一种NCEP/NCAR再分析气温数据的统计降尺度方法

A Statistical Downscaling Approach of NCEP/NCAR Reanalysis Temperature Data

近地表气温是影响陆表过程的一项重要因素，中高分辨率的栅格化气温数据是生态环境、水文、水循环等模拟和分析的重要参数，获得准确和较高空间分辨率的栅格化气温数据对更好地理解陆地表面过程和全球变化等有非常重要的意义。本文提出一种基于气温与地形、植被等环境因子，以及地理位置的统计降尺度方法，以回归树模型建立气温与NDVI、DEM及地理位置之间的定量关系，将NCEP／NCAR近地表气温数据降尺度到公里级分辨率，并利用该方法得到2000年1月至2010年12月全国陆地范围ikm分辨率逐月月平均气温数据。最后，采用全国380个气象站点的观测数据对结果进行对比分析，结果表明：该方法得到的气温数据可以有效地反映全国陆地范围气温空间分布特点和月际变化趋势，验证结果的R2范围在0．861—0．95之间，RMSE范围在1．88-2．68℃之间。

Near-surface air temperature is an important controlling parameter for land surface processes, and is critical to ecological, environmental and hydrological modeling. Temperature records observed at meteorological stations have been widely used, but there has been an increasing need for temperature data in grid for modeling purposes. Although grid temperature can be estimated from in-situ temperature records using interpolation algo- rithm, low accuracy have been reported due to limited ground stations and their clustering distribution, especially when there were insufficient sites to represent all land cover types and terrain conditions in the area. NCEP/ NCAR reanalysis project uses a frozen state-of-art global data assimilation system and a database as complete as possible. Although the NCEP/NCAR data has a coarse resolution （0.5 degree）, it provides global, consistent, and long term estimation of climate variables. This paper presents a downscaling approach to derive monthly temper- ature at lkm resolution from the NCEP/NCAR by utilizing derived relationships between monthly aggregated NCEP/NCAR temperature and other ground elements, i.e., terrain, vegetation and geographic locations. Regres- sion tree model was chosen to detect the possible relationships. Monthly temperature with lkm resolution for China land area from 2000 to 2010 has been produced using the approach. The final predicted temperatures were compared with observed records at 380 meteorological stations in China. The results indicate that the down- scaled estimations can represent spatial distribution and trends and the magnitude of inter-month temperature with R2 ranging from 0.861 to 0.95, and RMSE from 1.88℃ to 2.681℃.