| 摘要: |
| 【目的】以黄土高原地区为例，对温度和降水量2种气象要素的栅格化方法进行研究，为非气象站点所在地区或宏观大尺度区域气象要素数据的获取提供参考。【方法】 以直接插值法为对照，运用多元线性回归法和趋势面分析法，对分布于黄土高原及其周边地区127个气象站点1971-2000年的月平均温度和降水量与3种宏观地理因子（经度、纬度、海拔高度）之间分别建立回归关系，并在此基础上结合数字高程模型（DEM）和反距离权重法（IDW），对研究区2种气象要素数据进行栅格化，并选用8个气象站点对栅格化结果进行检测。【结果】 在对检验站点月平均温度的模拟中，直接插值法的平均绝对误差（MAE）大于1.0 ℃，而多元线性回归法和趋势面分析法的MAE值分别为0.485~0.776和0.242~0.509 ℃，多元线性回归法和趋势面分析法明显优于直接插值法，而趋势面分析法较多元线性回归法更优；但在检验站点月平均降水量的模拟中，3种栅格化方法在模拟精度上并无明显差别。3种宏观地理因子中，海拔高度是影响黄土高原地区温度空间分布最主要的因素，而纬度则对研究区范围内降水量的空间分布影响最大。【结论】 将宏观地理因子作为参数纳入到温度空间分布模型的构建当中，可以有效地提高大尺度范围内温度的模拟精度，并且趋势面分析法比多元线性回归法更具有优势；对降水量空间分布的模拟还存在诸多不确定性因素，需在方法上做进一步研究和探讨。 |
| 关键词: 黄土高原 温度 降水量 栅格化 直接插值法 多元线性回归法 趋势面分析法 |
| DOI： |
| 分类号: |
| 基金项目:中国科学院知识创新工程重要方向项目(KZCX2-YW-424)与重大项目(KSCX-YW-09-07) |
|
| Studying the methods for rasterizing meteorological variables in the Loess Plateau |
|
|
| Abstract: |
| 【Objective】 With the aim to provide some clues in acquiring meteorological information for those ungauged or large scale regions,this study took the Loess Plateau as the target area,which involved two types of meteorological variables temperature and precipitation to search proper methods for rasterization of meteorological data.【Method】 Here,multiple linear regression (MLR) and trend surface analysis (TSA) methods were employed to build regression relationships between the measured meteorological data and macro geographic factors (latitude,longitude and elevation).Then these relationships were combined with Digital Elevation Model (DEM) and Inverse Distance Weighting interpolation (IDW) in rasterizing monthly mean temperature and precipitation data derived from 127 meteorological stations for the period of 1971－2000 and data from 38 meteorological stations were taken to test the result.【Result】 It was found that,30 year monthly mean temperature and the MAE values were all above 1.0 ℃ under the scenario of applying direct interpolation,while the MAE values of multiple linear regression method ranged from 0.485 to 0.776 ℃ and trend surface analysis method from 0.242 to 0.509 ℃.So it can be seen that the latter two methods incorporated geographical factors much better than direct interpolation and trend surface analysis method performed better than multiple linear regression method.However,there was no big difference in precipitation between the results of the three methods.Among the three geographic factors,elevation was the most effective factor in predicting the spatial distribution of temperature over the Loess Plateau and latitude was the most influential factor affecting precipitation.【Conclusion】 In general,those methods like multiple linear regression and trend surface analysis methods involving macro geographical factors have great potential in improving the precision of temperature rasterization,particularly the trend surface analysis method.But for precipitation rasterization,there are so many uncertainties waiting to be explored further. |
| Key words: the Loess Plateau temperature precipitation rasterization direct interpolation multiple linear regression trend surface analysis |
