average error of 0. 147 which stands for better simulation precision. Compared with Guman model, linear unmixing model had better simulation precision.