Fig 3 - uploaded by David Favis-Mortlock
Content may be subject to copyright.
Mean annual precipitation (mm) for Northern Ireland (NIR), 1961–1990 (after Betts 2002b). Also shown are locations of places mentioned in text
Source publication
Statistical downscaling bridges the gap between General Circulation Model (GCM) output and climate-impacts modellers' requirements. Recent development of user-friendly software has facilitated use of statistical downscaling methods by the wider climate-impacts community. Simplicity of use, however, does not imply underlying conceptual simplicity. O...
Similar publications
A procedure for correcting systematic biases across multiple variables is presented. This procedure operates in the frequency domain, using the cross-spectrum across variables to correct bias across each frequency band. The proposed approach is termed multivariate frequency bias correction or MFBC. The approach is illustrated using global climate m...
Climate projections simulated by Global Climate Models (GCMs) are often used
for assessing the impacts of climate change. However, the relatively coarse
resolutions of GCM outputs often preclude their application to
accurately assessing the effects of climate change on finer regional-scale
phenomena. Downscaling of climate variables from coarser to...
Climate projections simulated by Global Climate Models (GCM) are often used for assessing the impacts of climate change. However, the relatively coarse resolutions of GCM outputs often precludes their application towards accurately assessing the effects of climate change on finer regional scale phenomena. Downscaling of climate variables from coars...
Downscaling of atmospheric climate parameters is a sophisticated tool to develop statistical relationships between large-scale atmospheric variables and local-scale meteorological variables. In this study, the variables selected from the National Center for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) reanalysis...
The gap in resolution between existing global climate model output and that sought by decision-makers drives an ongoing need for climate downscaling. Here we test the extent to which developments in deep learning can out-perform existing statistical approaches for downscaling historical rainfall in the highly complex terrain setting of New Zealand....
Citations
... The importance of variable selection for the SD models is even more significant in the Susurluk Basin. While establishing the SD methods, long-distance grids can also affect the local climate in different periods (Wilby and Wigley 2000;Crawford et al. 2007;Borges et al. 2017). Also, resolution differences between GCMs and the predictor set can source additional uncertainty (Amjad et al. 2020). ...
... Using Spearman correlation at a significance level of 1%, the choice of the predictors was performed with the grid values, in which the related meteorology station is located. In addition to the predictor selection, using more than one grid is vital as remote grids may affect the local climate at different times (Wilby and Wigley 2000;Crawford et al. 2007;Borges et al. 2017). Also, the resolution difference between the predictor set and the GCMs may be an additional cause of uncertainty (Amjad et al. 2020). ...
... For this reason, C2 and 1.50°9 1.50°grid resolutions were chosen because the mean of GCM in CMIP6 resolutions was about 1.50°9 1.50°. So, it can be said that long-distance grids effectively predict precipitation (Wilby and Wigley 2000;Crawford et al. 2007;Borges et al. 2017). ...
Unlabelled:
The impacts of climate change on current and future water resources are important to study local scale. This study aims to investigate the prediction performances of daily precipitation using five regression-based statistical downscaling models (RBSDMs), for the first time, and the ERA-5 reanalysis dataset in the Susurluk Basin with mountain and semi-arid climates for 1979-2018. In addition, comparisons were also performed with an artificial neural network (ANN). Before achieving the aim, the effects of atmospheric variables, grid resolution, and long-distance grid on precipitation prediction were holistically investigated for the first time. Kling-Gupta efficiency was modified and used for holistic evaluation of statistical moments parameters at precipitation prediction comparison. The standard triangular diagram, quite new in the literature, was also modified and used for graphical evaluation. The results of the study revealed that near grids were more effective on precipitation than single or far grids, and 1.50° × 1.50° resolution showed similar performance to 0.25° × 0.25° resolution. When the polynomial multivariate adaptive regression splines model, which performed slightly higher than ANN, tended to capture skewness and standard deviation values of precipitations and to hit wet/dry occurrence than the other models, all models were quite well able to predict the mean value of precipitations. Therefore, RBSDMs can be used in different basins instead of black-box models. RBSDMs can also be established for mean precipitation values without dry/wet classification in the basin. A certain success was observed in the models; however, it was justified that bias correction was required to capture extreme values in the basin.
