基于古气候数据同化框架的黄河中游地区近500a降水重建

Precipitation reconstruction in the middle reaches of the Yellow River over the last 500 years based on a paleoclimate data assimilation framework

关键词:  黄河中游地区  降水重建  古气候数据同化  LMR  CMIP6

Key words:  the middle reaches of the Yellow River  precipitation reconstruction  paleoclimate data assimilation  LMR  CMIP6

摘要:

黄河中游地区降水空间分布不均、年际变化大、降水季节集中，研究其百年时间尺度的演变规律， 为进一步分析该区域周期性和趋势性干湿变化，极端气候事件的发生，以及未来降水变化等提供更可靠的依据。文章通过古气候数据同化框架，将树轮代用资料分别与CMIP6全球气候模式中MIROC-ES2L和 MPI-ESM1-2-LR模式模拟结果相结合，重建黄河中游地区过去500 a （1501—2000年） 5—9月平均降水变化，利用CN05.1降水观测资料基准期（1961—2000年）评估两种模式重建该地区降水的能力。研究结果表明：利用这两种气候模式的降水重建结果与CN05.1降水观测资料的相关性都较好 （P＜0.01），该结果反映的过去500 a部分年际、年代际极端干湿变化和近几十年变干的趋势性变化与该地区及附近区域的重建结果一致；与MIROC-ES2L相比，MPI-ESM1-2-LR对黄河中游地区降水变化的模拟能力更强，且重建结果的可靠性也更高，表明气候模式的选择对于保证气候重建结果的可靠性至关重要。

Abstract:

Background, aim, and scope Precipitation in the middle reaches of the Yellow River has considerable regional differences, large inter-annual variability, and non-uniform seasonal precipitation. Due to these complex characteristics, the regional water resources have high sensitivity to climate change, significantly affected by frequent droughts and floods. However, most previous studies about the regional precipitation change are based on the precipitation data over the past few decades, which is difficult to show its evolution over long time scales.Therefore, we utilized a paleoclimate data assimilation framework for the last millennium reanalysis (LMR) to reconstruct the precipitation in the middle reaches of the Yellow River over the past 500 years, and compare the reconstruction capabilities of two global climate models (GCMs) including MIROC-ES2L and MPI-ESM1-2-LR in CMIP6. The aim of this study is to provide a reference for the selection of GCMs using data assimilation methods for precipitation reconstruction. Materials and methods This paper takes the middle reaches of the Yellow River (32°— 42°N, 100°—116°E) as the research region, uses MIROC-ES2L and MPI-ESM1-2-LR in CMIP6 to provide prior estimates of the precipitation, and selects the tree-ring proxy records with a Pearson correlation coefficient of CN05.1 instrumental precipitation data not less than 0.2. Here, LMR is used to combine GCM data and proxy records to reconstruct annual resolution precipitation from May to September (during 1501—2000) over the middle reaches of the Yellow River. Taking 1961—2000 as the reference period, the performance of precipitation reconstruction in the region by the MIROC-ES2L and MPI-ESM1-2-LR in CMIP6 was evaluated against CN05.1 instrumental precipitation data during this reference period. Results Our results show that MPI-ESM1-2-LR (r=0.83, RMSE=1.42, MB=0.81) has stronger spatial correlation and smaller RMSE with the CN05.1 instrumental data, compared with MIROC-ES2L (r=0.79, RMSE=2.44, MB=1.94). Moreover, the corresponding regional precipitation is much closer to the instrumental data. These results indicate that the MPI-ESM1-2-LR model has an enhanced ability to simulate precipitation from May to September in this region. According to the spatio-temporal validation of the reconstructed precipitation in the reference period, it can be seen that the reconstruction results of these two models have good correlation with the instrumental data (P＜0.01), and the MPI-ESM1-2-LR model (r=0.59, RMSE=0.29, MB=0.04) has stronger reconstruction skill than MIROC-ES2L (r=0.52, RMSE=0.31, MB=0.02). Through the analysis of reconstruction sequences in the past 500 years, it was found that the reconstruction results are basically consistent with several dry and wet periods for interannual and interdecadal scales found in other research results. Discussion Compared with MIROC-ES2L, the MPI-ESM1-2-LR model has a stronger correlation with instrumental data. The reconstruction results of the reference period also show that MPI-ESM1-2-LR has an enhanced ability to reconstruct precipitation as a prior estimate. It indicates that the selection of climate models is significant for the reliability of climate reconstruction results. It is noted that several dry-wet periods in the reconstruction results are inconsistent with the results of other studies. On the one hand, the reconstruction region and timespan of this study are not completely consistent with those of other studies, resulting in slight deviations in the interannual changes in the reconstruction results; on the other hand, the proxy records used in this study are sparsely distributed in the marginal areas of the middle reaches of the Yellow River, and the ability of GCMs to simulate climate fields needs to be improved, which limits the accuracy of precipitation reconstruction results. Conclusions The precipitation reconstruction results in the middle reaches of the Yellow River over the past 500 years using the LMR framework are basically consistent with several dry and wet periods in other research results in the region and nearby. In addition, compared with MIROC-ES2L, the MPI-ESM1-2-LR model has a stronger ability to simulate precipitation fields and reconstruct precipitation, indicating that the choice of GCMs can have a significant impact on the results of precipitation reconstruction and improving the ability of climate models to simulate climate fields can effectively improve the reliability of climate reconstruction. Recommendations and perspectives This study can provide a reference for GCM selection when using data assimilation methods for precipitation reconstruction, and show an in-depth revelation of the periodic and trend changes in the regional climate on a long-time scale. This research is of great significance for formulating climate change adaptation strategies and protecting the ecological environment. However, the reconstruction research has certain limitations due to the lack of available proxy records and the large difference between GCMs and instrumental data. Therefore, in the future, deep-learning downscaling methods can be used to enhance their ability to simulate the climate field by increasing the spatial resolution of GCMs, thereby improving the reliability of reconstructing the climate field using paleoclimate data assimilation methods.