Coupling statistical and dynamical methods for spatial downscaling of precipitation

The resolution of General Circulation Models (GCMs) is too coarse for climate change impact studies at the catchment or site-specific scales. To overcome this problem, both dynamical and statistical downscaling methods have been developed. Each downscaling method has its advantages and drawbacks, which have been described in great detail in the literature. This paper evaluates the improvement in statistical downscaling (SD) predictive power when using predictors from a Regional Climate Model (RCM) over a GCM for downscaling site-specific precipitation. Our approach uses mixed downscaling, combining both dynamic and statistical methods. Precipitation, a critical element of hydrology studies that is also much more difficult to downscale than temperature, is the only variable evaluated in this study. The SD method selected here uses a stepwise linear regression approach for precipitation quantity and occurrence (similar to the well-known Statistical Downscaling Model (SDSM) and called SDSM-like herein). In addition, a discriminant analysis (DA) was tested to generate precipitation occurrence, and a weather typing approach was used to derive statistical relationships based on weather types, and not only on a seasonal basis as is usually done. The existing data record was separated into a calibration and validation periods. To compare the relative efficiency of the SD approaches, relationships were derived at the same sites using the same predictors at a 300km scale (the National Center for Environmental Prediction (NCEP) reanalysis) and at a 45km scale with data from the limited-area Canadian Regional Climate Model (CRCM) driven by NCEP data at its boundaries. Predictably, using CRCM variables as predictors rather than NCEP data resulted in a much-improved explained variance for precipitation, although it was always less than 50?% overall. For precipitation occurrence, the SDSM-like model slightly overestimated the frequencies of wet and dry periods, while these were well-replicated by the DA-based model. Both the SDSM-like and DA-based models reproduced the percentage of wet days, but the wet and dry statuses for each day were poorly downscaled by both approaches. Overall, precipitation occurrence downscaled by the DA-based model was much better than that predicted by the SDSM-like model. Despite the added complexity, the weather typing approach was not better at downscaling precipitation than approaches without classification. Overall, despite significant improvements in precipitation occurrence prediction by the DA scheme, and even going to finer scales predictors, the SD approach tested here still explained less than 50?% of the total precipitation variance. While going to even smaller scale predictors (10–15?km) might improve results even more, such smaller scales would basically transform the direct outputs of climate models into impact models, thus negating the need for statistical downscaling approaches.     

Temporal statistical downscaling of precipitation and temperature forecasts using a stochastic weather generator

Statistical downscaling is based on the fact that the large-scale climatic state and regional/local physiographic features control the regional climate. In the present paper, a stochastic weather generator is applied to seasonal precipitation and temperature forecasts produced by the International Research Institute for Climate and Society(IRI). In conjunction with the GLM(generalized linear modeling) weather generator, a resampling scheme is used to translate the uncertainty in the seasonal forecasts(the IRI format only specifies probabilities for three categories: below normal, near normal, and above normal) into the corresponding uncertainty for the daily weather statistics. The method is able to generate potentially useful shifts in the probability distributions of seasonally aggregated precipitation and minimum and maximum temperature, as well as more meaningful daily weather statistics for crop yields, such as the number of dry days and the amount of precipitation on wet days. The approach is extended to the case of climate change scenarios, treating a hypothetical return to a previously observed drier regime in the Pampas.     

Evaluation of a WRF dynamical downscaling simulation over California

This paper presents results from a 40 year Weather Research and Forecasting (WRF) based dynamical downscaling experiment performed at 12 km horizontal grid spacing, centered on the state of California, and forced by a 1° × 1.25° finite-volume current-climate Community Climate System Model ver. 3 (CCSM3) simulation. In-depth comparisons between modeled and observed regional-average precipitation, 2 m temperature, and snowpack are performed. The regional model reproduces the spatial distribution of precipitation quite well, but substantially overestimates rainfall along windward slopes. This is due to strong overprediction of precipitation intensity; precipitation frequency is actually underpredicted by the model. Moisture fluxes impinging on the coast seem to be well-represented over California, implying that precipitation bias is caused by processes internal to WRF. Positive-definite moisture advection and use of the Grell cumulus parameterization result in some decrease in precipitation bias, but other sources are needed to explain the full bias magnitude. Surface temperature is well simulated in all seasons except summer, when overly-dry soil moisture results in a several degree warm bias in both CCSM3 and WRF. Additionally, coastal temperatures appear to be too warm due to a coastal sea surface temperature bias inherited from CCSM3. Modeled snowfall/snowmelt agrees quite well with observations, but snow water equivalent is found to be much too low due to monthly reinitialization of all regional model fields from CCSM3 values.     

