Projecting Spring Consecutive Rainfall Events in the Three Gorges Reservoir Based on Triple-Nested Dynamical Downscaling

Spring consecutive rainfall events(CREs) are key triggers of geological hazards in the Three Gorges Reservoir area(TGR), China. However, previous projections of CREs based on the direct outputs of global climate models(GCMs) are subject to considerable uncertainties, largely caused by their coarse resolution. This study applies a triple-nested WRF(Weather Research and Forecasting) model dynamical downscaling, driven by a GCM, MIROC6(Model for Interdisciplinary Research on Climate, version 6), to improve the historical simulation and reduce the uncertainties in the future projection of CREs in the TGR. Results indicate that WRF has better performances in reproducing the observed rainfall in terms of the daily probability distribution, monthly evolution and duration of rainfall events, demonstrating the ability of WRF in simulating CREs. Thus, the triple-nested WRF is applied to project the future changes of CREs under the middle-of-the-road and fossil-fueled development scenarios. It is indicated that light and moderate rainfall and the duration of continuous rainfall spells will decrease in the TGR, leading to a decrease in the frequency of CREs. Meanwhile, the duration, rainfall amount, and intensity of CREs is projected to regional increase in the central-west TGR. These results are inconsistent with the raw projection of MIROC6. Observational diagnosis implies that CREs are mainly contributed by the vertical moisture advection. Such a synoptic contribution is captured well by WRF, which is not the case in MIROC6,indicating larger uncertainties in the CREs projected by MIROC6.

Assessing the Performance of a Dynamical Downscaling Simulation Driven by a Bias-Corrected CMIP6 Dataset for Asian Climate

In this study,we aim to assess dynamical downscaling simulations by utilizing a novel bias-corrected global climate model(GCM)data to drive a regional climate model(RCM)over the Asia-western North Pacific region.Three simulations were conducted with a 25-km grid spacing for the period 1980–2014.The first simulation(WRF_ERA5)was driven by the European Centre for Medium-Range Weather Forecasts Reanalysis 5(ERA5)dataset and served as the validation dataset.The original GCM dataset(MPI-ESM1-2-HR model)was used to drive the second simulation(WRF_GCM),while the third simulation(WRF_GCMbc)was driven by the bias-corrected GCM dataset.The bias-corrected GCM data has an ERA5-based mean and interannual variance and long-term trends derived from the ensemble mean of 18 CMIP6 models.Results demonstrate that the WRF_GCMbc significantly reduced the root-mean-square errors(RMSEs)of the climatological mean of downscaled variables,including temperature,precipitation,snow,wind,relative humidity,and planetary boundary layer height by 50%–90%compared to the WRF_GCM.Similarly,the RMSEs of interannual-tointerdecadal variances of downscaled variables were reduced by 30%–60%.Furthermore,the WRF_GCMbc better captured the annual cycle of the monsoon circulation and intraseasonal and day-to-day variabilities.The leading empirical orthogonal function(EOF)shows a monopole precipitation mode in the WRF_GCM.In contrast,the WRF_GCMbc successfully reproduced the observed tri-pole mode of summer precipitation over eastern China.This improvement could be attributed to a better-simulated location of the western North Pacific subtropical high in the WRF_GCMbc after GCM bias correction.

Downscaling Seasonal Precipitation Forecasts over East Africa with Deep Convolutional Neural Networks

This study assesses the suitability of convolutional neural networks(CNNs) for downscaling precipitation over East Africa in the context of seasonal forecasting. To achieve this, we design a set of experiments that compare different CNN configurations and deployed the best-performing architecture to downscale one-month lead seasonal forecasts of June–July–August–September(JJAS) precipitation from the Nanjing University of Information Science and Technology Climate Forecast System version 1.0(NUIST-CFS1.0) for 1982–2020. We also perform hyper-parameter optimization and introduce predictors over a larger area to include information about the main large-scale circulations that drive precipitation over the East Africa region, which improves the downscaling results. Finally, we validate the raw model and downscaled forecasts in terms of both deterministic and probabilistic verification metrics, as well as their ability to reproduce the observed precipitation extreme and spell indicator indices. The results show that the CNN-based downscaling consistently improves the raw model forecasts, with lower bias and more accurate representations of the observed mean and extreme precipitation spatial patterns. Besides, CNN-based downscaling yields a much more accurate forecast of extreme and spell indicators and reduces the significant relative biases exhibited by the raw model predictions. Moreover, our results show that CNN-based downscaling yields better skill scores than the raw model forecasts over most portions of East Africa. The results demonstrate the potential usefulness of CNN in downscaling seasonal precipitation predictions over East Africa,particularly in providing improved forecast products which are essential for end users.