Supplementary information:
The online version contains supplementary material available at 10.1007/s00477-022-02345-5.
... In addition to model structure, downscaled series are also sensitive to the choice of predictor variables (Crawford et al. 2007;Borges et al. 2017). Other work warns about equifinality and transferability issues-whereby alternative predictor sets yield equivalent skill during model calibration and validation but different projections when applied under changed climate conditions (Fu et al. 2018). ...
Global efforts to upgrade water, drainage, and sanitation services are hampered by hydrometeorological data-scarcity plus uncertainty about climate change. Intensity–duration–frequency (IDF) tables are used routinely to design water infrastructure so offer an entry point for adapting engineering standards. This paper begins with a novel procedure for guiding downscaling predictor variable selection for heavy rainfall simulation using media reports of pluvial flooding. We then present a three-step workflow to: (1) spatially downscale daily rainfall from grid-to-point resolutions; (2) temporally scale from daily series to sub-daily extreme rainfalls and; (3) test methods of temporal scaling of extreme rainfalls within Regional Climate Model (RCM) simulations under changed climate conditions. Critically, we compare the methods of moments and of parameters for temporal scaling annual maximum series of daily rainfall into sub-daily extreme rainfalls, whilst accounting for rainfall intermittency. The methods are applied to Kampala, Uganda and Kisumu, Kenya using the Statistical Downscaling Model (SDSM), two RCM simulations covering East Africa (CP4 and P25), and in hybrid form (RCM-SDSM). We demonstrate that Gumbel parameters (and IDF tables) can be reliably scaled to durations of 3 h within observations and RCMs. Our hybrid RCM-SDSM scaling reduces errors in IDF estimates for the present climate when compared with direct RCM output. Credible parameter scaling relationships are also found within RCM simulations under changed climate conditions. We then discuss the practical aspects of applying such workflows to other city-regions.
... material for reference. Although many statistical downscaling comparisons have focused on the methodology applied to a single predictor set(Gutmann et al. 2014;Dixon et al. 2016;Lanzante et al. 2018), other studies have highlighted the importance of predictor set Accepted for publication in Journal of Hydrometeorology. DOI 10.1175/JHM-D-21-0142.1.Crawford et al. 2007;González-Rojí et al. 2019), thus a comparison evaluating both the predictor selection and the estimation algorithm is a valuable next step.e. MetricsIn evaluating climate downscaling methods, there are several key metrics to assess. First and foremost, climate downscaling methods must predict the expected mean and variability observed in ...
Statistical processing of numerical model output has been a part of both weather forecasting and climate applications for decades. Statistical techniques are used to correct systematic biases in atmospheric model outputs and to represent local effects that are unresolved by the model, referred to as downscaling. Many downscaling techniques have been developed, and it has been difficult to systematically explore the implications of the individual decisions made in the development of downscaling methods. Here we describe a unified framework that enables the user to evaluate multiple decisions made in the methods used to statistically post-process output from weather and climate models. The semble eneralized nalog egression ownscaling (En-GARD) method enables the user to select any number of input variables, predictors, mathematical transformations, and combinations, for use in parametric or non-parametric downscaling approaches. En-GARD enables explicitly predicting both the probability of event occurrence and the event magnitude. Outputs from En-GARD include errors in model fit, enabling the production of an ensemble of projections through sampling of the probability distributions of each climate variable. We apply En-GARD to regional climate model simulations to evaluate the relative importance of different downscaling method choices on simulations of the current and future climate. We show that choice of predictor variables is the most important decision affecting downscaled future climate outputs, while having little impact on the fidelity of downscaled outcomes for current climate. We also show that weak statistical relationships prevent such approaches from predicting large changes in extreme events on a daily time scale.
... GCMs predictors and local predictant are considered as inputs and output components of transfer functions, respectively. In these downscaling methods, the most crucial step is to discover the most reliable predictors to establish a statistical relationship (Crawford et al. 2007). Many researchers consider a physical relation as a proper criterion while some of them express that behavior of rainfall is a consequence of one specific variable (Goyal and Ojha 2012), so they selected single ones or considered a set of predictors. ...
