Bias Correction Spatial Disaggregation Research Articles

Projection of future drought characteristics in the Great South of Madagascar using CMIP6 and bias-correction spatial disaggregation method

Projection of future drought characteristics in the Great South of Madagascar using CMIP6 and bias-correction spatial disaggregation method

Climate change projections of continental-scale streamflow across the Mississippi River Basin

A large body of scientific research has demonstrated a changing climate, which affects river flow regimes and extreme flood frequencies and magnitudes. The magnitude and frequency of extreme events are of critical importance in the evaluation of river systems to inform flood risk reduction under current and future conditions. The global climate projections from the Coupled Model Intercomparison Project, Phase 5 (CMIP5) datasets were used by the Variable Infiltration Capacity (VIC) land surface model to produce a runoff dataset, implementing a Bias-Correction Spatial Disaggregation (BCSD) approach. The resulting runoff was then used as input to the Routing Application for Parallel computatIon of Discharge (RAPID) river routing model to simulate daily flows within all 1.2 million Mississippi River Basin river reaches for years 1950 through 2099. This research effort analyzed the performance of the models for the historical time period, comparing with the observations at 64 gage locations for 16 different climate models. A recurrence interval analysis was performed to determine the 2-, 5-, 10-, 50-, 100-, 500-, and 1000-year events within both the historical and projected time periods, highlighting the relative changes predicted into the future. Anticipated seasonal changes are demonstrated by comparing monthly average streamflows for three different time periods (1951–2005, 2006–2049, and 2050–2099). Results indicate that the hydrologic conditions of the Lower Mississippi River are not stationary. Based on all 16 models considered in this study, the median of the model projections shows an 8% increase in the 100-year return period discharge at Vicksburg, Mississippi, into the future time period, although the full range of 16 models varies widely from − 11 to + 85% change in the 100-year discharge in the future.

How Predictable is Short-Term Drought in the Northeastern United States?

Abstract We investigated the predictability (forecast skill) of short-term droughts using the Palmer drought severity index (PDSI). We incorporated a sophisticated data training (of decadal range) to evaluate the improvement of forecast skill of short-term droughts (3 months). We investigated whether the data training of the synthetic North American Multi-Model Ensemble (NMME) climate has some influence on enhancing short-term drought predictability. The central elements are the merged information among PDSI and NMME with two postprocessing techniques. 1) The bias correction–spatial disaggregation (BC-SD) method improves spatial resolution by using a refined soil information introduced in the available water capacity of the PDSI calculation to assess water deficit that better estimates drought variability. 2) The ensemble model output statistic (EMOS) approach includes systematically trained decadal information of the multimodel ensemble simulations. Skill of drought forecasting improves when using EMOS, but BC-SD does not increase the forecast skill when compared with an analysis using BC (low spatial resolution). This study suggests that predictability forecast of drought (PDSI) can be extended without any change in the core dynamics of the model but instead by using the sophisticated EMOS postprocessing technique. We pointed out that using NMME without any postprocessing is of limited use in the suite of model variations of the NMME, at least for the U.S. Northeast. From our analysis, 1 month is the most extended range we should expect, which is below the range of the seasonal scale presented with EMOS (2 months). Thus, we propose a new design of drought forecasts that explicitly includes the multimodel ensemble signal.

NASA Global Daily Downscaled Projections, CMIP6

We describe the latest version of the NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP-CMIP6). The archive contains downscaled historical and future projections for 1950–2100 based on output from Phase 6 of the Climate Model Intercomparison Project (CMIP6). The downscaled products were produced using a daily variant of the monthly bias correction/spatial disaggregation (BCSD) method and are at 1/4-degree horizontal resolution. Currently, eight variables from five CMIP6 experiments (historical, SSP126, SSP245, SSP370, and SSP585) are provided as procurable from thirty-five global climate models.

