结合TRMM数据的区域降水高精度曲面建模研究
作者简介:张 涛(1987-),男,四川泸县人,博士生,研究方向为遥感环境变化检测。E-mail: zhangtao@lreis.ac.cn
收稿日期: 2015-01-18
要求修回日期: 2015-03-05
网络出版日期: 2015-08-05
基金资助
国家“973”计划项目(2015CB954101、2015CB954103)
国家科技支撑计划项目(2012BAH33B01)
Analysis on High Accuracy Surface Modeling in Regional Rainfall Estimation Combined with TRMM Data
Received date: 2015-01-18
Request revised date: 2015-03-05
Online published: 2015-08-05
Copyright
高精度曲面建模方法(High Accuracy Surface Modeling, HASM),从理论上解决了传统方法在插值过程中峰值削平和边界震荡等问题。其模拟精度相对于经典插值方法有很大提高,已成功应用于人口密度、土壤属性,以及气候要素等领域的空间制图。然而,由于地面气象站点数量和分布的限制,使得HASM仅依靠站点数据难以得到高精度的空间降水估计数据,因此,本文以地貌与气候类型复杂多样的我国中西部地区2010年年降水量空间分布模拟为例,采用混合插值法进行HASM区域降水模拟。结果表明,TRMM作为背景场的HASM模拟的年降水量精度,在全局和局部明显优于IDW、Spline和Kriging等经典插值方法的结果,作为背景场的HASM模拟精度,MAE和RMSE分别为125.15 mm和155.80 mm,其他方法最好的模拟结果比其误差值分别高出53.6%和54.5%;其模拟误差在不同子区域都较小;各种方法在平原的精度都高于山区的精度。
关键词: 高精度曲面建模(HASM); TRMM; 空间插值; 降水
张涛 , 李宝林 , 赵娜 , 许丽丽 . 结合TRMM数据的区域降水高精度曲面建模研究[J]. 地球信息科学学报, 2015 , 17(8) : 895 -901 . DOI: 10.3724/SP.J.1047.2015.00895
High Accuracy Surface Modeling (HASM) method has theoretically solved issues of hill peak smoothing and oscillation phenomenon at edges, and its modeling accuracy is much better than the traditional interpolation methods such as Inverse Distance Weighting (IDW), Spline and Kriging. HASM has been successfully applied to the spatial mapping in multiple fields, such as population density, soil properties and climatic elements, etc. However, as the number and distribution of meteorological rain gauges are limited, getting the accurate precipitation distribution maps based on HASM is still a challenge. Additionally, remote sensing rainfall estimation data, which can provide better spatial information of the precipitation, but without accurate rainfall values, may play an important role. Therefore, in this study, we combine these two data sets together based on HASM model to estimate regional rainfall. Central and western China (25°~35°N, 105°~115°E), which are featured by extensive high mountains and plains, is chosen as the study area to model the spatial distribution of its total precipitation in 2010. Using satellite rainfall estimation, the Tropical Rainfall Measuring Mission (TRMM) 3B43 data is chosen as the background field for HASM modeling. Then, we compare its results with respect to the classical methods (including IDW, Spline and Kriging) based (also used as background fields) HASM modeling. Results show that TRMM based HASM method has higher accuracy and its results exhibit a better spatial pattern for precipitation simulation than those from the other methods. The MAE and RMSE of TRMM based HASM simulation results are 125.15 mm and 155.80 mm, respectively. The simulation errors of the best simulation results using the other methods are respectively 53.6% and 54.5% higher than TRMM based HASM simulation results. Besides, its relative error in each sub-region is also smaller than the other methods. In the multiple applications of spatial elements modeling, e.g. meteorological elements modeling, where there is not enough sampling sites to characterize the spatial structure of an element, the accuracy of HASM modeling will be limited. Therefore, combining it with supplementary information to compensate the deficiency of limited sampling sites will contribute to the production of better results for HASM applications.
Fig. 1 Locations of the study area and meteorological stations图1 研究区位置与气象站点分布 |
Fig. 2 Spatial distribution of rainfall data obtained using HASM model with different background fields图2 不同背景场HASM模拟的降水量分布 |
Tab. 1 Accuracy of HASM with different trend surfaces表1 不同背景场的HASM计算精度(mm) |
验证点 | 指标 | HASM_I | HASM_S | HASM_K | HASM_T |
---|---|---|---|---|---|
全局 | MAE | 212.32 | 233.59 | 192.27 | 125.15 |
RMSE | 260.27 | 327.75 | 240.69 | 155.80 | |
局部 | MAE | 195.95 | 197.77 | 196.84 | 167.53 |
RMSE | 262.61 | 259.71 | 256.39 | 228.81 |
Fig. 3 Relative errors of HASM rainfall modeling with different background fields at validation stations图3 验证站点处不同背景场的HASM降水模拟相对误差 |
Fig. 4 Average slope in the study area图4 研究区平均坡度 |
Tab. 2 Root mean square of the relative errors for validation station groups classified by mean slope表2 按平均坡度分级后各组验证点相对误差的均方根值 |
区域类型 | 坡度范围(°) | 验证点数(个) | HASM_I(%) | HASM_S(%) | HASM_K(%) | HASM_T(%) |
---|---|---|---|---|---|---|
G1 | 0~8 | 6 | 31.38 | 37.72 | 25.79 | 13.89 |
G2 | 8~15 | 6 | 13.68 | 12.16 | 12.66 | 9.05 |
G3 | 15~30 | 4 | 29.40 | 30.88 | 27.91 | 23.78 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
|
[19] |
|
/
〈 | 〉 |