Seasonal Prediction of Indian Summer Monsoon Using WRF: A Dynamical Downscaling Perspective

Seasonal forecasting of the Indian summer monsoon by dynamically downscaling the CFSv2 output using a high resolution WRF model over the hindcast period of 1982-2008 has been performed in this study. The April start ensemble mean of the CFSv2 has been used to provide the initial and lateral boundary conditions for driving the WRF. The WRF model is integrated from 1st May through 1st October for each monsoon season. The analysis suggests that the WRF exhibits potential skill in improving the rainfall skill as well as the seasonal pattern and minimizes the meteorological errors as compared to the parent CFSv2 model. The rainfall pattern is simulated quite closer to the observation (IMD) in the WRF model over CFSv2 especially over the significant rainfall regions of India such as the Western Ghats and the central India. Probability distributions of the rainfall show that the rainfall is improved with the WRF. However, the WRF simulates copious amounts of rainfall over the eastern coast of India. Surface and upper air meteorological parameters show that the WRF model improves the simulation of the lower level and upper-level winds, MSLP, CAPE and PBL height. The specific humidity profiles show substantial improvement along the vertical column of the atmosphere which can be directly related to the net precipitable water. The CFSv2 underestimates the specific humidity along the vertical which is corrected by the WRF model. Over the Bay of Bengal, the WRF model overestimates the CAPE and specific humidity which may be attributed to the copious amount of rainfall along the eastern coast of India. Residual heating profiles also show that the WRF improves the thermodynamics of the atmosphere over 700 hPa and 400 hPa levels which helps in improving the rainfall simulation. Improvement in the land surface fluxes is also witnessed in the WRF model.

A generative adversarial network-based unified model integrating bias correction and downscaling for global SST

本文提出了一种基于生成对抗网络的全球海表面温度(sea surface temperature,SST)偏差订正及降尺度整合模型.该模型的生成器使用偏差订正模块将数值模式预测结果进行校正,再用可复用的共享降尺度模块将订正后的数据分辨率逐次提高.该模型的判别器可鉴别偏差订正及降尺度结果的质量,以此为标准进行对抗训练。同时,在对抗损失函数中含有物理引导的动力学惩罚项以提高模型的性能.本研究基于分辨率为1°的GFDL SPEAR模式的SST预测结果,选择遥感系统(Remote Sensing System)的观测资料作为真值,面向月尺度ENSO与IOD事件以及天尺度海洋热浪事件开展了验证试验:模型在将分辨率提高到0.0625°×0.0625°的同时将预测误差减少约90.3%,突破了观测数据分辨率的限制,且与观测结果的结构相似性高达96.46%.

Predictor Selection for CNN-based Statistical Downscaling of Monthly Precipitation

Convolutional neural networks(CNNs) have been widely studied and found to obtain favorable results in statistical downscaling to derive high-resolution climate variables from large-scale coarse general circulation models(GCMs).However, there is a lack of research exploring the predictor selection for CNN modeling. This paper presents an effective and efficient greedy elimination algorithm to address this problem. The algorithm has three main steps: predictor importance attribution, predictor removal, and CNN retraining, which are performed sequentially and iteratively. The importance of individual predictors is measured by a gradient-based importance metric computed by a CNN backpropagation technique, which was initially proposed for CNN interpretation. The algorithm is tested on the CNN-based statistical downscaling of monthly precipitation with 20 candidate predictors and compared with a correlation analysisbased approach. Linear models are implemented as benchmarks. The experiments illustrate that the predictor selection solution can reduce the number of input predictors by more than half, improve the accuracy of both linear and CNN models,and outperform the correlation analysis method. Although the RMSE(root-mean-square error) is reduced by only 0.8%,only 9 out of 20 predictors are used to build the CNN, and the FLOPs(Floating Point Operations) decrease by 20.4%. The results imply that the algorithm can find subset predictors that correlate more to the monthly precipitation of the target area and seasons in a nonlinear way. It is worth mentioning that the algorithm is compatible with other CNN models with stacked variables as input and has the potential for nonlinear correlation predictor selection.