Station‐based non‐linear regression downscaling approach: A new monthly precipitation downscaling technique

ABSTRACTStatistical downscaling is an effective way to downscale General Circulation Model (GCM) outputs to a finer temporal and spatial scale. Here, we proposed and trained a new station‐based non‐linear regression downscaling (SNRD) approach, and quantitatively demonstrated its better performance than linear regression model. In addition, we also compared the SNRD method with the Bias Correction Spatial Disaggregation (BCSD) method, which has been widely applied on downscaling precipitation. Results indicated that RMSEs and residuals of the BCSD method are 4.6–31 times as ones of SNRD in each region of China during validation period (2004–2015). Besides, with historical in‐situ observed monthly mean precipitation as the benchmark, SNRD downscaled model modified the over‐estimation (1.25–2 times as observed precipitation) flaw of BCSD model in rainy season (June–August) of China, which could provide assistance on the further researches on drought events in rainy season in future. Further, we also surprisingly found that the 2,500 m was the threshold height for statistical downscaling models. Both BCSD and SNRD downscaled precipitation displayed more steadily and accurately above 2,500 m, since there are less intense human activities in China. Thus, we recommend further studies should fully consider the height element in downscaling models to improve the accuracies.

Deep-Learning-Based Gridded Downscaling of Surface Meteorological Variables in Complex Terrain. Part II: Daily Precipitation

AbstractStatistical downscaling (SD) derives localized information from larger-scale numerical models. Convolutional neural networks (CNNs) have learning and generalization abilities that can enhance the downscaling of gridded data (Part I of this study experimented with 2-m temperature). In this research, we adapt a semantic-segmentation CNN, called UNet, to the downscaling of daily precipitation in western North America, from the low resolution (LR) of 0.25° to the high resolution (HR) of 4-km grid spacings. We select LR precipitation, HR precipitation climatology, and elevation as inputs; train UNet over the subset of the south- and central-western United States using Parameter–Elevation Regressions on Independent Slopes Model (PRISM) data from 2015 to 2018, and test it independently in all available domains from 2018 to 2019. We proposed an improved version of UNet, which we call Nest-UNet, by adding deep-layer aggregation and nested skip connections. Both the original UNet and Nest-UNet show generalization ability across different regions and outperform the SD baseline (bias-correction spatial disaggregation), with lower downscaling error and more accurate fine-grained textures. Nest-UNet also shares the highest amount of information with station observations and PRISM, indicating good ability to reduce the uncertainty of HR downscaling targets.

Comparison of Statistical and Dynamic Downscaling Techniques in Generating High-Resolution Temperatures in China from CMIP5 GCMs

AbstractIn aiming for better access to climate change information and for providing climate service, it is important to obtain reliable high-resolution temperature simulations. Systematic comparisons are still deficient between statistical and dynamic downscaling techniques because of their inherent unavoidable uncertainties. In this paper, 20 global climate models (GCMs) and one regional climate model [Providing Regional Climates to Impact Studies (PRECIS)] are employed to evaluate their capabilities in reproducing average trends of mean temperature (Tm), maximum temperature (Tmax), minimum temperature (Tmin), diurnal temperature range (DTR), and extreme events represented by frost days (FD) and heat-wave days (HD) across China. It is shown generally that bias of temperatures from GCMs relative to observations is over ±1°C across more than one-half of mainland China. PRECIS demonstrates better representation of temperatures (except for HD) relative to GCMs. There is relatively better performance in Huanghuai, Jianghuai, Jianghan, south Yangzi River, and South China, whereas estimation is not as good in Xinjiang, the eastern part of northwest China, and the Tibetan Plateau. Bias-correction spatial disaggregation is used to downscale GCMs outputs, and bias correction is applied for PRECIS outputs, which demonstrate better improvement to a bias within ±0.2°C for Tm, Tmax, Tmin, and DTR and ±2 days for FD and HD. Furthermore, such improvement is also verified by the evidence of increased spatial correlation coefficient and symmetrical uncertainty, decreased root-mean-square error, and lower standard deviation for reproductions. It is seen from comprehensive ranking metrics that different downscaled models show the most improvement across different climatic regions, implying that optional ensembles of models should be adopted to provide sufficient high-quality climate information.

