基于CART模型陇西黄土高原潜在NDVI模拟

根据研究区30a的气象观测数据,使用栅格化技术,建成陇西黄土高原分布式气候模型(降水量、气温、蒸散发量);然后以人类活动影响较少的森林、灌木林和十草原作为该处的本底NDVI数据,使用分类与回归树模型建立潜在NDVI与栅格化的气候要素的关系,模拟了陇西黄土高原的潜在NDVI;在此基础上分析了外界压力对潜在NDVI的影响.结果表明:1)陇西黄土高原的多年平均降水量有从东南到西北的递减趋势,气温随海拔高度变化明显,多年平均生长季蒸散发量可达300 mm;2)使用CART模型模拟潜在NDVI精度较高,其总体精度达到了82.6%,可满足潜在NDVI的模拟;3)外界压力下的NDVI与潜在NDVI在空间分布上呈现很大的不一致.

Potential NDVI in Longxi Loess Plateau based on CART model

Data from some meteorologic stations were used to build a distributed climate models (precipitation, temperature and evapotranspiration), and then an approach was presented for estimating potential NDVI from climate variables and training data of actual NDVI in nature reserves. The actual data from the baseline generally correspond with land cover types in the western China where there are fewer human activities. The results show: 1) the precipitation had a trend from southeast to northwest, the temperature changed with elevation and the average evapotranspiration reached 30 cm; 2) The CART model was accurate enough in estimating the potential NDVI of Longxi Loess Plateau, with the accuracy reaching 82.6%; 3) There was an great difference between the actual NDVI and potential NDVI.