未来30年亚洲季风系统演变的动力-统计情景预估研究

This proposal aims at establishing a hybrid dynamical-statisticalscheme for reducing the uncertainties in future monsoon estimate in Asia. To estimate drought or flood in East China has become a hard task for scientists due to chaotic oscillations of Asian summer monsoon. There are apparent bias with severe climate drafts in future climate estimates from the multi-model ensemble means of the Coupled Model Intercomparison Project Phase 5 (CMIP5) under different scenarios (RCPs). So, the deviations must be corrected for reducing the uncertainties in the estimates. The error-correction model (ECM) should be designed against non-stationary error signals so as to be used in the corrections of CMIP5 future monsoon estimates, in which the monsoon is approximately represented by low-layer wind vector or moisture transport in this proposal. There are several steps toward the corrections. We, firstly, analyze the monsoon errors in CMIP5 historical simulations and remove the climate draft in them through the comparison with 20 century reanalysis datasets. Second, the non-stationary part in the corrected data is reduced by multi-difference method we developed. The third is to establish ECM by Bayesian model average (BMA) for the quasi-stationary monsoon data, and then the monsoon estimate is divided into two parts: internal modes and forced modes with the help of the orthogonal wavelet decomposition. The former is improved again by use of multi-deference regression basing on the air-sea coupled modes from reanalysis data. The corrected historical monsoon field is composed by the two parts, hereafter, and meanwhile the related parameters in ECM are updated. The two parts of future Asian monsoon fields are established respectively by the ECMs basing on CMIP5 multi-model outputs under different scenarios, which is the dynamic-statistical monsoon estimates with uncertainties for Asia. The decadal drying or flood episodes in East Asia are then estimated with the improved monsoon data for coming 30 years. The results can serve as a scientific reference against future climate hazards in East China.

中国东部旱涝格局变化与季风涛动密切关联，其预估一直是个难题。已经发现，国际海气耦合模式比较计划CMIP5的气候预估存在明显漂移或误差，必须进行有效订正，以减少其不确定性。因误差演变的非平稳性，需要设计压制非平稳信号的误差订正模型，以适用于未来季风情景的订正。首先，比较CMIP5历史气候模拟与20世纪再分析资料，分析低层风场或水汽输送场误差特征，剔除系统性漂移；其次用多尺度差分削弱模式或再分析数据的非平稳性，再用多尺度的贝叶斯模式平均(BMA)方法集成，给出CMIP5历史季风的订模型，对其用正交小波分解，分别建立季风自然变化和人类活动影响的订正模型，用计算的海气耦合模对季风自然变化订正模型做二次修正，二者的结合构成准平稳的误差订正模型，将其用于3种情景下亚洲季风的订正，得到未来30年亚洲季风的动力-统计情景预估, 评估季风变异对东亚年代尺度旱涝格局的影响并做不确定性分析，为防灾减灾服务。

本项目基于CMIP5数据研究未来30年亚洲季风的预估及偏差订正问题。 项目首先确定了用降水及地面到500mb水汽通量表示季风，克服了地形对季风表示的影响。其次，指出了降水预估的主要困难在于降水趋势太弱，甚至在许多地区无趋势，导致用多模式集合平均难以确定未来降水趋势的变化，传统降水预估结果的不确定性太大，难于应用。针对该问题，本项目检验了多种线性或广义线性偏差订正模型，提出基于观测记录的偏差订正方案，发现偏差订正方法具有区域性，即在30oN以北和以南分别用对数及差分回归效果最好。据此，制作了2021-2050亚洲区域经偏差订正后的暖季（5-10月）降水距平预估结果。这种基于观测记录的偏差订的预估属于一段年代的气候背景预测，可供气象或水文业务部门制作年际或年代尺度气候预测参考。与观测对比检验表明，该订正方法成功地估计出中国大陆汛期由“南涝北旱”转为“北涝南韩旱”的格局，即江南中部和西南干旱，大雨带北移，致使三江源区、西北东部、黄淮及东北南部多雨等。这与近年来我国汛期降水异常的旱涝分布形势基本一致。更有价值的是它对极端降水的落区也有指示意义，例如，2022年汛期亚洲的极端降水事件几乎也都落入该降水气候背景预测中的多雨区，如，夏季韩国首尔暴雨、阿富汗和巴基斯坦东南部暴雨、我国东北南部、华北、新疆南部、青海等地的极端降水。近几年南方的干旱，特别是2022年高温干旱也与该降水背景预测的干旱区基本吻合。此外，基于对流层低层水汽通量的季风预估结果指出，未来30年南亚索马里急流增强，向南亚西北部的水汽输送增强，南亚向东及向我国西南部的季风水汽输送减弱，会导致南亚西北部洪涝和我国西南部的干旱，这与近些年，特别是2022年汛期亚洲的旱涝分布形式基本一致。以上研究结果表明，利用CMIP5/CMIP6多模式预估数据，经适当的偏差订正后可以有效地应用于年际或年代尺度气候背景的预测估计，为气象业务部门制作年际或年代尺度气候预测提供了新的思路和方法，有利于提高气候预测的水平，为防灾减灾服务。

内容获取失败，请点击重试