A new statistical downscaling approach for global evaluation of the CMIP5 precipitation outputs: Model development and application

Outputs of the Coupled Model Intercomparison Project Phase 5 (CMIP5) models have been widely used in studies of climate changes related to scenarios at global and regional scales. However, CMIP5 outputs cannot be used directly in analysis of climate changes due to coarse spatial resolution. Here, we proposed a new statistical downscaling method for the downscaling practice of the CMIP5 outputs, i.e. Bias-corrected and station-based Non-linear Regression Downscaling method based on Randomly-Moving Points (BNRD). And up to now, there are only two global downscaled CMIP5 precipitation datasets, i.e. NASA daily downscaled CMIP5 precipitation product and BCSD-based (Bias Correction Spatial Disaggregation) monthly downscaled CMIP5 precipitation product available online, which are both based on BCSD downscaling method. Hence, we evaluated downscaling performance of BNRD by comparing it with the downscaled CMIP5 outputs using the BCSD method in this current study. The results indicate that: (1) during the period for development of the model (1964–2005), the error between downscaled CMIP5 precipitation and GPCC ranges between −50 mm–50 mm at monthly scale. When compared to BCSD-downscaled CMIP5 precipitation, BNRD-downscaled CMIP5 precipitation well reduces errors and avoids underestimation and overestimation of GPCC by BCSD-downscaled CMIP5 precipitation; (2) during period for verification of the downscaling models (2006–2013), the maximum (182 mm), minimum (15 mm) and average (68 mm) RMSEs between BNRD-downscaled CMIP5 precipitation and GPCC are all lower than those between BCSD-downscaled CMIP5 precipitation and GPCC at continental scales. Besides, from the average precipitation viewpoint, BNRD-downscaled CMIP5 precipitation is in higher correlation (around 0.75) with GPCC than BCSD-downscaled CMIP5 precipitation under RCP4.5 and RCP8.5 scenarios at continental scales; (3) BNRD resolved the negative relation to GPCC in the areas near equator, including north part of the South America, southern Africa, northern Australia. In all, BNRD downscaling method developed in this study performs better in describing GPCC changes in both space and time when compared to BCSD and can be used for downscaling practice of CMIP5 and even potentially CMIP6 precipitation outputs over the globe.

Impact of climate change on precipitation patterns in Houston, Texas, USA

Abstract Extreme precipitation events damage infrastructure and property; thus, predicting future precipitation patterns in the context of climate change is important. In this study, precipitation projections from 36 downscaled General Circulation Models (GCMs) under two Representative Concentration Pathway (RCP) scenarios (RCP2.6 and RCP8.5) enabled examination of projected changes in future precipitation for the Clear Creek watershed in Houston, Texas, USA. Precipitation from 1950 to 2099 simulated with GCM were downscaled using the Bias-Correction Spatial Disaggregation method. Ten precipitation indices that represent precipitation amount, precipitation intensity, precipitation duration, and precipitation frequency evaluated how precipitation patterns will likely change. Results show that, at the annual scale, mean precipitation will significantly decrease based on RCP8.5, or remain relatively constant based on RCP2.6. Precipitation intensity and precipitation variability, however, will likely increase. Dry periods will lengthen significantly, whereas the length of wet spells will generally remain unchanged. At the monthly scale, the amount of precipitation, precipitation intensity, precipitation frequency and the length of wet spells will likely increase in September. In contrast, precipitation will likely decrease and dry spells will lengthen in April, May, August, November, and December. This finding illustrates that the intra-annual variability in precipitation will increase. The projected changes in precipitation under RCP8.5 are generally greater compared with RCP2.6. Differences between the scenarios are more pronounced towards the end of the century. Houston has recently experienced substantial precipitation variability, including severe drought and record-breaking precipitation from Hurricane Harvey in 2017. These events are consistent with the long-term GCM projections. Findings from this study can be applied to help manage water resources and enhance adaptability to climate change.

