| Date: | 2016-08-24 |
|---|---|
| Version: | 4 |
| Web site: | http://gdo-dcp.ucllnl.org/downscaled_cmip_projections/ |
Contents
The web-service currently only supports the option of downloading the downscaled data in text format if the request specifies a location or statistics. For all other requests, the netCDF format is used, mainly because its much more efficient in terms of processing requirements at the archive backend.
See the answer to question 1.3 for more information on working with netCDF files.
This occasionally happens when a spam filter catches the notification email in the requestor's email system. The sender of email sent from the archive appears as "no-reply@gdo-dcp.ucllnl.org".
There has been one occasion where the website was overwhelmed with requests causing a processing wait of more than one day. There is now a feature that reports an estimated processing time when the request is submitted. On rare occasions, there have been hardware and networking problems. So please let us know if it seems that something is broken.
Large requests of daily LOCA may take more than a day or two to process. And if there are already jobs in the queue ahead of your request it could be several days before processing is complete.
Utilities are available that allow the reading and conversion of the netcdf format. See the 'Tools' section at the bottom of the Links tab.
Another option, if you're a Matlab user, is to update your software to Release 2008b or later, which has embedded netCDF utilities.
This is in referrence to the checkboxes in 'Step 2.6: Emissions Scenarios, Climate Models and Runs' under the 'Projections: Subset Request' tab.
The runs represent initial atmosphere and ocean conditions for the climate projection simulation. These depend on the 20th century "control" simulation used to define initial conditions in the 21st century. The number of runs for a given model was a choice made by the different modeling centers.
More information can be found on the PCMDI site (http://cmip-pcmdi.llnl.gov/) and in our downscaling technical document (http://gdo-dcp.ucllnl.org/downscaled_cmip_projections/techmemo/downscaled_climate.pdf).
We dont't produce a shapefile for the area delineated in those requests. The *.cdl file that are output as part of the process is used to mask the netcdf data. They contain most the information you would need to produce a shapefile with GIS software (latitudes, longitudes, and mask value). These are text files and you could possibly make an ascii grid file from one of these.
We don't use ArcMap for viewing/processing NetCDF files, so probably cannot help with this issue. For viewing netcdf files in Windows, you might try something like Panoply .
CMIP3 Daily projections were added to the archive in August 2011. CMIP5 daily projections were added in May 2013.
An alternative method of deriving daily data from monthly downscaled projections--For impacts to hydrology or other resources where spatial correlation is important, the simplest (and most common) way to obtain daily values is to resample a gridded daily observed data set, and for each month in the time series a single month/year from the observed dataset for rescaling all grid cells. There is a link to one dataset, which also is the observed baseline for the downscaled monthly data, available here.
The scenarios, models, and runs that have been downscaled are only those currently listed on our web site. These were the data that were available and could be successfully downscaled at the time of the downscaling activites in the summer of 2007 (BCSD-cmip3), fall of 2010 (BCCA-cmip3), and fall of 2012 (BCSD/BCCA-cmip5).
The archive offers monthly and daily climate time series projections of temperature and precipitation. There are currently no plans to downscale other variables.
CMIP5 content was made available through this archive in May, 2013. This effort applied two statistical downscaling techniques -- monthly bias-correction and spatial disaggregation (BCSD) and daily bias-correction and constructed analogs (BCCA) to a large ensemble of new climate projections released through the World Climate Research Programme (WCRP) Coupled Model Intercomparison Project phase 5 (CMIP5). The downscaled projections are being developed over the contiguous U.S.
The new CMIP5 content was developed at the same spatial and temporal specifications as the archive's downscaled CMIP3 content, thereby permitting easy comparison between the two WCRP sets of information. CMIP5 BCSD content includes >80 historical climate simulations setting up >200 future projections representing four representative concentration pathways (2.6, 4.5, 6.0 and 8.5). CMIP5 BCCA content includes >40 historical climate simulations and > 130 future projections.
BCSD-CMIP5 Climate monthly Projections are currently only available for direct download here. Please see the "Tools" section at the bottom of the "Links" page for software that my be used to subset these files once downloaded.
There are currenly no daily projections available on a global scale.
The underlying mask and elevation data are all from the NLDAS project.
GCM output will be essentially the same for the historic period (meaning 1950-1999). This is by construct, since the bias correction step applied to each GCM ensures that the statistical properties of the observed data (though NOT year to year correspondence) for 1950-1999 temperatures and precipitation are reproduced by the downscaled data (see the archive's About tab, Methodology subtab for illustration; also see the this presentation for additional illustration).
It is emphasized that the outcome of the bias-correction of a 1950-2099 climate simulation is that the portion of the simulation during the bias-identification period (1950-1999) will have 50-year monthly period statistics that match that of the gridded observed (Maurer et al. 2002). However, the 1950-1999 sequencing of climate variations will NOT match that of what was observed. This means that sub-period statistics within this 50-year historical simulation period (e.g., 1961-1990) will vary from climate simulation to simulation, even after bias-correction relative to 1950-1999 using the procedure described.