Statistical Downscaling Retrieval of Land Surface Temperature in an Area with Complex Landforms in the Eastern Qinling Mountains of China Based on Sentinel-2/3 Satellite Data

The study of land surface temperature(LST)is of great significance for ecosystem monitoring and ecological environmental protection in the Qinling Mountains of China.In view of the contradicting spatial and temporal resolutions in extracting LST from satellite remote sensing(RS)data,the areas with complex landforms of the Eastern Qinling Mountains were selected as the research targets to establish the correlation between the normalized difference vegetation index(NDVI)and LST.Detailed information on the surface features and temporal changes in the land surface was provided by Sentinel-2 and Sentinel-3,respectively.Based on the statistically downscaling method,the spatial scale could be decreased from 1000 m to 10 m,and LST with a Sentinel-3 temporal resolution and a 10 m spatial resolution could be retrieved.Comparing the 1 km resolution Sentinel-3 LST with the downscaling results,the 10 m LST downscaling data could accurately reflect the spatial distribution of the thermal characteristics of the original LST image.Moreover,the surface temperature data with a 10 m high spatial resolution had clear texture and obvious geomorphic features that could depict the detailed information of the ground features.The results showed that the average error was 5 K on April 16,2019 and 2.6 K on July 15,2019.The smaller error values indicated the higher vegetation coverage of summer downscaling result with the highest level on July 15.

Evaluating the effects of topographical factors on the precipitation simulated by kilometer-scale versus quarter-degree dynamical downscaling models in eastern China

与传统的1/4度(≈25-30 km)动力降尺度模拟相比,公里尺度模拟的降水空间分布与观测结果更为接近.为了研究这一差异原因,本研究以华东地区为例,探究了地形因子在观测和模拟的降水中的作用.为了更好地体现地形因子对降水分布非均匀性的影响,以及不同地形因子作用的尺度差异,本研究采用多尺度地理加权回归模型,对五个主要地形因子与公里尺度和1/4度分辨率模拟的降水的关系进行了评估.基于观测数据的研究结果显示地形起伏度,地形高程和离海岸线距离对华东地区降水分布的非均匀性都有重要影响,其中地形起伏度在研究区大部分站点降水分布中起主导作用;公里尺度模拟结果基本反映了地形起伏度的主导作用;而1/4度模拟结果表现出降水对地形高程的过度依赖.本研究揭示了公里尺度地形分布对中国东部降水的非均匀分布的关键作用,研究结果可以为改进降水模拟提供新的思路.

基于GRACE/GRACE-FO数据降尺度方法反演库尔勒东区地下水储量变化

GRACE与GRACE-FO重力卫星为全球中大尺度地下水反演监测提供了新的手段,但难以提供小尺度上较高空间分辨率的地下水储量变化(GWSA)信息。本文针对新疆库尔勒东部缺资料区,采用动力降尺度方法提高GRACE/GRACE-FO反演GWSA数据的空间分辨率,分析GWSA的时空分布规律。首先,基于数据融合方法构建了库尔勒东区GWSA数值模型;然后,利用优化后的模型将GWSA反演数据的分辨率从1°降尺度至0.25°和0.05°,将反演的GWSA与水井监测的地下水位(GWL)数据进行对比验证;最后,利用0.05°GWSA数据分析研究区GWSA态势。结果表明:①与降尺度前1°GWSA数据相比,降尺度转换后的高分辨率GWSA数据在空间上更加平滑、展示了更加丰富的细节,且提高了与水井GWL监测数据的相关性,改进了反演精度和可靠性;②小尺度上,降尺度后的GWSA数据能够反映水源地地下水的季节性、年际和长期开采下的亏损等动态特征;③研究区GWSA呈现出时空分布差异性,2005—2020年区内地下水储量变化率为-1~1 mm·a-1,总体上呈南增、北减态势,南、北部山区的变幅大于中部相对平坦区域。

基于GWR的陕西省夏季GPM降水融合降尺度研究

以陕西省为研究目标,基于GWR方法,使用NDVI对GPM降水进行融合降尺度分析,结合87个地面雨量计数据,分析融合前后降水在空间上的差异和原因;同时,考虑陕西省特殊地形,将陕西分为陕北、关中和陕南三部分,利用雨量计数据分析这3个区域的降尺度效果.研究发现:1)GPM和融合降尺度数据均能反映降水在空间上的分布特征,融合降尺度数据更能呈现降水细节;2)参数计算显示,融合降尺度数据与站点数据更接近;3)当具体到某一个较小的范围时,GPM和融合降尺度数据精度仍有待提高;4)GPM和融合降尺度数据能反映陕北、关中和陕南不同的降水特征,且关中和陕南观测质量优于陕北;5)GPM和融合降尺度数据对高降水量的记录更为准确,对小降水量的记录能力有待提高.