Estimating the reliability of a rainwater catchment system using the output data of general circulation models for the future period (case study: Birjand City, Iran)

The evaporation loss is a key component that affects managing the water resources of arid and semi-arid regions, where the resources are not uniformly distributed. Due to the climate condition and the physical characteristics of arid regions, a major proportion of precipitation is often unavailable through the flash floods, while merely a small fraction recharges the groundwater aquifers. Therefore, to achieve sustainable development, managing the water resources based on rainwater harvesting systems is inevitable. The main aim of this study was to assess the reliability of the rainwater harvesting systems designed for a future period (2017–2030). Thus, monthly climate data (e.g., precipitation) were simulated by using the outputs of general circulation models (GCMs) of the newest generation in the coupled model intercomparison project phase 5 (CMIP5) as the first step under two representative concentration pathways (RCPs), i.e., RCP2.6 and RCP8.5. Then, the data were downscaled spatially through bias-correction spatial disaggregation (BCSD) method for 20 grid points surrounding Birjand rain gauge station, east of Iran. Monthly precipitation of Birjand rain gauge station was interpolated automatically by means of the ordinary Kriging method during a future period, between 2017 until 2030. Data pre-processing in geostatistical methods, including investigating the normality, isotropic, trend analysis, and semi-variogram selection, was also carried out automatically through coding in MATLAB. Finally, using the interpolated monthly precipitation time series, the reliability of the precipitation harvesting systems was assessed for a different range of rooftop areas and storage tank capacities. Results indicated that this process can meet a significant volume of the household non-potable water demand by RWHS. Similar reliability values based on the projected monthly precipitation due to two GCMs and two RCPs were extracted for future period. The reliability of RWHS acquired by RCP2.6 will not widely diverse from RCP8.5. Totally, for semi-arid region, it is possible to supply about 20% of non-potable water demand in the future periods. Although this amount seems to be a low value, it should be noted that RWHS may prevent extra groundwater withdrawal and thus enhance the sustainability of the water resources.

The impact of climate change on inland waterway transport: Effects of low water levels on the Mackenzie River

Potential impact of climate change on the inland waterway transport of the Mackenzie River Basin (MRB) have been investigated through simulations of the Hydrologiske Byrån avdeling för Vattenbalans (HBV) hydrological model, the HBV-light, for the baseline, 1974–2004, and future periods of 2041–2070 and 2071–2100 over the MRB driven by precipitation and air temperature data of Representative Concentration Pathways, RCP4.5 and RCP8.5 climate scenarios of 10 Global Climate Models statistically downscaled by the Pacific Climate Impacts Consortium using the Bias-Correction Spatial Disaggregation. The average onset of spring snowmelt between 2041 and 2100 is projected to occur up to two weeks earlier than the historical (1974–2004) climate. Projected warmer temperature and higher precipitation will lead to higher runoff for MRB in winter and spring, at the expense of decreased summer streamflow and water level at Fort Simpson and Arctic Red River stations partly because projected increase in evapotranspiration will offset the projected increase in precipitation. Under a warmer climate, navigation issues related to low water levels are expected to increase, e.g., under RCP8.5 at Arctic Red River station, the number of days the water level above 5 m is projected to decrease from 74 days in 1974–2000 to 42 days in the 2080s. The summer durations that water levels of MRB will be at or above 3, 4, and 5 m are projected to decrease by 2.8%–22.2%, 10.0%–34.5%, and 16.2%–43.4%, respectively, which mean navigation problems would increase because safe transit through MRB depends on its water levels.

Intercomparison of machine learning methods for statistical downscaling: the case of daily and extreme precipitation

Statistical downscaling of Global Climate Models (GCMs) allows researchers to study local climate change effects decades into the future. A wide range of statistical models have been applied to downscaling GCMs but recent advances in machine learning have not been explored compared to traditional approaches. In this paper, we compare five Perfect Prognosis (PP) approaches, Ordinary Least Squares, Elastic-Net, and Support Vector Machine along with two machine learning methods Multi-task Sparse Structure Learning (MSSL) and Autoencoder Neural Networks. In addition, we introduce a hybrid Model Output Statistics and PP approach by modeling the residuals of Bias Correction Spatial Disaggregation (BCSD) with MSSL. Metrics to evaluate each method’s ability to capture daily anomalies, large-scale climate shifts, and extremes are analyzed. Generally, we find inconsistent performance between PP methods in their ability to predict daily anomalies and extremes as well as monthly and annual precipitation. However, results suggest that L1 sparsity constraints aid in reducing error through internal feature selection. The MSSL+BCSD coupling, when compared with BCSD, improved daily, monthly, and annual predictability but decreased performance at the extremes. Hence, these results suggest that the direct application of state-of-the-art machine learning methods to statistical downscaling does not provide direct improvements over simpler, longstanding approaches.