Evaluation of statistical downscaling in short range precipitation forecasting

The objective of this study is to compare several statistical downscaling methods for the development of an operational short-term forecast of precipitation in the area of Bilbao (Spain). The ability of statistical downscaling methods nested inside numerical simulations run by both coarse and regional model simulations is tested with several selections of predictors and domain sizes. The selection of predictors is performed both in terms of sound physical mechanisms and also by means of “blind” criteria, such as “give the statistical downscaling methods all the information they can process”.Results show that the use of statistical downscaling methods improves the ability of the mesoscale and coarse resolution models to provide quantitative precipitation forecasts. The selection of predictors in terms of sound physical principles does not necessarily improve the ability of the statistical downscaling method to select the most relevant inputs to feed the precipitation forecasting model, due to the fact that the numerical models do not always fulfil conservation laws or because precipitation events do not reflect simple phenomenological laws. Coarse resolution models are able to provide information usable in combination with a statistical downscaling method to achieve a quantitative precipitation forecast skill comparable to that obtained by other systems currently in use.     

白洋淀流域未来日最高最低气温变化的统计降尺度分析

大气环流模型（GCMs）预测的气候变化情景空间分辨率低,不能满足气候变化对水资源影响进行评估的需要.利用统计降尺度模型可以解决GCMs预测的气候变化情景空间分辨率低的缺陷.在白洋淀流域应用统计降尺度模型（SDSM）,选取日平均气温作为预报量,根据NCEP再分析数据与站点实测数据序列的相关关系选择合适的预报因子,建立大气环流因子与各站点日最高气温和最低气温之间的统计关系.将数据序列分为1961-1975年和1976-1990年两个时段,对SDSM进行率定和验证.最后将HadCM3输出的未来情景降尺度到站点尺度,模拟白洋淀流域未来时期三个时段2020s（2010-2039年）、2050s（2040-2069年）和2080s（2070-2099年）的日最高气温和最低气温时间序列.结果表明：SDSM在白洋淀流域的模拟效果较好.白洋淀流域日最高气温和最低气温在A2和B2两种情景下均呈现上升趋势,且A2情景下的增幅高于B2情景,山区的增幅高于平原,日最高气温的增幅大于日最低气温.     

统计降尺度方法在天津小时降水和气温精细化预报中的应用

基于ECWMF模式预报数据对2018年3—11月降水和2 m温度进行统计降尺度,利用先频率匹配法、再阈值法对插值后的降水订正,利用Kalman滤波型的递减平均统计降尺度法对插值后的温度订正,最终获得逐小时降水量和温度的预报。结果表明：(1)对于晴雨预报准确率,绝大多数预报时效频率匹配法和阈值法均对其有明显提高,前者最大改进幅度可达20%以上。对于相对误差,阈值法对空报现象有较显著改进。对于1 h降雨量大于等于20 mm的短时强降水,频率匹配法订正后的TS评分有明显提高。对2018年“安比”台风事件,除具有以上改进效果外,频率匹配法提高了降水主体形态和量级的预报水平,阈值法对空报站订正正确。(2)对于温度的ECWMF模式预报检验,几乎在任何预报时效内都是3月的绝对误差最大。通过Kalman滤波型的递减平均统计降尺度法后,各月的绝对误差都有不同程度减小。总体上,订正后的绝对误差曲线仍具有订正前的周期性波动,波峰、波谷位置也与订正前基本一致,且绝对误差越大,订正幅度越大。个例分析也表明订正后保留了温度预报空间分布的准确性,且绝对误差有明显下降。     

Probabilistic estimates of recent changes in temperature: a multi-scale attribution analysis

The role of anthropogenic forcings in temperature changes during recent decades is investigated over a range of spatial scales. Changes in the annual mean surface temperature and also in the warmest night of the year, which has implications for human health, are considered. Distributions of regional trends with and without the effect of human activity are produced, using constraints from a global optimal detection analysis. Anthropogenic forcings are estimated to have more than doubled the likelihood of mean warming in all regions considered except central North America, where results are more model dependent. The likelihood of warming of the warmest night has also increased, but the estimated change is more uncertain. Inferences on sub-continental scales are indicative rather than definitive because of the absence of locally important forcings and processes in model simulations, as well as model biases. As model inconsistencies may impact regional analyses, a multi-model approach is essential.     