基于聚类的HPO-BILSTM光伏功率短期预测

考虑到光伏发电功率在不同天气类型下的波动性和不确定性,对此提出一种基于模糊C均值聚类算法(FCM)和猎食者优化算法(HPO)优化双向长短期记忆网络(BILSTM)的光伏发电短期功率预测模型。首先对光伏发电数据进行处理和分析,再进行主成分分析(PCA)降维和FCM聚类算法将数据按天气类型分为阴、晴、雨;最后通过HPO筛选得出BILSTM神经网络的最佳超参数,避免因超参数设置不佳对实验带来的影响,进一步提高实验的准确性和模型的泛化能力。最后通过预测和对比实验进行分析,验证所提方法的优越性。

结合光谱降维的IPSO-SVR水体总磷浓度预测模型

[目的]选择最优模型对水体中总磷浓度进行预测,为准确、实时、高效检测水资源状况提供支持。[方法]以2021年在长江中下游武汉—安徽地区采集的水质样本作为研究对象,首先,对采集到的长江光谱数据进行最大最小归一化和均值中心化两种预处理操作以便统一数据的范围和均值点,并使用核主成分分析(KPCA)技术对预处理后的光谱数据进行降维操作。选取方差解释率为99.6%下的6个特征向量进行后续预测模型的训练,接着在原有粒子群算法的基础上引入自适应惯性权重更新公式和遗传—模拟退火变异思想,提高算法的寻优能力。使用改进的粒子群优化算法对支持向量回归模型中的超参数组合进行寻优,对支持向量回归模型使用输出的结果进行预测模型的训练,最后使用测试集数据进行总磷浓度的预测。[结果]提出了一种结合光谱降维的改进粒子群优化算法(IPSO)结合支持向量回归(SVR)的水体总磷含量预测模型。通过和当前预测性能较好的几种机器学习模型进行精度的比较发现,该试验模型对长江水体总磷浓度进行预测时决定系数(R^(2))为0.973920,均方根差(RMSE)为0.003012,平均绝对误差(MAE)为0.002105。[结论]使用光谱数据结合降维技术、粒子群优化算法和机器学习模型的算法融合模型检测水体总磷浓度可行性强,精确度高,且拟合效果良好。

基于红光波段冠层SIF降尺度的小麦条锈病遥感监测

为减弱冠层几何结构等因素对传感器探测到的冠层日光诱导叶绿素荧光(Solar-induced chlorophyll fluorescence, SIF)的影响,探讨了条锈病胁迫下红光波段荧光(Red SIF,RSIF)的响应特性,并以RSIF为自变量构建了小麦条锈病遥感监测的线性回归(Simple linear regression, SLR)及非线性回归(Non-linear regression, NLR)模型。结果表明:叶片尺度RSIF在小麦条锈病遥感监测中具有较大优势,其与小麦条锈病病情严重度(Severity level, SL)间相关系数较远红光波段SIF(Far-red SIF,FRSIF)提高13.2%,以叶片尺度RSIF为自变量构建的SLR及NLR模型预测D_(SL)与实测D_(SL)之间R^(2)较FRSIF分别增加9.8%、38.9%,RMSE分别降低23.1%、36.4%。此外,降尺度处理能够提高RSIF监测小麦条锈病的精度,叶片尺度RSIF与D_(SL)之间R^(2)较冠层尺度增加126.3%,以叶片尺度RSIF为自变量构建的SLR和NLR模型预测D_(SL)与实测D_(SL)间R^(2)较冠层尺度分别提高114.3%和233.3%,RMSE分别降低16.7%、15.4%。本文提出方法可提高小麦条锈病遥感监测精度,同时对其它胁迫的遥感监测具有一定的参考价值。

