BLIPMAP Model Notes

 BLIPMAP = Boundary Layer Information Predictions Map
Updated 5 Aug 2006


DrJack sez:
      This page gives some guidelines on when model predictions are particularly likely to be in error and is intended to aid those who have little knowledge of what meteorological models can or cannot do well.  While BLIPMAPs provide an "intelligent machine" forecast, human evaluation of BLIPMAPs together with other information should produce a superior forecast - but of course that depends on the correctness of the "added value" that the human provides.  One way that "human intelligence" comes into play is through knowledge of model strengths and weakness, particularly by recognizing when the latter are making a model forecast inaccurate. 
      At present the notes are very rudimentary, but better something than nothing.  Generally notes result either from my noticing something about a particular forecast or because I am asked a question, so they are not logically structured. 

Note #1:  Remember folks, this is a model not a crystal ball.  The atmospheric system is complex and detailed but the description of that system provided by model approximations is relatively crude.  More detailed information on numerical weather models and their errors can be found at the How Does a Meteorological Model Work ? webpage.

General Prediction Accuracy:  To give an overall perspective, forecast accuracy of the many parameters predicted by a meteorological model can be generally ordered, from most accurate to least accurate, as:  (1) Winds,  (2) Thermal parameters,  (3) Moisture parameters,  (4) Cloud parameters,  (5) Rainfall. 

Surface Type:  All soaring pilots know that thermal strengths vary greatly with the land surface type, from forests to vegetated fields to bare soil/rock to irrigated crops and etc.  Meteorological models try to incorporate the effect of difference surfaces, but must do so in a very crude manner by estimating the "average" surface type over the grid area, which for NAM is 12x12km and 20x20km for RAP.  When you consider the wide variety of surfaces over such an area you can appreciate how inaccurate such an estimate can be.  Moreover, surface type determinations are usually made on a much coarser scale and such determinations are usually based upon satellite-based estimates of surface type not on actual inspection of the surface itself.  Seasonal adjustments are made using a monthly database of vegetative fraction but this considers only a limited number of seasonal effects (the effect of snow cover is also included in the model).  This note came about because I was asked if the model takes account of the flooding of the rice fields in California's Central Valley around this time of year, which obviously has a big impact on the thermal strengths there - and the answer is that although I cannot actually examine the monthly database, the ratio of specific to latent surface heat fluxes forecast by the model indicates that the model surface moisture is much drier than the actual surface, so thermal strengths are being over-predicted.  I have provided maps of surface type variation over the different RAP regions at the regional grid orientation webpage.

Rainfall and Soil Moisture:  Soil moisture greatly affects thermal predictions since solar energy which goes into evaporating surface moisture is not available to heat the surface.  All good atmospheric models include many soil moisture processes including vertical percolation of rain into the ground, ground runoff into adjacent grid cells, and of course evaporation into the atmosphere.  But these contributions can only be estimated crudely since including all the complexities of soil hydrology would require calculations as involved as those of the atmosphere itself!.  One significant problem is that this is not a parameter which is ever verified against actual data, so there is nothing to correct any model biases.  Based on limited reports, I've gotten the impression that the model tends to greatly underpredict soil moisture when there has been a heavy rain.
      Above all the soil moisture is driven by the amount of rainfall that the model predicts will occur - so if the actual rainfall is significantly more or less than that predicted then actual soaring conditions will be poorer or better, respectively, than forecast by the model.  Unfortunately rainfall is the most poorly predicted of all parameters (partly due to its intermittent on/off nature), so predicting rainfall influences on soaring conditions is very iffy.  Rainfall rates are not given in the BLIPMAP forecasts (!) but can be obtained from standard model reports such as at the "Maps of standard meteorological parameters from GSD RAP forecast products webpage" link on the regional BLIPMAP page - note in particular the 12 hr accumulated precipitation prediction there, which could, in theory, be checked against observed values.  You can't check on the predicted (or observed) soil moisture itself, but another clue can be obtained by comparing observed vs predicted surface dew point temperatures (though much more goes into that than just ground water evaporation).  That comparison is also be useful in determining whether the model is likely to over- or under-predict BL cloud formation.  A good source for the surface dew point forecast is the forecast meteogram available at the link "Time-series of standard RAP meteorological parameters" on the regional BLIPMAP page.  That link also provides other parameters which are useful for comparing to observations to assess "how the model is doing", such as its display of "Cloud Base Height" (though that is reported in pressure rather than MSL height).

Clouds:  Predicting clouds is always a challenge, with difficulties increasing as clouds decrease in size, vertically or horizontally, since clouds smaller than the grid spacing cannot be directly predicted by fundamental equations.  Some ad hoc supplementary equations are used to estimate effects such as reduction of surface heating by unresolved clouds, but these cannot be very accurate.  The RAP model can only predict that a cube the size of a model grid cell is either completely clear or completely filled with condensed water, but the NAM model does try to provide for partial cloudiness in a grid cell. For RAP, the cloud top height and cloud base height predictions are available on the GSD RAP forecast products page.
      A BLIPMAP cloud issue, separate from model prediction issues, is that its thermal strength forecasts do not include the contribution of condensation heating aloft produced by cloud formation (sometimes elegantly referred to as "cloudsuck") - so expect stronger than predicted thermals to occur below clouds when they are present in the BL.

Thin Cloud Layers:  If thin cloud layers are present, then BLIPMAP predictions are particularly suspect.  Models often fail to forecast cirrus and other thin cloud layers, largely due to the finite thickness of the model grid layers - and since grid vertical spacing increases with height, this problem exacerbates at upper levels.  Unfortunately it does not take a very thick cloud layer to greatly reduce the solar radiation reaching the ground, so thermal strengths and heights are often over-predicted when such layers are present.  To allow for such clouds when they are not forecast by the model is a frequent challenge.  Users can use satellite photos to spot the layers and anticipate their movement based upon upper level winds - but this works only for existing layers which are transported, not for those which develop later.  One useful reality check is the model-predicted surface heating parameter - if clouds are predicted patches of decreased surface heating will appear, so the existence of actual cloud layers without such surface heating decreases is a warning of a model's failure to predict that cloudiness.

Low Visibility:  The models do not adequately allow for the reduction in surface solar radiation resulting from large "aerosol" concentrations such as dust or smoke.  For example, reductions in visibility and surface solar radiation such as caused by a recent transport of Asian dust to the West Coast will be completely missed.  And the model knows nothing about forest fires!  Also, my impression is that the models do not adequately allow for the reduction in surface solar radiation created the haze associated with pollution or coastal conditions (the model also knows nothing about pollution or atmospheric sea salt, both of which cause water to condense into haze droplets).  Therefore, if visibility is poor, expect thermal strengths to be lower than predicted.



Link to the BLIPMAPs for all regions