High-resolution precipitation and temperature downscaling for glacier models

The spatial resolution gap between global or regional climate models and the requirements for local impact studies motivates the need for climate downscaling. For impact studies that involve glacier modelling, the sparsity or complete absence of climate monitoring activities within the regions of interest presents a substantial additional challenge. Downscaling methods for this application must be independent of climate observations and cannot rely on tuning to station data. We present new, computationally-efficient methods for downscaling precipitation and temperature to the high spatial resolutions required to force mountain glacier models. Our precipitation downscaling is based on an existing linear theory for orographic precipitation, which we modify for large study regions by including moist air tracking. Temperature is downscaled using an interpolation scheme that reconstructs the vertical temperature structure to estimate surface temperatures from upper air data. Both methods are able to produce output on km to sub-km spatial resolution, yet do not require tuning to station measurements. By comparing our downscaled precipitation (1 km resolution) and temperature (200 m resolution) fields to station measurements in southern British Columbia, we evaluate their performance regionally and through the annual cycle. Precipitation is improved by as much as 30% (median relative error) over the input reanalysis data and temperature is reconstructed with a mean bias of 0.5°C at locations with high vertical relief. Both methods perform best in mountainous terrain, where glaciers tend to be concentrated.     

中国地面气温统计降尺度预报方法研究

利用中国752个基本、基准地面气象观测站2000—2010年地面温度日值数据,采用具有自适应特征的Kalman滤波类型的递减平均统计降尺度技术,对中国地面温度进行精细化预报研究。分析该方案的降尺度效果,并与常用插值降尺度方法进行比较。结果表明:1)递减平均统计降尺度技术相比插值方法有较大的提高,显著减小东西部预报效果差异,1~3 d预报的均方根误差减小了1.4℃;2)该方案1~3 d预报的均方根误差为1.5℃,预报误差从东南地区(均方根误差为1.4℃)向西北地区(均方根误差为1.8℃)逐渐增大,并且预报效果夏季优于冬季。因此,递减平均统计降尺度技术对中国地面温度进行精细化预报是可行的。     

On regional dynamical downscaling for the assessment and projection of temperature and precipitation extremes across Tasmania, Australia

The ability of an ensemble of six GCMs, downscaled to a 0.1° lat/lon grid using the Conformal Cubic Atmospheric Model over Tasmania, Australia, to simulate observed extreme temperature and precipitation climatologies and statewide trends is assessed for 1961–2009 using a suite of extreme indices. The downscaled simulations have high skill in reproducing extreme temperatures, with the majority of models reproducing the statewide averaged sign and magnitude of recent observed trends of increasing warm days and warm nights and decreasing frost days. The warm spell duration index is however underestimated, while variance is generally overrepresented in the extreme temperature range across most regions. The simulations show a lower level of skill in modelling the amplitude of the extreme precipitation indices such as very wet days, but simulate the observed spatial patterns and variability. In general, simulations of dry extreme precipitation indices are underestimated in dryer areas and wet extremes indices are underestimated in wetter areas. Using two SRES emissions scenarios, the simulations indicate a significant increase in warm nights compared to a slightly more moderate increase in warm days, and an increase in maximum 1- and 5-day precipitation intensities interspersed with longer consecutive dry spells across Tasmania during the twenty-first century.     

A comparison study of three statistical downscaling methods and their model-averaging ensemble for precipitation downscaling in China

