DOC PREVIEW
Utility of Daily vs. Monthly Large-scale Climate Data

This preview shows page 1-2-3-4 out of 13 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 13 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 13 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 13 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 13 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 13 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Hydrol. Earth Syst. Sci., 12, 551–563, 2008www.hydrol-earth-syst-sci.net/12/551/2008/© Author(s) 2008. This work is distributed underthe Creative Commons Attribution 3.0 License.Hydrology andEarth SystemSciencesUtility of daily vs. monthly large-scale climate data: anintercomparison of two statistical downscaling methodsE. P. Maurer1and H. G. Hidalgo21Civil Engineering Department, Santa Clara University, Santa Clara, CA, USA2Scripps Institution of Oceanography, Div. of Climate Atmospheric Sciences & Physical Oceanography, La Jolla, CA, USAReceived: 16 August 2007 – Published in Hydrol. Earth Syst. Sci. Discuss.: 12 September 2007Revised: 20 December 2007 – Accepted: 5 February 2008 – Published: 13 March 2008Abstract. Downscaling of climate model data is essential tolocal and regional impact analysis. We compare two meth-ods of statistical downscaling to produce continuous, grid-ded time series of precipitation and surface air temperatureat a 1/8-degree (approximately 140 km2per grid cell) reso-lution over the western U.S. We use NCEP/NCAR Reanaly-sis data from 1950–1999 as a surrogate General CirculationModel (GCM). The two methods included are constructedanalogues (CA) and a bias correction and spatial downscal-ing (BCSD), both of which have been shown to be skillfulin different settings, and BCSD has been used extensivelyin hydrologic impact analysis. Both methods use the coarsescale Reanalysis fields of precipitation and temperature aspredictors of the corresponding fine scale fields. CA down-scales daily large-scale data directly and BCSD downscalesmonthly data, with a random resampling technique to gener-ate daily values. The methods produce generally comparableskill in producing downscaled, gridded fields of precipita-tion and temperatures at a monthly and seasonal level. Fordaily precipitation, both methods exhibit limited skill in re-producing both observed wet and dry extremes and the dif-ference between the methods is not significant, reflecting thegeneral low skill in daily precipitation variability in the re-analysis data. For low temperature extremes, the CA methodproduces greater downscaling skill than BCSD for fall andwinter seasons. For high temperature extremes, CA demon-strates higher skill than BCSD in summer. We find that thechoice of most appropriate downscaling technique dependson the variables, seasons, and regions of interest, on theavailability of daily data, and whether the day to day cor-respondence of weather from the GCM needs to be repro-Correspondence to: E. P. Maurer([email protected])duced for some applications. The ability to produce skillfuldownscaled daily data depends primarily on the ability of theclimate model to show daily skill.1 IntroductionClimate models are the primary tool to evaluate the pro-jected future response of the atmosphere-land-ocean systemto changing atmospheric composition (MacCracken et al.,2003; Stocker et al., 2001), and they underpin most climatechange impacts studies (Wilby and Harris, 2006). Howeverthere is a mismatch between the grid resolution of current cli-mate models (generally hundreds of kilometers), and the res-olution needed by environmental impacts models (typicallyten kilometers or less). Downscaling is the process of trans-forming information from climate models at coarse resolu-tions to a fine spatial resolution. Downscaling is necessary,as the underlying processes described by the environmentalimpact models are very sensitive to the nuances of local cli-mate, and the drivers of local climate variations, such as to-pography, are not captured at coarse scales.There are two broad categories of downscaling: dynamic(which simulates physical processes at fine scales) and sta-tistical (which transforms coarse-scale climate projections toa finer scale based on observed relationships between theclimate at the two spatial resolutions) (Christensen et al.,2007). Dynamic downscaling, nesting a fine scale climatemodel in a coarse scale model, produces spatially completefields of climate variables, thus preserving some spatial cor-relation as well as physically plausible relationships betweenvariables. However, dynamic downscaling is very computa-tionally intensive, making its use in impact studies limited,and essentially impossible for multi-decade simulations withPublished by Copernicus Publications on behalf of the European Geosciences Union.552 E. P. Maurer and H. G. Hidalgo: Daily vs. monthly climate data in statistical downscalingdifferent global climate models and/or multiple greenhousegas emission scenarios. Thus, most impacts studies rely onsome form of statistical downscaling, where variables of in-terest can be downscaled using historical observations. Therehas been extensive work developing and intercomparing sta-tistical downscaling techniques for climate impact studies(Giorgi et al., 2001; Wilby and Wigley, 1997).Statistical downscaling is typically used to predict onevariable at one site, though some techniques for simultane-ous downscaling to multiple sites for precipitation have beendeveloped (Harpham and Wilby, 2005; Wilks, 1999). How-ever, for studies of some climate impacts such as river basinhydrology, it is important to downscale simultaneous valuesof multiple variables (such as precipitation and temperature)over large, heterogeneous areas, while maintaining physi-cally plausible spatial and temporal relationships, though fewdownscaling techniques have been developed to do this.In this study we compare two methods of statistical down-scaling to produce gridded time series of precipitation andsurface air temperature at a fine resolution over a large spa-tial domain. These two methods are termed constructed ana-logues (CA, Hidalgo et al., 2008; van den Dool, 1994) andbias correction and spatial downscaling (BCSD, Wood et al.,2004). The CA method has been shown to have significantskill in reproducing the variability of daily precipitation andtemperature over the contiguous United States (U.S.), in par-ticular in the western coast (Hidalgo et al., 2008). The BCSDmethod has been shown to provide downscaling capabilitiescomparable to other statistical and dynamical methods in thecontext of hydrologic impacts (Wood et al., 2004).The main conceptual difference between the two methodscompared here is that the daily correspondence of the coarseresolution and the fine resolution patterns is maintained inthe CA method, while in the BCSD the monthly patternsare conserved but daily


Utility of Daily vs. Monthly Large-scale Climate Data

Download Utility of Daily vs. Monthly Large-scale Climate Data
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Utility of Daily vs. Monthly Large-scale Climate Data and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Utility of Daily vs. Monthly Large-scale Climate Data 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?