Water Resources Criticality Due to Future Climate Change and Population Growth: Case of River Basins in Utah, USA

This study investigates annual river basin–level water criticality (ratio of available water to withdrawals) considering effects of climate change on supply and of future population change on demand. A steady-state water balance model was developed to estimate the water mass budget and disaggregate the internal and external water supply sources at a river basin level. Future precipitation and evapotranspiration were dynamically downscaled under a moderate greenhouse gas emission scenario to 4-km horizontal resolution using a regional climate model for a decade centered on 2090. The climate data were also statistically downscaled via the bias correction spatial disaggregation (BCSD) method applied to the CMIP5 (the fifth phase of the Coupled Model Intercomparison Project) archives for four emission scenarios for decades centered on 2040 and 2090. Bootstrapping and k-nearest-neighbor (k-NN) algorithms were applied to simulate future water demand and external basin supply with uncertainty. Water stress is classified into four levels: (i) “very high-water stress” when the water criticality ratio <1.25, (ii) “high water stress” when criticality ranges from 1.25 to 2.5, (iii) “moderate water stress” when criticality ranges from 2.5 to 10, and (iii) “no water stress” when criticality >10. A basin with a criticality ratio of 1 indicates that the basin demand has been exactly met by the available supply sources. Most river basins have current water criticality less than 2 and are dependent on inflow from other basins (i.e., are not self sustaining). Future projections indicate modest increases in net available water for Utah through the end of the current century from climate change, with increasing vulnerability largely driven by population growth. Out of 11 basins, 4 achieve a high and 3 achieve a very high water stress status by the 2040s. Four basins achieve a very high water stress status by the 2090s compared to only two in a very high water stress status in the 2010s.

A rank-based approach for correcting systematic biases in spatial disaggregation of coarse-scale climate simulations

Use of General Circulation Model (GCM) precipitation and evapotranspiration sequences for hydrologic modelling can result in unrealistic simulations due to the coarse scales at which GCMs operate and the systematic biases they contain. The Bias Correction Spatial Disaggregation (BCSD) method is a popular statistical downscaling and bias correction method developed to address this issue. The advantage of BCSD is its ability to reduce biases in the distribution of precipitation totals at the GCM scale and then introduce more realistic variability at finer scales than simpler spatial interpolation schemes. Although BCSD corrects biases at the GCM scale before disaggregation; at finer spatial scales biases are re-introduced by the assumptions made in the spatial disaggregation process. Our study focuses on this limitation of BCSD and proposes a rank-based approach that aims to reduce the spatial disaggregation bias especially for both low and high precipitation extremes.BCSD requires the specification of a multiplicative bias correction anomaly field that represents the ratio of the fine scale precipitation to the disaggregated precipitation. It is shown that there is significant temporal variation in the anomalies, which is masked when a mean anomaly field is used. This can be improved by modelling the anomalies in rank-space. Results from the application of the rank-BCSD procedure improve the match between the distributions of observed and downscaled precipitation at the fine scale compared to the original BCSD approach. Further improvements in the distribution are identified when a scaling correction to preserve mass in the disaggregation process is implemented. An assessment of the approach using a single GCM over Australia shows clear advantages especially in the simulation of particularly low and high downscaled precipitation amounts.

Evaluation of Statistical Downscaling of North American Multimodel Ensemble Forecasts over the Western United States

AbstractThe skill of two statistical downscaled seasonal temperature and precipitation forecasts from the North American Multimodel Ensemble (NMME) was evaluated across the western United States at spatial scales relevant to local decision-making. Both statistical downscaling approaches, spatial disaggregation (SD) and bias correction spatial disaggregation (BCSD), exhibited similar correlative skill measures; however, the BCSD method showed superior tercile-based skill measures since it corrects for variance deflation in NMME ensemble averages. Geographic and seasonal variations in downscaled forecast skill revealed patterns across the complex topography of the western United States not evident using coarse-scale skill assessments, particularly in regions subject to inversions and variability in orographic precipitation ratios. Similarly, differences in the skill of cool-season temperature and precipitation forecasts issued when the fall El Niño–Southern Oscillation (ENSO) signal was strong versus ENSO-neutral years were evident across topographic gradients in the northwestern United States.

