大气强迫数据的不确定性对地表温度模拟结果的影响研究——以黄河流域上中游为例

地表温度Land Surface Temperature(LST)是气候变化的重要指标，地表温度时空分布的准确估计，对加强地表水文过程的理解以及预测未来干旱状况至关重要。通用陆面模型模拟结果的准确性与大气强迫数据的质量紧密相关，强迫数据自身的误差会传递至模拟结果。本研究利用了Climatic Research Unit-National Centers for Environmental Prediction(CRUNCEP)、Global Soil Wetness Project(GSWP)及China Meteorological Forcing Dataset(CMFD)三套大气强迫数据集、中国地面气候资料日值数据集(V3.0)、中分辨率成像光谱仪Moderate-resolution Imaging Spectroradiometer(MODIS)反演的地表温度数据集，以通用陆面模型Community Land Model 5.0(CLM5.0)为研究工具，探讨了大气强迫数据的不确定性对模型模拟LST的影响，分析了黄河流域上中游LST时空分布特征以及不同大气强迫数据集驱动下LST的模拟特点。研究表明:(1)2003-2010年，三套大气强迫数据中的气温和入射太阳辐射存在一定的不确定性。在气温的时间变化方面，三套大气强迫数据变化趋势较一致，波动与不同数据间差距都较小，呈现GSWP的气温最高，CRUNCEP次之，CMFD最低的局面。在入射太阳辐射的时间变化方面，三套大气强迫数据年际变化趋势有一定差异，CRUNCEP数据在数值上与另两套数据有较大差别，是三套中最高的。入射太阳辐射在秋冬季，CMFD数据最低;在春夏季，CRUNCEP数据最高。空间上，三套数据在四季均表现出东南部气温较高，东北和西部气温较低的特征。CRUNCEP和GSWP数据空间分布相似，CMFD数据的气温空间分布更细致，与另两套数据的差异较大。入射太阳辐射呈现出东南部较低，西部青藏高原区域较高的特征;CRUNCEP的数据较高，波动较小，CMFD数据空间分布波动较大。与中国地面气候资料日值数据集相比发现，研究区东部的数据精度显著高于研究区西部的数据精度。且CMFD数据的不确定性最小，GSWP数据次之，CRUNCEP数据不确定性最大。总的来说，由于CMFD数据在现有的多个数据资料的基础上融合了更多地面气象站观测数据，分辨率也较高，因此，CMFD数据提供的气温和入射太阳辐射数据在三套数据中最佳。(2)2003-2010年，三套大气强迫数据驱动下的黄河流域上中游地表温度年际变化趋势较一致，但CLM-GSWP模拟的地表温度最高，CLM-CRUNCEP次之，CLM-CMFD最低。在空间上，三套大气强迫数据模拟的地表温度均能较好地刻画出MODIS地表温度数据的空间分布特征:研究区西部LST较低，关中平原LST较高，均体现出LST随纬度、季节、地形、和土地利用类型变化而变化的特点。在不确定性方面，三套数据的模拟结果都表现出研究区东部不确定性明显低于西部区域不确定性。除冬季CLM-CMFD结果不是最佳外，其他三个季节CLM-CMFD方案的模拟结果均最佳。综合分析后，得出结论:CLM-CMFD方案的LST模拟结果最优，但其驱动模拟冬季LST时的准确性有待进一步提高。(3)三种LST模拟结果在时间尺度上的大小排序、在空间上的分布特征以及不确定性的特点与三套大气强迫数据集中的气温特征是一致的。CMFD数据驱动的模拟结果高估了关中平原的地表温度，主要是来自大气强迫数据中气温的不确定性。春秋季节，三种方案的LST模拟结果都表现出与气温和入射太阳辐射较高的相关性，而在冬夏季节，相关性较低;相比于入射太阳辐射，LST与气温有着更加显著的相关性，说明CLM在模拟LST时，受大气强迫数据中的气温影响更大，在今后对LST进行研究时，应尤其注意选择气温准确性较高的强迫数据集驱动模型。总的来说，大气强迫数据集中的气温和入射太阳辐射的不确定性会通过模型传递到地表温度的模拟结果，未来需要不断地提高大气强迫数据集的精确性来提高在黄河流域上中游地表温度模拟的精确性。