Also, the sequencing over the historical period (1950-1999) will be identical for a given model/run across scenarios. In other words, all run 1's for a model will be exactly the same for sresa1b, sresa2, and sresb1 emission paths from 1950-1999, but will be different over the period 2000-2099.
BCSD: For the coarser GCM-scale 2-degree (1-degree for CMIP5) data, the bias correction in BCSD ensures that there will be no bias relative to observations for the historical bias correction period (for summary statistics for the entire historical period). For any individual year, or any sub-period not matching the full base historical period, however, differences between the downscaled and observed statistics will be evident. During the spatial disaggregation step, as the anomalies are interpolated to each fine-scale grid cell and applied to local climatological mean values, additional bias relative to observations will exist, since there is no correction for biases relative to fine-scale observations. The only guarantee is that they match statistically at the 2-degree scale (1-degree for CMIP5).
BCCA: Since BCCA shares the same GCM-scale (1- or 2-degree) bias correction as BCSD (though applied to daily rather than monthly data), the same as above at the GCM-scale is true. The spatial interpolation of BCCA biases uses domain wide pattern-matching on a day-by-day basis, the emphasis is aimed at maintaining plausible spatial storm structures, and to let new patterns emerge as simulated by GCMs. By not being tied as strictly to the observed climatology (at either the coarse or fine scale) as BCSD, biases relative to observations at the fine scale will exist. For precipitation, these biases can be in the total precipitation, the number of rainy days, and other statistics, and will vary considerably for different regions. That the CA spatial disaggregation step introduces biases relative to observations in some regions has been noted in some studies (see, for example, Fig. 2 in the Maurer et al. (2010) paper.).
If low bias at the fine scale is necessary, a secondary bias correction can be applied after spatial disaggregation for either BCSD or BCCA. This was implemented experimentally by Maurer et al. (2010), using the same quantile-mapping technique as in BCSD and BCCA but at the fine scale, though the impact of fine scale bias correction on simulated hydrology for that study region was mixed. A simpler approach to remove local scale biases could be a simple shifting of the downscaled data by the differences in the mean values for the historical period. Ultimately, the final application of the data will drive the choice of approach for final bias correction.
The downscaled projections are on the NLDAS 1/8th degree grid (datum WGS84).
July 17, 2013: It has come to our attention that the BCCA precipitation has a dry bias over much of the contiguous U.S., and more especially over the central and eastern U.S. Please see the Errata document for more information regarding this potential issue.
April 4, 2014: On July 17, 2013 users were notified that the BCCA precipitation had a dry bias over much of the contiguous U.S., and more especially over the central and eastern U.S. This bias has been corrected with a simple bias-correction in the monthly means.
In summary, the daily BCCA precipitation for each projection was adjusted to do a better job of matching the monthly mean gridded observed values on which the downscaling was based. The basic approach was to develop smoothed (3-month running mean) monthly mean values and base the adjustment on that, to avoid abrupt discontinuities between months. This means each month isn't a perfect match, but the differences are small. The same was done for the 12 monthly means for the observed data. Monthly 'correction' ratios were derived by dividing the mean monthly observed values by the mean monthly historical BCCA projection values. These ratios were then applied to the BCCA projection's historical and future periods to produce the 'corrected' values. The period 1961-1999 was used to develop the correction ratios for CMIP3 projections. The period 1950-1999 was used for CMIP5 projections.
The corrected BCCA precipitation archive files are referred to as BCCAv2. Documentation including the Errata document will be updated soon to reflect this correction.
Please see the Errata document for more information regarding this issue.
July 8, 2014: It has come to our attention that the GCM modeling results for the BNU-ESM model have problems. So downscaled products derived from the model are no longer available through the archive. It has been left in the subsetting interface tables as a place holder, in case corrected GCM modeling results become available and downscaled products are reproduced.
Please refer to the reference here: E. P. Maurer, L. Brekke, T. Pruitt, B. Thrasher, J. Long, P. Duffy, M. Dettinger, D. Cayan, and J. Arnold, 2014: An Enhanced Archive Facilitating Climate Impacts and Adaptation Analysis. Bull. Amer. Meteor. Soc., 95, 10111019. (http://journals.ametsoc.org/doi/abs/10.1175/BAMS-D-13-00126.1)
This topic is discussed in that paper, and references are made to some other papers that go into greater detail.
The downscaled monthly (BCSD) and daily (BCCA) projections were developed in different efforts using different methodologies from monthly and daily GCM projections. They both start with coarse scale GCM data, but the processes are unrelated and were performed independently. So there's no reason that the BCCA daily values for a given projection aggregated/averaged to monthly should match the corresponding BCSD monthly projection values.
| UCRL-WEB-236256 | Last modified July 8, 2014 |