To turn General Circulation Models (GCMs) projection toward better assessment, it is crucial to employ a downscaling process to get more reliability of their outputs. The data-driven based downscaling techniques recently have been used widely, and predictor selection is usually considered as the main challenge in these methods. Hence, this study aims to examine the most common approaches of feature selection in the downscaling of daily rainfall in two different climates in Iran. So, the measured daily rainfall and National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) predictors were collected, and Support Vector Machine (SVM) was considered as downscaling methods. Also, a complete set of comparative tests considering all dimensions was employed to identify the best subset of predictors. Results indicated that the skill of various selection methods in different tests is significantly different. Despite a few partial superiorities viewed between selection models, they not presented an obvious distinction. However, regarding all related factors, it may be deduced that the Stepwise Regression Analysis (SRA) and Bayesian Model Averaging (BMA) are better than others. Also, the finding of this study showed that there are some weaknesses in the interpretation of SRA, so concerning this issue, it may be concluded that BMA has more reliable performance. Furthermore, results indicated that generally, the downscaling procedure has more accuracy in arid climate than cold-semi arid climate.
... Furthermore, the variability in seasons which is controlled by the monsoon rains in the region is increased when climate stations are sparse as was in the study area (Wilby 2008;Wilby and Dawson 2013). The next major limitation was the choice of large atmospheric variables (predictors) for the calibration of SDSM which has been reported to be sensitive during downscaling (Crawford et al. 2007;Koukidis and Berg 2009;Mullan et al. 2012). Moreover, studies such as Haylock et al. (2006), Chen et al. (2010) and Liu et al. (2011) found rainfall downscaling in summer months which was mostly the peak of rainfall in Ghana to be difficult in other regions. ...
The slightest change in rainfall could have a significant impact on rain-fed agriculture in countries like Ghana. This study evaluated for the first time the performance of the statistical downscaling model (SDSM-DC) at 2m spatial resolution in simulating rainfall in Ghana for the base period 1981–2010. It further analysed the projected changes in seasonal rainfall pattern across different agro-ecological zones for the twenty-first century under RCP 4.5 and 8.5 emission scenarios over Ghana. Ensemble mean of simulated rainfall data (2011–2099) generated by 43 GCMs in the Coupled Model Intercomparison Project Phase 5 (CMIP5) were used as base factors for local future climate scenarios generation. Performance analysis of SDSM-DC shows a Nash–Sutcliffe efficiency, percent bias and RMSE observations standard deviation ratio of 0.88, −19 and 0.34, respectively. Generally, seasonal rainfall amount is expected to increase between 10 and 40% in all the agro-ecological zones in Ghana by the end of the twenty-first century. Off-season rainfall in December–February shows more than 100% increase in the Guinea Savannah zone. Rainfall projected under RCP 4.5 was on average 2% higher than RCP 8.5 in all the seasons throughout the century. Based on these results, it is appropriate to suggest a high incidence of flooding across Ghana in the twenty-first century. This could have dire consequences on agriculture which contribute to a large proportion of Ghana’s GDP. Therefore, for sustainable food production and security in the twenty-first century, Ghana needs climate adaptation policies and programmes that encourage the design and implementation of early warning systems of meteorological hazards and the introduction of new crop varieties that are flood tolerant.
... The historical data shows that the frequency of these events is increasing year after year. As per the report of the Inter-governmental Panel on Climate Change (IPCC) 2013-AR5, the predicted increase in temperature from the year 1990 to 2100 will be approximately 1.4°C to 5.8° C, which will melt the glaciers of the state [5][6]. The report also said that the change in local precipitation and temperature due to climate change might increase the hazards like droughts and floods as well as it may increase their severity [7][8]. ...