Land Surface Temperature (LST) is a vital indicator of climate change. Accurate estimation of the spatiotemporal distribution of LST is critical for advancing the understanding of surface hydrological processes and forecasting future drought conditions. The accuracy of simulations generated by land surface models is closely linked to the quality of atmospheric forcing datasets, and errors inherent in these forcing datasets can propagate into the model outputs. In this study, we employed three atmospheric forcing datasets: Climatic Research Unit-National Centers for Environmental Prediction (CRUNCEP), Global Soil Wetness Project (GSWP), and China Meteorological Forcing Dataset (CMFD), alongside the Daily Dataset of Chinese Surface Climate Data (V3.0) and the LST dataset retrieved from the Moderate-resolution Imaging Spectroradiometer (MODIS). Using the Community Land Model 5.0 (CLM5.0) as the research tool, we examined the impact of uncertainties in atmospheric forcing datasets on model-simulated LST, and analyzed the spatiotemporal distribution characteristics of LST in the upper and middle reaches of the Yellow River Basin, as well as the simulation performance of LST driven by different atmospheric forcing datasets. The findings of this study are as follows: (1) From 2003 to 2010, the air temperature and incident solar radiation among the three atmospheric forcing datasets exhibited certain uncertainties. Regarding the temporal variation of air temperature, the three datasets exhibited consistent trends, with minimal fluctuations and small disparities between different datasets; specifically, GSWP recorded the highest air temperature, followed by CRUNCEP, while CMFD had the lowest. For the temporal variation of incident solar radiation, the interannual trends of the three datasets showed certain discrepancies. CRUNCEP had notably different numerical values compared to the other two datasets, being the highest among the three. Regarding seasonal distribution, CMFD had the lowest incident solar radiation in autumn and winter, while CRUNCEP had the highest values in spring and summer. Spatially, across all four seasons, all three datasets displayed the pattern that air temperature was higher in the southeastern region and lower in the northeastern and western parts of the study area. CRUNCEP and GSWP had similar spatial distributions, whereas CMFD exhibited a more detailed spatial distribution of air temperature, with larger deviations from the other two datasets. Incident solar radiation showed the pattern of being lower in the southeast and higher in the western Tibetan Plateau region; CRUNCEP had relatively higher values with small fluctuations, while CMFD had a spatially more fluctuating distribution. Compared with the Daily Dataset of Chinese Surface Climate Data (V3.0), it was found that the data accuracy in the eastern part of the study area was significantly higher than that in the western part. Additionally, CMFD had the smallest uncertainty, followed by GSWP, while CRUNCEP exhibited the largest uncertainty among the three datasets. Overall, as CMFD integrates more ground meteorological station observation data based on multiple existing datasets and has a higher spatial resolution, the air temperature and incident solar radiation data provided by CMFD are the most optimal among the three datasets. (2) From 2003 to 2010, the interannual variation trends of LST in the upper and middle reaches of the Yellow River Basin driven by the three atmospheric forcing datasets were relatively consistent, yet the CLM-simulated LST under the CLM-GSWP scheme was the highest, followed by CLM-CRUNCEP, while CLM-CMFD yielded the lowest LST values. Spatially, the LST simulated by all three atmospheric forcing datasets could well characterize the spatial distribution patterns of the MODIS LST dataset: the western part of the study area had lower LST, while the Guanzhong Plain had higher LST, all reflecting that LST varies with latitude, season, topography, and land use type. In terms of uncertainty, the simulation results of all three datasets showed that the uncertainty in the eastern part of the study area was significantly lower than that in the western region. Except for the CLM-CMFD scheme not performing optimally in winter, the CLM-CMFD scheme achieved the best simulation performance in the other three seasons. Through comprehensive analysis, we concluded that the LST simulation results of the CLM-CMFD scheme are the most optimal, but its accuracy in simulating winter LST requires further improvement. (3) The ranking of the three LST simulation results across temporal scales, their spatial distribution characteristics, and the patterns of uncertainty are consistent with the air temperature characteristics of the three atmospheric forcing datasets. The simulation results driven by CMFD overestimated the LST in the Guanzhong Plain, which was mainly attributed to the uncertainty of air temperature in the atmospheric forcing datasets. In spring and autumn, the LST simulation results of the three schemes showed relatively high correlations with air temperature and incident solar radiation, while in winter and summer, the correlations were lower. Compared with incident solar radiation, LST had a more significant correlation with air temperature, indicating that CLM is more affected by air temperature in the atmospheric forcing datasets when simulating LST. Therefore, when conducting LST research in the future, special attention should be paid to selecting forcing datasets with higher air temperature accuracy to drive the model. Overall, the uncertainties of air temperature and incident solar radiation in atmospheric forcing datasets can be propagated to LST simulation results via the model, and the accuracy of LST simulation in the upper and middle reaches of the Yellow River Basin needs to be continuously improved by enhancing the precision of atmospheric forcing datasets in the future.

搜集汇总

背景与挑战

背景概述

该数据集基于一项研究，探讨了三种大气强迫数据集（CRUNCEP、GSWP、CMFD）在黄河流域上中游地区对地表温度模拟不确定性的影响，使用CLM5.0陆面模型和MODIS数据进行分析。研究发现CMFD数据因融合更多地面观测且分辨率高，其模拟结果最优，但冬季准确性需改进，同时模型模拟更易受气温不确定性影响。数据集包含4.4 KB的文本格式数据，聚焦于大气强迫数据不确定性、地表温度模拟和陆面模型应用。