This study evaluated the performance of three frequently applied statistical downscaling tools including SDSM, SVM, and LARS-WG, and their model-averaging ensembles under diverse moisture conditions with respect to the capability of reproducing the extremes as well as mean behaviors of precipitation. Daily observed precipitation and NCEP reanalysis data of 30 stations across China were collected for the period 1961–2000, and model parameters were calibrated for each season at individual site with 1961–1990 as the calibration period and 1991–2000 as the validation period. A flexible framework of multi-criteria model averaging was established in which model weights were optimized by the shuffled complex evolution algorithm. Model performance was compared for the optimal objective and nine more specific metrics. Results indicate that different downscaling methods can gain diverse usefulness and weakness in simulating various precipitation characteristics under different circumstances. SDSM showed more adaptability by acquiring better overall performance at a majority of the stations while LARS-WG revealed better accuracy in modeling most of the single metrics, especially extreme indices. SVM provided more usefulness under drier conditions, but it had less skill in capturing temporal patterns. Optimized model averaging, aiming at certain objective functions, can achieve a promising ensemble with increasing model complexity and computational cost. However, the variation of different methods' performances highlighted the tradeoff among different criteria, which compromised the ensemble forecast in terms of single metrics. As the superiority over single models cannot be guaranteed, model averaging technique should be used cautiously in precipitation downscaling.     

基于转移累计概率分布统计降尺度方法的未来降水预估研究:以湖南省为例

基于最新一代CMIP5（Coupled Model Intercomparison Project Phase 5）模式历史情景和未来RCP4.5情景下的模式逐日降水数据,使用转移累计概率分布（CDF-t）统计降尺度方法,从空间变化和时间变率两个方面评估该降尺度方法对湖南日降水量模拟能力的改善效果,并在此基础上对未来降水量变化进行预估。结果表明, CMIP5气候模式由于分辨率较低,无法细致反映湖南地形变化和大气环流影响导致的区域降水变化特征。经过CDF-t统计降尺度处理之后,模式对湖南降水的时、空分布模拟与实况更为接近,绝大部分模式对降水空间结构的模拟能力都有显著提高。基于CDF-t统计降尺度的多模式集合预估结果表明,21世纪湖南省日降水量呈弱的增多趋势（0.95%/（10 a））。21世纪初、中和末期相对于1986—2005年的气候平均态,湖南省日降水量分别增加了4.6%、5%和5.2%。3个时期湖南省日平均降水变化的空间分布存在较强的一致性,皆表现为湖南西北、东北和东南3个地区降水增幅最为显著,且随着辐射强迫的增大,3个地区降水增幅也呈递增趋势。需要指出的是,预估结果在模式之间存在一定差异,并且这种差异随着辐射强迫的增大而增大。      

An estimate of future climate change for western France using a statistical downscaling technique

A statistical downscaling procedure based on an analogue technique is used to determine projections for future climate change in western France. Three ocean and atmosphere coupled models are used as the starting point of the regionalization technique. Models' climatology and day to day variability are found to reproduce the broad main characteristics seen in the reanalyses. The response of the coupled models to a similar CO₂ increase scenario exhibit marked differences for mean sea-level pressure; precipitable water and temperature show arguably less spread. Using the reanalysis fields as predictors, the statistical model parameters are set for daily extreme temperatures and rain occurrences for seventeen stations in western France. The technique shows some amount of skill for all three predictands and across all seasons but failed to give reliable estimates of rainfall amounts. The quality of both local observations and large-scale predictors has an impact on the statistical model skill. The technique is partially able to reproduce the observed climatic trends and inter annual variability, showing the sensitivity of the analogue approach to changed climatic conditions albeit an incomplete explained variance by the statistical technique. The model is applied to the coupled model control simulations and the gain compared with direct model grid-average outputs is shown to be substantial at station level. The method is then applied to altered climate conditions; the impact of large-scale model uncertain responses and model sensitivities are quantified using the three coupled models. The warming in the downscaled projections are reduced compared with their global model counterparts.     

Drought episodes over Greece as simulated by dynamical and statistical downscaling approaches

Drought over the Greek region is characterized by a strong seasonal cycle and large spatial variability. Dry spells longer than 10 consecutive days mainly characterize the duration and the intensity of Greek drought. Moreover, an increasing trend of the frequency of drought episodes has been observed, especially during the last 20 years of the 20th century. Moreover, the most recent regional circulation models (RCMs) present discrepancies compared to observed precipitation, while they are able to reproduce the main patterns of atmospheric circulation. In this study, both a statistical and a dynamical downscaling approach are used to quantify drought episodes over Greece by simulating the Standardized Precipitation Index (SPI) for different time steps (3, 6, and 12 months). A statistical downscaling technique based on artificial neural network is employed for the estimation of SPI over Greece, while this drought index is also estimated using the RCM precipitation for the time period of 1961–1990. Overall, it was found that the drought characteristics (intensity, duration, and spatial extent) were well reproduced by the regional climate models for long term drought indices (SPI12) while ANN simulations are better for the short-term drought indices (SPI3).

