Saturday, December 5, 2015

ERA Reanalysis Precipitation

Returning to the subject of normal winter snowfall patterns across the interior, Peter Bieniek of UAF has very graciously provided the results of his downscaled ERA reanalysis for the period 1979-2013.  The ERA reanalysis is a product of the highly-regarded European Centre for Medium-Range Weather Forecasts (ECMWF) and is independent of the CFS reanalysis that I looked at earlier.  What the UAF group has done is to run a higher-resolution numerical model over an Alaska-centric domain while constraining the model by the (global) ERA reanalysis; the result is a higher resolution ("downscaled") historical analysis for Alaska.  It's a great idea and should be a very useful product.

The map below shows the results of the UAF/ERA reanalysis for mean November-March precipitation over the interior, just as in the earlier map for the CFSR data (also included below).  Note that the shaded boxes represent the native resolution of the model in the case of the CFSR, but I've interpolated to a finer grid in the case of the UAF/ERA data.  The actual resolution of the UAF model runs was 20km, compared to about 30km for CFSR.




It's interesting to note the dramatic differences between the two reanalysis products, which highlights the tremendous uncertainties in the output from these models.  The UAF/ERA precipitation shows very much less sensitivity to the relatively modest terrain features of interior Alaska, but it does show a pronounced maximum on the south slopes of the western Brooks Range.  The high precipitation from Kaltag and the southern Nulato Hills south along the Yukon River is one of the best points of agreement between the models, which supports the original idea that Kaltag is one of the snowiest lowland locations in the interior.

The upper Tanana River valley is strangely moist in the UAF/ERA data, as there is little sign of the precipitation minimum that we know is observed in that area.  The chart below shows the same point-by-point comparison that I showed earlier, but including the UAF/ERA results.  Northway and Gulkana are both much too wet in the UAF/ERA output, and consequently the overall correlation to the observed values for these 8 locations is lower than for the CFSR data.  My next step will be to assess the performance of the two models in terms of the interannual variability of winter precipitation at these locations.


11 comments:

  1. Does this have a place in your analysis?

    http://www.nohrsc.noaa.gov/interactive/html/map.html?ql=station&zoom=&loc=Latitude%2CLongitude%3B+City%2CST%3B+or+Station+ID&var=ssm_depth_gc&dy=2015&dm=11&dd=30&dh=0&snap=1&o9=1&o12=1&o13=1&lbl=m&mode=pan&extents=ak&min_x=-169&min_y=51.749999999995&max_x=-129&max_y=74.249999999995&coord_x=-149&coord_y=62.999999999995&zbox_n=&zbox_s=&zbox_e=&zbox_w=&metric=0&bgvar=dem&shdvar=shading&width=800&height=450&nw=800&nh=450&h_o=0&font=0&js=1&uc=0

    Gary

    ReplyDelete
    Replies
    1. No. The NOHRSC product might be superior, because it includes correction for station precipitation observations; but it only goes back to 2002 (it would appear).

      Delete
  2. Rick Lader at UAF (currently with IARC) did his Master's Thesis looking at all reanalysis products and how they compare in Alaska. Here is a video of his thesis defense in 2014: https://vimeo.com/99756721

    ReplyDelete
    Replies
    1. Thanks Brian, I was not aware of that. The downscaled data provides a new member in the ensemble of available reanalyses.

      Delete
    2. Thanks also Brian. The download will make a good am vid during coffee waiting for the sun to appear.

      Gary

      Delete
  3. A question from Iceland: Actual precipitation=measured precipitation? or is (measurement) wind loss accounted for in the above figures? If not the computers might be "right" with their considerably higher precipitation ammounts during the snow season. This is the case in parts of Iceland.

    ReplyDelete
    Replies
    1. Trausti,

      Thanks for the question. I used the reported/measured liquid-equivalent precipitation for "actual" in the chart; there is no correction for potential wind losses. You raise an interesting point as we would expect there to be some loss in windy locations; I'm not sure how significant that would be, but I'll try to learn more. I will say that some interior locations like Fairbanks have relatively little wind and I would expect wind loss to be quite minimal.

      Delete
    2. I meant to add - I would be interested to learn more about the wind loss estimates from Iceland.

      Delete
    3. Lots of experiments:

      http://www.wmo.int/pages/prog/www/IMOP/intercomparisons/SPICE/SPICE.html

      Iceland:

      http://www.precipitation-intensity.it/pdf/arctic_iceland.pdf

      Gary

      Delete
    4. Interesting stuff, Gary. Thanks.

      Delete
    5. I threw it out for the bibliography.

      Measurement error and bias exist, the question is how to deal with it and the resulting noise and best relate the observed data to the reanalysis. Perhaps a correction factor?

      Hopefully the stations use the same equipment and are situated properly.

      As an aside, walk across a flat parking lot covered with snow and note the surface variability due to ? Now expect a gauge to capture the average depth of that field. That's Nature for you.

      Gary

      Delete