Intercomparison of multiple statistical downscaling methods: multi-criteria model selection for South Korea

A number of statistical downscaling methodologies have been introduced to bridge the gap in scale between outputs of climate models and climate information needed to assess potential impacts at local and regional scales. Four statistical downscaling methods [bias-correction/spatial disaggregation (BCSD), bias-correction/constructed analogue (BCCA), multivariate adaptive constructed analogs (MACA), and bias-correction/climate imprint (BCCI)] are applied to downscale the latest climate forecast system reanalysis (CFSR) data to stations for precipitation, maximum temperature, and minimum temperature over South Korea. All methods are calibrated with observational station data for 19 years from 1973 to 1991 and validated for the more recent 19-year period from 1992 to 2010. We construct a comprehensive suite of performance metrics to inter-compare methods, which is comprised of five criteria related to time-series, distribution, multi-day persistence, extremes, and spatial structure. Based on the performance metrics, we employ technique for order of preference by similarity to ideal solution (TOPSIS) and apply 10,000 different weighting combinations to the criteria of performance metrics to identify a robust statistical downscaling method and important criteria. The results show that MACA and BCSD have comparable skill in the time-series related criterion and BCSD outperforms other methods in distribution and extremes related criteria. In addition, MACA and BCCA, which incorporate spatial patterns, show higher skill in the multi-day persistence criterion for temperature, while BCSD shows the highest skill for precipitation. For the spatial structure related criterion, BCCA and MACA outperformed BCSD and BCCI. From the TOPSIS analysis, we found that MACA is the most robust method for all variables in South Korea, and BCCA and BCSD are the second for temperature and precipitation, respectively. We also found that the contribution of the multi-day persistence and spatial structure related criteria are crucial to ranking the skill of statistical downscaling methods.

The effects of climate downscaling technique and observational data set on modeled ecological responses.

Assessments of future climate change impacts on ecosystems typically rely on multiple climate model projections, but often utilize only one downscaling approach trained on one set of observations. Here, we explore the extent to which modeled biogeochemical responses to changing climate are affected by the selection of the climate downscaling method and training observations used at the montane landscape of the Hubbard Brook Experimental Forest, New Hampshire, USA. We evaluated three downscaling methods: the delta method (or the change factor method), monthly quantile mapping (Bias Correction-Spatial Disaggregation, or BCSD), and daily quantile regression (Asynchronous Regional Regression Model, or ARRM). Additionally, we trained outputs from four atmosphere-ocean general circulation models (AOGCMs) (CCSM3, HadCM3, PCM, and GFDL-CM2.1) driven by higher (A1fi) and lower (B1) future emissions scenarios on two sets of observations (1/8º resolution grid vs. individual weather station) to generate the high-resolution climate input for the forest biogeochemical model PnET-BGC (eight ensembles of six runs).The choice of downscaling approach and spatial resolution of the observations used to train the downscaling model impacted modeled soil moisture and streamflow, which in turn affected forest growth, net N mineralization, net soil nitrification, and stream chemistry. All three downscaling methods were highly sensitive to the observations used, resulting in projections that were significantly different between station-based and grid-based observations. The choice of downscaling method also slightly affected the results, however not as much as the choice of observations. Using spatially smoothed gridded observations and/or methods that do not resolve sub-monthly shifts in the distribution of temperature and/or precipitation can produce biased results in model applications run at greater temporal and/or spatial resolutions. These results underscore the importance of carefully considering field observations used for training, as well as the downscaling method used to generate climate change projections, for smaller-scale modeling studies. Different sources of variability including selection of AOGCM, emissions scenario, downscaling technique, and data used for training downscaling models, result in a wide range of projected forest ecosystem responses to future climate change.