Uttarakhand, a Himalayan state of India, may experience an increase in temperature of 1.4°C to 5.8°C by 2100 due to global warming. The rise in temperature may melt the glaciers of the state and may have some significant impact on the rainfall. In this study, we have quantified the changes in the rainfall of the state. Also, an attempt has been made to evaluate the impact of climate change on rainfall. The future rainfall can be estimated by using a global circulation model (GCM). However, due to the very coarse spatial resolution of the different GCM, we cannot use them directly. For matching this spatial inequality between the GCM output and historical precipitation data, we used the statistical downscaling technique. In the present study, we have examined the suitability of the artificial neural network with principal component analysis for downscaling the rainfall for different hilly districts of the state. We used the GCM model developed by Canadian Earth System Model, and the Indian metrological department gridded rainfall data. We performed the analysis for the different scenarios to visualize the impact of climate change on rainfall trends for all nine hilly districts of Uttarakhand. Results show that there was a clear indication of climate change in upper Himalayan Districts like Pithoragarh, Rudraprayag, and Chamoli, which was observed from the peak of monthly rainfall. The percentage change of monsoon rainfall in the future may go up to 200 % in the case of RCP8.5, and the change maybe around 180% for RCP4. Also, the volume of rainfall may increase in the case of RCP8.5 from July to September as compared to the historical data, i.e., there may be a shifting of monsoon rainfall in the future.
... For a more detailed analysis of the predictors and in order to identify potential better correlations with the historical records, the lagging of daily predictor variables could be applied as well as suggested for some authors (e.g., Harpham and Wilby 2005;Crawford et al. 2007) with the purpose of revealing hidden direct relationships between predictand and predictors; this is because predictors from distant grid-boxes may also influence the local climate in distinct time. ...
Background
Considering the lack of research over this region the Statistical Downscaling Model (SDSM) was used as a tool for downscaling meteorological data statistically over four representative regions in the eastern side of Colombia. Data from the two Global Climate Models CanESM2 and IPSL-CM5A-MR, which are part of the CMIP5-project have been used to project future maximum and minimum temperature, precipitation and relative humidity for the periods 2021–2050 and 2071–2100. For both models, the Representative Concentration Pathways RCP2.6 and RCP8.5 were considered, representing two different possible future emission trajectories and radiative forcings. Predictor variables from the National Centre for Environmental Prediction (NCEP-DOE 2) reanalysis dataset, together with analyzed correlation coefficient (R) and root mean square error (RMSE) were used as performance indicators during the calibration and validation process.
Results
Results indicate that Maximum and minimum temperature is projected to increase for both Global Climate Models and both Representative Concentration Pathways; relative humidity shows a decreasing trend for all scenarios and all regions; and precipitation shows a slight decrease over three regions and an increase over the warmest region. As expected, the results of the simulation for the period 2071–2100 show a more drastic change when compared to the baseline period of observations.
Conclusions
The SDSM model proves to be efficient in the downscaling of maximum/minimum temperature as well as relative humidity over the studied regions; while showing a lower performance for precipitation, agreeing with the results for other statistical downscaling studies. The results of the projections offer good information for the evaluation of possible future-case scenarios and decision-making management.
... This implementation of SDSM follows the procedures customarily applied before (Wilby et al. 1998Gulacha and Mulungu 2017). Examples of more specific applications of SDSM can be found elsewhere (Hanssen-Bauer et al. 2005;Huth 2005;Crawford et al. 2007;Babel 2013, 2014;Wilby and Dawson 2013). ...
There have been numerous statistical and dynamical downscaling model comparisons. However, differences in model skill can be distorted by inconsistencies in experimental set-up, inputs and output format. This paper harmonizes such factors when evaluating daily precipitation downscaled over the Iberian Peninsula by the Statistical DownScaling Model (SDSM) and two configurations of the dynamical Weather Research and Forecasting Model (WRF) (one with data assimilation (D) and one without (N)). The ERA-Interim reanalysis at resolution provides common inputs for spinning-up and driving the WRF model and calibrating SDSM. WRF runs and SDSM output were evaluated against ECA&D stations, TRMM, GPCP and EOBS gridded precipitation for 2010–2014 using the same suite of diagnostics. Differences between WRF and SDSM are comparable to observational uncertainty, but the relative skill of the downscaling techniques varies with diagnostic. The SDSM ensemble mean, WRF-D and ERAI have similar correlation scores (–0.7), but there were large variations amongst SDSM ensemble members (–0.6). The best Linear Error in Probability Space (–0.007) and simulations of precipitation amount were achieved by individual members of the SDSM ensemble. However, the Brier Skill Score shows these members do not improve the prediction by ERA-Interim, whereas precipitation occurrence is reproduced best by WRF-D. Similar skill was achieved by SDSM when applied to station or gridded precipitation data. Given the greater computational demands of WRF compared with SDSM, clear statements of expected value-added are needed when applying the former to climate impacts and adaptation research.
