User's Guide to the PSD Model Evaluation Web Site for the New England High Resolution Temperature Program
The PSD web site
contains the description of PSD's contribution to NEHRTP. From there one can
click on Model Verification or go directly to
to view the model verification web site.
This web site contains model/observation comparisons for 8 wind profiler sites and 8 NWS METAR surface meteorology sites located in New England. The 8 NWS surface sites are used by the energy industry in forecasting energy loads, and are referred to as "Energy" sites on the web site. The models shown include present NCEP operational models as well as research or developmental models being run at NCEP, FSL, and NSSL. Each model is evaluated over a 48 hour forecast period. On all diagrams time is plotted using the meteorological convention of time increasing to the left.
Buttons or tabs on the left or top of the plotting area allow the user to select the time, date, location, model, and observational parameters being compared to the models:
Time and Location
On the top left of the plotting area is a box for selecting the model cycle
initialization (00 UTC or 12 UTC). All models are run for both cycles,
with two exceptions
- the ensemble models shown for the energy sites, which are
available only for the 12 UTC cycle, and
- the NMM, which is only run at 18 UTC.
The default for the web page is the current date. To select any previous date from the start of the field program on 1 July 2003, click on the date in the calendar box at the bottom left. Arrows next to the month at the top of the calendar box changes the box to previous or later months.
Also on the left of the plot are buttons to choose between the 8 Profiler Sites or the 8 Energy Sites. After making this selection, the 8 available sites are listed in the drop-down box immediately below. Clicking on any of these site names changes the plots to that particular site. One site, Concord NH, is common to both the Profiler and Energy Sites.
Models and Observations
Additional tabs at the top of the plots allow the user to select the models and types of observations for comparison. The choices for these depend on whether
the Profiler Sites or Energy Sites buttons have been selected.
If the Profiler Site button has been selected, two further sets of choices for models and observations become available, and are displayed at the top of the plot.
The first choice is for the model, and there are 7 tabs that allow the user to
select among the various deterministic models available: GFS, Eta-12,
NMM-8, RUC-10, RUC-20, WRF-10 and WRF-20. The first three of these models
are run at NCEP, the later four are provided by FSL. Model predictions are displayed as soon as they are available, and this is typically for the entire 48 h forecast period, although for the RUC and WRF models part of the 48 h model forecast period may appear first, with later hours being filled in gradually. The Eta-12, NMM-8, and RUC-20 model predictions are available at hourly increments. The GFS, RUC-10, WRF-10, and WRF-20 are available at 3 hourly increments. All model data is shown at the model native levels.
The second set of choices is for the type of observation available at the
wind profiler sites. The three choices are SNR/Winds, RASS/Winds,
and Sfc Met.
For SNR/Winds, two time-height cross-section panels are displayed, the top one showing the wind profiler observations and the lower panel showing the model predictions. The wind profiler data in this plot has a vertical resolution of 60m in the lowest 500m, and
100m above (except for Pease, which has a fixed 60m resolution). The color
background shows the profiler Signal-to-Noise ratio. Moderate values of
SNR can indicate the depth of the convective PBL, and large values
indicate the presence of precipitation. Profiler data are hour averages,
and are added to the plot on an hourly basis as they become available,
typically a few minutes after the end of each hour averaging period.
When SNR/Winds is selected, the model plots also show the predicted depth of the PBL, shown as black dots, overlaid on the wind barbs.
For RASS/Winds, the profiler wind data is again shown, but with a color background showing the RASS virtual temperatures. The vertical scale shown in these plots is reduced from 4km to 1.5km, due to the more limited vertical range of the RASS temperature measurements. Virtual temperatures are computed for the model output and are displayed on the lower panel.
Surface meteorological observations are also taken at the wind profiler sites,
and can be viewed by clicking on the Sfc Met tab at the top of the
plot. The observations at the wind profiler sites are 10m speed and direction,
2m T and Td, pressure and precipitation, and solar and net radiation.
Observed values are plotted at 5 min intervals and can appear as almost
continuous lines, while the model predicted values are shown as larger open symbols at either 1 or 3-h intervals.
The Energy site comparisons are viewed by clicking on the button at the left of the plotting area, and then selecting a site from the drop-down-list. In this case only two tabs are then shown at the top of the plotting area, which allow one to choose between the deterministic or ensemble models.
If the Deterministic tab is selected, a four panel set of spaghetti-type plots is displayed, showing predictions of each of the 7 deterministic models, in addition to those from the GFS-MOS, Eta-MOS, and NGM-MOS. These later three MOS products start six hours into the 48 hour forecast period displayed. Also, since the NMM forecast cycle begins at 18 UTC, it is only included in the 00 UTC model initialization cycle plots, and it ends 6 hours before the end of the 48 hours shown on the plots.
If the Ensemble tab is selected, another set of spaghetti-type
panels is shown. On these the various curves indicate the predictions
for 4 ensemble products provided by NSSL:
For comparison, the GFS-MOS, Eta-MOS, and NGM-MOS are also shown. The NSSL
ensemble predictions are available only for the 12 UTC initialization cycle.
Therefore for the 00 UTC cycle, only the 3 MOS products are compared to the
- the mean of a 23 member ensemble, generated from all of the
independent 32-km ensemble members produced at NCEP for their various
developmental ensemble products,
- the bias-corrected 23 member ensemble mean,
- a 30 member ensemble mean, generated by adding 7 additional
OAR (FSL + NSSL) members, and
- the bias-corrected mean of the 30 member ensemble.