干旱区研究 ›› 2022, Vol. 39 ›› Issue (6): 1706-1716.doi: 10.13866/j.azr.2022.06.02

• 天气与应用气候 • 上一篇    下一篇

青海湖流域及周边区域TRMM 3B43降水数据降尺度方法对比分析

李炎坤1,2(),高黎明1,2,3,张乐乐1,2,3(),吴雪晴1,2,刘轩辰1,2,祁闻1,2   

  1. 1.青海师范大学地理科学学院,青海 西宁 810008
    2.青海师范大学青藏高原地表过程与生态保育教育部重点实验室,青海 西宁 810008
    3.青海省人民政府-北京师范大学高原科学与可持续发展研究院,青海 西宁 810008
  • 收稿日期:2022-04-16 修回日期:2022-09-07 出版日期:2022-11-15 发布日期:2023-01-17
  • 通讯作者: 张乐乐
  • 作者简介:李炎坤(1997-),女,硕士研究生,研究方向为遥感信息分析及地学应用. E-mail: liyankunkkk@163.com
  • 基金资助:
    国家自然科学基金(42171467);国家自然科学基金(42001060);青海省基础研究计划(2021-ZJ-947Q)

Comparison of downscaling methods for TRMM 3B43 precipitation data in the Qinghai Lake Basin and its surrounding areas

LI Yankun1,2(),GAO Liming1,2,3,ZHANG Lele1,2,3(),WU Xueqing1,2,LIU Xuanchen1,2,QI Wen1,2   

  1. 1. College of Geography Science, Qinghai Normal University, Xining 810008, Qinghai, China
    2. Key Laboratory of Tibetan Plateau Land Surface Processes and Ecological Conservation (Ministry of Education), Qinghai Normal University, Xining 810008, Qinghai, China
    3. Academy of Plateau Science and Sustainability, People’s Government of Qinghai Province and Beijing Normal University, Xining 810008, Qinghai, China
  • Received:2022-04-16 Revised:2022-09-07 Online:2022-11-15 Published:2023-01-17
  • Contact: Lele ZHANG

摘要:

利用多元线性回归模型(MLR)、主成分逐步回归分析模型(PCSR)、克里金插值法(Kriging),将青海湖流域及周边区域0.25°分辨率的TRMM 3B43降水数据降尺度至0.01°分辨率,并选取研究区范围内20个气象站点的实测降水数据,利用相关系数(CC)、均方根误差(RMSE)和相对偏差(Bias)对降尺度结果进行评价,以此选取更适用于研究区的降尺度方法。结果表明:(1) 基于 TRMM及3种降尺度方法获取的研究区降水空间分布具有一致性,年均及春、夏、秋季的降水量均表现为北部高,西部及西北部低,而冬季降水量表现为南部与西北部高,中部低。(2) 研究区降水量随着海拔的增高,以3800 m为界整体上呈现先升高后降低的趋势。(3) 精度评价的结果表明,年尺度上的Kriging精度表现最好;在空间上TRMM及3种降尺度数据在东部地区精度最优。在季尺度上,数据精度表现为PCSR>Kriging>TRMM>MLR;在月尺度上,PCSR数据精度最优。(4) 海拔对研究区内TRMM及3种降尺度数据的影响较小,但随着海拔的升高,遥感数据逐渐出现低估降水的现象,其可能原因在于降水的低估与微波降水率反演时对对流性降水的低估有关。综合降水空间分布一致性分析与精度评价,认为PCSR最适合于青海湖流域及周边区域的TRMM 3B43降水数据降尺度方法。

关键词: TRMM 3B43, 降尺度, 多元线性回归, 主成分逐步回归, 克里金插值, 青海湖流域

Abstract:

Using multiple linear regression (MLR), principal component stepwise regression (PCSR), and Kriging, the TRMM 3B43 precipitation data in the Qinghai Lake Basin and surrounding areas with a resolution of 0.25° were downscaled to a resolution of 0.01°. The measured precipitation data of 20 meteorological stations in the study area were selected, and the correlation coefficient, root mean square error, and relative deviation (Bias) were used to evaluate the downscaling results. Downscaling methods for the study area. The results show a consistent spatial distribution of precipitation in the study area based on TRMM and the three downscaling methods. The annual average precipitation and the three seasons of spring, summer, and autumn are all high in the north, low in the west and northwest, and winter precipitation. The performance was high in the south and northwest and low in the middle. With increased altitude, the precipitation in the study area showed an overall trend of first increasing and then decreasing with 3800 meters as the boundary. The results of the accuracy evaluation show that the Kriging accuracy on the annual scale has the best performance. On the spatial scale, the TRMM and the three downscale data have the best accuracy in the eastern region. On the quarterly scale, the data precision is PCSR > Kriging > TRMM > MLR. On the monthly scale, the PCSR data accuracy is the best. The effect of altitude on the TRMM and the three downscaling data in the study area is small. However, with increasing altitude, the remote sensing data gradually underestimates the precipitation phenomenon, which may be due to the underestimation of precipitation and convection during microwave precipitation rate inversion underestimation of precipitation. From the comprehensive precipitation spatial distribution consistency analysis and precision evaluation, it is considered that PCSR is the most suitable downscaling method for TRMM 3B43 precipitation data in the Qinghai Lake Basin and surrounding areas.

Key words: TRMM 3B43, downscaling, multiple linear regression, principal component stepwise regression, kriging, Qinghai Lake Basin