降雨监测与预报技术在防洪减灾中的应用进展

洪水灾害突发性强,成灾速度快,对人民生命和财产安全造成较大的威胁。降雨作为洪水灾害致灾因子,数据的精确度对防洪减灾具有重要意义。以降雨监测与预报技术为切入点,对雨量站点观测、天气雷达降雨估计及预报、降雨数值预报、卫星遥感反演的现状进行了总结,通过分析时空降尺度方法及多源数据融合技术在降雨监测与预报中的应用,揭示了其在提升降雨数据“量”与“型”准确度方面的效果。研究表明:降雨监测与预报技术在当前取得了显著进展,但在山丘区和城市环境空间的复杂地形方面仍面临分辨率受到限制及精确性、时效性不足的问题。多源数据融合能提高降雨数据精度、时空覆盖能力和预测准确性,优化算法模型、融合“空-天-地”多源数据形成高分辨率预报是未来的研究方向。

基于多种统计降尺度方法的未来降水预估研究——以青藏高原为例

虽然第六次耦合模式比较计划(Coupled Model Intercomparison Project 6,CMIP6)能很好地预测大尺度气候要素,但是其在预测流域尺度方面的效果与实测数据仍有差别,尤其是在青藏高原这种高海拔、地形复杂地区,气候模式所产生的误差更大。基于最新一代高分辨率CMIP6模式历史情景和SSP126、SSP245、SSP370、SSP585等多种未来气候排放情景,研究使用包括偏差校正、KNN、SDSM等多种统计降尺度方法进行降尺度分析,并对各自的预测性能进行了评估,在此基础上使用性能最佳的统计降尺度方式预估青藏高原地区的未来降水,对最终得到的预估降水的时空演变特征进行了详细的分析,并与青藏高原的历史降水情况进行了对比。结果表明,3种统计降尺度在青藏高原的适用性差异较大,线性回归降尺度方法的性能最佳,其次为偏差校正方法,最差为KNN类比方法。从未来降水预估情况分析,青藏高原未来80 a平均降水、降水极值等总体呈上升趋势但上升幅度较小,且空间分布情况变化不大。研究结果可为青藏高原水资源评价及规划与管理提供科学依据。

Convection-Permitting Simulations of Current and Future Climates over the Tibetan Plateau

The Tibetan Plateau(TP)region,also known as the“Asian water tower”,provides a vital water resource for downstream regions.Previous studies of water cycle changes over the TP have been conducted with climate models of coarse resolution in which deep convection must be parameterized.In this study,we present results from a first set of highresolution climate change simulations that permit convection at approximately 3.3-km grid spacing,with a focus on the TP,using the Icosahedral Nonhydrostatic Weather and Climate Model(ICON).Two 12-year simulations were performed,consisting of a retrospective simulation(2008–20)with initial and boundary conditions from ERA5 reanalysis and a pseudoglobal warming projection driven by modified reanalysis-derived initial and boundary conditions by adding the monthly CMIP6 ensemble-mean climate change under the SSP5-8.5 scenario.The retrospective simulation shows overall good performance in capturing the seasonal precipitation and surface air temperature.Over the central and eastern TP,the average biases in precipitation(temperature)are less than−0.34 mm d−1(−1.1℃)throughout the year.The simulated biases over the TP are height-dependent.Cold(wet)biases are found in summer(winter)above 5500 m.The future climate simulation suggests that the TP will be wetter and warmer under the SSP5-8.5 scenario.The general features of projected changes in ICON are comparable to the CMIP6 ensemble projection,but the added value from kilometer-scale modeling is evident in both precipitation and temperature projections over complex topographic regions.These ICON-downscaled climate change simulations provide a high-resolution dataset to the community for the study of regional climate changes and impacts over the TP.