Hydrologic extremes – an intercomparison of multiple gridded statistical downscaling methods

Abstract. Gridded statistical downscaling methods are the main means of preparing climate model data to drive distributed hydrological models. Past work on the validation of climate downscaling methods has focused on temperature and precipitation, with less attention paid to the ultimate outputs from hydrological models. Also, as attention shifts towards projections of extreme events, downscaling comparisons now commonly assess methods in terms of climate extremes, but hydrologic extremes are less well explored. Here, we test the ability of gridded downscaling models to replicate historical properties of climate and hydrologic extremes, as measured in terms of temporal sequencing (i.e. correlation tests) and distributional properties (i.e. tests for equality of probability distributions). Outputs from seven downscaling methods – bias correction constructed analogues (BCCA), double BCCA (DBCCA), BCCA with quantile mapping reordering (BCCAQ), bias correction spatial disaggregation (BCSD), BCSD using minimum/maximum temperature (BCSDX), the climate imprint delta method (CI), and bias corrected CI (BCCI) – are used to drive the Variable Infiltration Capacity (VIC) model over the snow-dominated Peace River basin, British Columbia. Outputs are tested using split-sample validation on 26 climate extremes indices (ClimDEX) and two hydrologic extremes indices (3-day peak flow and 7-day peak flow). To characterize observational uncertainty, four atmospheric reanalyses are used as climate model surrogates and two gridded observational data sets are used as downscaling target data. The skill of the downscaling methods generally depended on reanalysis and gridded observational data set. However, CI failed to reproduce the distribution and BCSD and BCSDX the timing of winter 7-day low-flow events, regardless of reanalysis or observational data set. Overall, DBCCA passed the greatest number of tests for the ClimDEX indices, while BCCAQ, which is designed to more accurately resolve event-scale spatial gradients, passed the greatest number of tests for hydrologic extremes. Non-stationarity in the observational/reanalysis data sets complicated the evaluation of downscaling performance. Comparing temporal homogeneity and trends in climate indices and hydrological model outputs calculated from downscaled reanalyses and gridded observations was useful for diagnosing the reliability of the various historical data sets. We recommend that such analyses be conducted before such data are used to construct future hydro-climatic change scenarios.

Assessing the impact of CMIP5 climate multi-modeling on estimating the precipitation seasonality and timing

This paper investigates the effect of a Bayesian Model Averaging (BMA) method on simulated precipitation over the Columbia River Basin using two statistically downscaled climate datasets, i.e., Bias-Correction Spatial Disaggregation (BCSD) and Multivariate Adaptive Constructed Analogs (MACA). To this end daily observed and simulated precipitation are used to calculate different indices focusing solely on seasonality, event-timing, and variability in timing (persistence) of the precipitation. The climate model weights are estimated for each cell (6 × 6 km) of the Columbia River Basin using daily time series for the historical period 1970–1999 from the ten Global Climate Models (GCMs) participating to the Coupled Model Intercomparison Project platform, Phase 5. The results show that BMA results in more than 15 % improvement/reduction in mean absolute error as compared to the individual GCMs. The improvement is, in general, higher for the MACA models than for the BCSD models. The results of variability in precipitation timing show that extreme precipitation events are mostly not persistent (i.e., occurring in different periods throughout the year), i.e., more than 75 % of the grid cells with an elevation above 900 m indicate persistence values less than 0.2 whereas nearly 70 % of the high elevation cells indicate such low persistence. Further we find that the variability in persistence is higher in high elevation cells than those with low elevation. The picture is different for MACA ensembles as the simulated persistence of extreme precipitation events is higher than that for the observed and BCSD datasets.

Projected Changes in Climate Extremes over the Northeastern United States

AbstractProjections of historical and future changes in climate extremes are examined by applying the bias-correction spatial disaggregation (BCSD) statistical downscaling method to five general circulation models (GCMs) from phase 5 of the Coupled Model Intercomparison Project (CMIP5). For this analysis, 11 extreme temperature and precipitation indices that are relevant across multiple disciplines (e.g., agriculture and conservation) are chosen. Over the historical period, the simulated means, variances, and cumulative distribution functions (CDFs) of each of the 11 indices are first compared with observations, and the performance of the downscaling method is quantitatively evaluated. For the future period, the ensemble average of the five GCM simulations points to more warm extremes, fewer cold extremes, and more precipitation extremes with greater intensities under all three scenarios. The changes are larger under higher emissions scenarios. The inter-GCM uncertainties and changes in probability distributions are also assessed. Changes in the probability distributions indicate an increase in both the number and interannual variability of future climate extreme events. The potential deficiencies of the method in projecting future extremes are also discussed.