     

Precipitation,temperature and wind in Norway: dynamical downscaling of ERA40

A novel downscaling approach of the ERA40 (ECMWF 40-years reanalysis) data set has been taken and results for comparison with observations in Norway are shown. The method applies a nudging technique in a stretched global model, focused in the Norwegian Sea (67°N, 5°W). The effective resolution is three times the one of the ERA40, equivalent to about 30 km grid spacing in the area of focus. Longer waves (<T42) in the downscaled solution are nudged towards the ERA40 solution, and thus the large-scale circulation is similar in the two data sets. The shorter waves are free to evolve, and produce high intensities of winds and precipitation. The comparison to observations incorporate numerous station data points of (1) precipitation (#357), (2) temperature (#98) and (3) wind (#10), and for the period 1961–1990, the downscaled data set shows large improvements over ERA40. The daily precipitation shows considerable reduction in bias (from 50 to 11%), and twofold reduction at the 99.9 percentile (from −59 to −29%). The daily temperature showed a bias reduction of about a degree in most areas, and relative large RMSE reduction (from 7.5 to 5.0°C except winter). The wind comparison showed a slight improvement in bias, and significant improvements in RMSE.     

Simulation of extreme precipitation indices in the Yangtze River basin by using statistical downscaling method (SDSM)

In this study, the applicability of the statistical downscaling model (SDSM) in modeling five extreme precipitation indices including R10 (no. of days with precipitation ≥10?mm?day^?1), SDI (simple daily intensity), CDD (maximum number of consecutive dry days), R1d (maximum 1-day precipitation total) and R5d (maximum 5-day precipitation total) in the Yangtze River basin, China was investigated. The investigation mainly includes the calibration and validation of SDSM model on downscaling daily precipitation, the validation of modeling extreme precipitation indices using independent period of the NCEP reanalysis data, and the projection of future regional scenarios of extreme precipitation indices. The results showed that: (1) there existed good relationship between the observed and simulated extreme precipitation indices during validation period of 1991–2000, the amount and the change pattern of extreme precipitation indices could be reasonably simulated by SDSM. (2) Under both scenarios A2 and B2, during the projection period of 2010–2099, the changes of annual mean extreme precipitation indices in the Yangtze River basin would be not obvious in 2020s; while slightly increase in the 2050s; and significant increase in the 2080s as compared to the mean values of the base period. The summer might be the more distinct season with more projected increase of each extreme precipitation indices than in other seasons. And (3) there would be distinctive spatial distribution differences for the change of annual mean extreme precipitation indices in the river basin, but the most of Yangtze River basin would be dominated by the increasing trend.     

Improving long-lead seasonal forecasts of precipitation over Southern China based on statistical downscaling using BCC_CSM1.1m

Long-lead precipitation forecasts for 1–4 seasons ahead are usually difficult in dynamical climate models due to the model deficiencies and the limited persistence of initial signals. But, these forecasts could be empirically improved by statistical approaches. In this study, to improve the seasonal precipitation forecast over the southern China (SC), the statistical downscaling (SD) models are built by using the predictors of atmospheric circulation and sea surface temperature (SST) simulated by the Beijing Climate Center Climate System Model version 1.1 m (BCC_CSM1.1 m). The different predictors involved in each SD model is selected based on both its close relationship with the target seasonal precipitation and its reasonable prediction skill in the BCC_CSM1.1 m. Cross and independent validations show the superior performance of the SD models, relative to the BCC_CSM1.1 m. The temporal correlation coefficient of SD models could reach > 0.4, exceeding the 95 % confidence level. The SC precipitation index can be much better forecasted by the SD models than by the BCC_CSM1.1 m in terms of the interannual variability. In addition, the errors of the precipitation forecast in all four seasons are significantly reduced over most of SC in the SD models. For the 2015/2016 strong El Niño event, the SD models outperform the dynamical BCC_CSM1.1 m model on the spatial and regional-average precipitation anomalies, mostly due to the effective SST predictor in the SD models and the weak response of the SC precipitation to El Niño-related SST anomalies in the BCC_CSM1.1 m.