20010314: Radar Composites

Jerrold,

I created the program to allow for a user specified time and time range.
The plots that are being created on the web pages are using the
current system time (but can be set to a specific time as well).
The time range I am currently using is 30 minutes ending at the time given.

The product I am generating is a gridded data set where all the 
individual images are projected to the grid domain. The user can specify
the grid domain, projection and number of grid points.

The program runs through all the NEXRAD stations, determines the most
recent product for the station within the time range and determines
if the NIDS product is within the composite domain.

Each gridpoint within the product domain is sampled for the actual data value
(eg the data value represented by the bin) so that clear air and precip mode
values of dDZ can be compared directly as reflectivity values. When a grid
point is stored, it is checked to see if there is already a value at that
point (eg more than 1 radar covers that area). Currently, I am keeping
the maximum value of the data points. Since the dBZ generally decreases
with altitude, this is a first order approximation of which radar is
closer to the grid point.

I initialize the entire grid dataset to missing values to start.
When a value of ND is stored from the radar product into the grid,
this then detects that there was data for the point- rather than
no data coverage for the area. So, I could display the value of
no data coverage in the plot as well. 

It takes about 1 minute to produce the CONUS composite.
Currently, I am creating a new grid mosaic every 5 minutes.

Compositing radial velocity is possible, however, a single NIDS
product is a 1-D (toward or away) data value. To actually compute
horizontal wind, you need an area of overlap between 2 radars
with an assumption of the vertical component (this is generally
better done with the doppler profiler array, or dual-doppler
arrays). One thing to consider is that the NIDS products
are not CAPPI displays, they are conical volumes. That means
that the N0V product is increasing with altitude as radial
distance increases. Without creating a horizontal slice though
the 3-D volume of tilts, the winds will be at different altitudes, 
and the region of overlap between radars will be at 2 different levels
(as I mentioned above in the dBZ comment). So, I wouldn't really
consider a wind product too valuable unless the data is sliced
to a constant level. At this point, we don't have the data from
all tilts. This is a much better problem for Level II radar data
and vis-ad probably.

By creating a gridded data set, I operate on the data with other
functions, suchg as the RUC freezing level data grids to
compute the precipitation type display, to filter (smooth) ground
clutter AP, and compute indicies such as the "interesting weather"
in the area index that I have for selecting the 5 floaters on the page.

I'll talk to Tom W. this week to see if he has more thoughts, questions.

Steve Chiswell
Unidata User Support



On Wed, 14 Mar 2001, Jerrold Robaidek wrote:

> 
> 
> Hello,
> 
> My name is Jerrold Robaidek, and I am responsible for the data in the
> University of Wisconsin SSEC Data Center.  I noticed that you are
> generating radar composites from the Level III Nexrad products
> distributed via the NOAAport broadcast. 
> http://motherlode.ucar.edu/unidata/images/nids/gdradr.html
> I was wondering if you could tell me how you generate them ... i.e. 
> what time range does each composite picture contain?  How do you select
> which stations reflections to keep and which ones you reject.   
> 
> The reason I ask, is that I have run into problems making the
> composites; often one site will not be updated when I generate the
> composite and can give a misleading representation of the current radar
> activity.  Also when a Station is not reporting at all, do you indicate
> it some how?  Otherwise it may appear as a clear slot when it may not
> be.  
> 
> And finally, have you looked into compositing radial velocity?  Is this
> even possible.  It seems that a composite of radar wind vectors would be
> useful, but I can't think of a way to do it.  
> 
> Here are some early prototypes of the composites that I and Rick Kohrs
> are working on:
> 
> 
> I've mentioned some of my questions to Tom Whittaker, and he may bring
> them up to you.
> 
> Thanks in advance for any information that you can provide.
> 
> Jerry
> 
> -- 
> Jerrold Robaidek                        Email:  address@hidden
> SSEC Data Center                        Phone: (608) 262-6025
> University of Wisconsin                 Fax: (608) 263-6738
> Madison, Wisconsin
>