... The choice of predictors and the selection of grid boxes in the LSV domain have an important impact on the performance of the downscaling (Crawford et al., 2007;Liu et al., 2013;Rashid et al., 2016). Many predictor selection methods were proposed by previous literatures, including circulation indices (Fan et al., 2015), wavelet techniques Rashid, 2016), step-wise screening (Wilby et al., 2002) and principal component analysis (PCA) (Huth, 1999). ...
Statistical downscaling (SD) of daily precipitation is a challenging task, and the identification of predictors is crucial for constructing SD models. This study focuses on identifying SD predictors for summer (June–September) daily precipitation in China. Six large-scale variables (LSVs) in ERA-Interim reanalysis were used to select predictors for 177 sites. For each site, the predictor identification was conducted by searching the grid box having the best correlation to precipitation in a three-dimensional way: across different grid boxes and multiple pressure levels. The result indicates that correlations are often sensitive to the pressure levels. Adjacent sites share similar spatial patterns of correlations, indicating regionally different physical relations between LSVs and precipitation. The predictor selection reasonably reflects the regional circulations related to precipitation. Twelve candidate predictors were used to train generalized linear models by least absolute shrinkage and selection operator (LASSO) algorithm. The validation indicates the models have generally high performance, and also shows relatively poor performance for the sites in North China, Northwest China, and Yunnan when compared to that in the east of China. The downscaled outputs can roughly reflect the annual variations of summer total precipitation and rainy days. Two experiments on the stationarity assumption of the models under different climate conditions were conducted, indicating that no areas/sites were found significantly violated the stationarity assumption. This study presents guidance on how to select suitable predictors for downscaling daily precipitation in different areas of China.
... Downscaling results may be sensitive to the predictor domain under consideration (Benestad, 2002;Mtongori et al., 2016). Many researchers have experimented with the effect of the domain size in their downscaling studies (Timbal and McAvaney, 2001;Wetterhall et al., 2007;Crawford et al., 2007;Yang et al., 2013). Benestad (2002) reported that the choice of predictor domain has a significant role in the downscaling process where a smaller area of the predictor domain may provide useful downscaling results over a target region. ...
Selection of suitable predictors for downscaling local‐scale precipitation from the wide range of large‐scale predictors available in National Center for Atmospheric Research/National Centers for Environmental Prediction (NCAR/NCEP) reanalysis is a challenging task because of the existence of the complex interactions between local‐scale predictands and large‐scale predictor fields. An attempt was made to assess how well different large‐scale predictors were able to reproduce local‐scale monsoon precipitation over seven homogeneous zones of India through statistical downscaling. For calibration of downscaling (DS) models, the principal component (PC)‐based multiple linear regression approach was adopted where each raw grid‐point predictor field transformed into PCs using empirical orthogonal function (EOF) analysis. The predictors consistently producing better downscaled results across four nonoverlapping calibration and validation periods were identified as “superior predictor” (SP). It was found that some common predictors like precipitable water; specific and relative humidity at different levels have emerged as SP predictors over several zones. In general, SP predictors have not been much sensitive with small changes in the domain size. However, a decline in performances of DS models was noticed for the majority of SP predictors for a large increase in the size of domains. Especially, the largest South Asia domain has been the most inappropriate domain as very few predictors found to be suitable for downscaling. In general, about 40% out of 36 numbers of combined predictors were identified as potential SP predictors over the majority of the zones. Several numbers of combined SP predictors have also produced slightly superior skills compared to single SP predictors. In many cases, predictors showing poor performance as single predictors have produced improved performances when combined with other predictors.