基于随机森林模型的GRACE数据3种空间降尺度对比

陆地水储量是赋存在陆地上各种形式水的综合体现,研究其时空变化对认识区域水循环过程和水资源调控等具有重要意义。然而现有陆地水储量变化数据实际分辨率较低,限制了其在中小流域或地区中的应用。针对这一问题,本文基于GRACE重力卫星和其后续卫星GRACE-FO反演的陆地水储量变化数据,首先采用随机森林模型,分别基于格点、区域(流域)和区域(全国)3种空间降尺度思路将GRACE数据降尺度至0.25°×0.25°,后结合GLDAS模型数据,基于水量平衡原理计算得到地下水储量变化数据,最后基于降尺度模型模拟效果和实测地下水位数据评估3种降尺度思路在全国的适用性。结果表明:随机森林模型能够较好地模拟驱动数据(降水、气温、植被条件指数和土壤水储量)与GRACE数据的统计关系,验证期格点降尺度思路的平均相关系数总体在0.6左右,区域降尺度思路的平均纳什效率系数、相关系数和均方根误差分别>0.5、>0.75和<6.6 cm,3种空间降尺度思路的模拟精度均满足基本要求;2003—2021年间,GRACE数据、格点降尺度、区域降尺度(流域)和区域降尺度(全国)得到的我国陆地水储量亏缺量分别约为119.5×10^(8)、62.4×10^(8)、121.1×10^(8)和121.8×10^(8)m^(3)/a,地下水储量亏缺量分别约为230.0×10^(8)、171.8×10^(8)、235.6×10^(8)和236.4×10^(8)m^(3)/a,受制于样本数较少等原因,格点降尺度结果精度较差;两种区域降尺度思路得到的水储量变化速率均和原始GRACE数据基本一致,模拟结果均优于格点降尺度,且细化到流域的区域降尺度对地下水储量变化验证精度有一定的改善。区域降尺度具有适用性强、模拟精度高、计算效率高的优势,研究结果可为流域水资源可持续利用以及水资源规划等提供精细化的水储量变化数据。

融合精细化气象因素与物理约束的深度学习模型在短期风电功率预测中的应用

现有基于深度学习方法的风电功率预测是一种以气象数据为输入的间接预测,其预测精度依赖于气象预报的准确率,然而现有气象预报资料普遍存在分辨率低,预报模式不稳定的问题。同时,深度学习模型完全依赖数据驱动,缺乏物理规律的指导,预测精度难以进一步提升。因此,提出一种精细化气象因素与物理深度学习相结合的方法。首先,通过降尺度与多模式集成技术,对数值天气预报数据进行处理,改善气象预报产品的低分辨率和准确率问题;其次,基于风电场尾流效应和功率曲线两种物理模型,一方面将物理模型嵌入神经网络损失函数作为正则化项,引入物理约束指导学习过程,以构建物理深度学习网络;另一方面,利用物理模型产生预训练样本,解决观测数据不足的情况,构建预训练模型,为后续有监督学习任务提供支持。最后,通过对某市近海风电场的实际数据进行仿真分析,验证了所提出方法的有效性和优越性。

喀斯特山区TRMM降水数据降尺度研究

获取高精度的卫星降水数据,为喀斯特区域旱涝灾害评估、水文预报等各研究领域提供数据基础。以热带降雨卫星(tropical rainfall measuring mission, TRMM)为数据源,采用多元线性回归法(ordinary least square, OLS)和地理加权回归法(geographically weighted regression, GWR),综合考虑高程、坡度、坡向、经纬度和归一化植被指数(normalized difference vegetation index, NDVI)等6个因子构建OLS和GWR降尺度模型进行年降尺度研究,并比较OLS和GWR两种降尺度模型在喀斯特山区的适用性。结果表明:1)TRMM数据与站点观测数据之间精度较好;2)降尺度后数据空间分辨率提升到1 km, GWR降尺度年降水量在多数年份比原始TRMM数据更接近实测值,高估现象得到改善;与OLS降尺度数据相比,其三项指标表现更好;3)单站点中,OLS降尺度数据在高程和NDVI突变区域易出现假性更优相关性。综合多指标评价,GWR降尺度数据在喀斯特山区总体精度更高。后续可通过划分植被区、岩溶区、增加环境因子、校正等使降水更贴合实测值。

遥感降水降尺度高精度校正及不确定性分析方法

为消除降水场同质部分影响,提升统计降水降尺度结果精度,提出了基于贝叶斯高精度曲面建模(Bayes-HASM)算法的遥感降水降尺度高精度校正方法。该方法通过引入模拟精度更高的高精度曲面建模方法,并结合贝叶斯优化算法,实现了模型参数自动优化选择和高精度降尺度校正,解决了现有降尺度残差校正方法存在的误差和多尺度问题。结果表明:贝叶斯优化使高精度曲面建模的不确定性显著减少;经过Bayes-HASM残差校正后,降尺度结果的散点分布更加接近1∶1线,年、季、月和旬尺度的精度指标均得到了显著的改善,CC和IA指标提高至0.9左右,RMSE下降明显,RB也显著改善。本方法能显著降低模型的不确定性并起到消除降水场同质部分影响的作用,有效提升降水降尺度结果精度。