[GRASS-user] Starspan/GRASS question

[Thomas Adams]

Jonathan,

I think you have the problem correct. Attached is a gif image that shows 
my 686 subbasins (polygons) in green. I need to get average values of 
temperature & precipitation for each of the polygon areas for numerous 
rasters. From the values, I then generate the time series files for each 
subbasin area (i.e., each polygon).

Thanks very much for your reply!

Regards,
Tom

Jonathan Greenberg wrote:
> So if I understand this correctly, you have some polygon surface that 
> defines the extent of your subbasin, which you want to use to query 
> and summarize the raster values falling within it (e.g. precip and 
> temp), correct?  How are you doing it currently?  The main speedup 
> that starspan gets you is that the polygon -> raster query should be 
> significantly faster than a zonalstats type analysis.  With that said 
> (and I'll need to talk to Carlos, the lead programmer of starspan, 
> about this) there might be some tricks you can use if you are 
> extracting using the same polygons and the same grids (in terms of 
> size of raster, number of samples, lines, geographic extent, etc...)  
> The idea is basically there are two processing bottlenecks: 1) 
> determining the polygon/raster intersection and 2) the i/o and actual 
> extraction of the data.  #1 could be made faster if you are always 
> using the same polys and grids, since you could (in theory) only 
> determine the intersection once...
>
> While we kick this around on our end, I'd recommend grabbing the 
> latest and greatest version of starspan, and trying it out!
>
> --j
>
> Thomas Adams wrote:
>> List:
>>
>> My apologies for the background info. leading-up to my question…
>>
>> I've seen mention of starspan previously and now I think it's time 
>> for me to learn more. Let me pose a problem to you to see if starspan 
>> would be of help. I am working on a modeling project with a couple of 
>> other people that involves the following:
>>
>> (1) downloading and decoding gridded fields of numerical weather 
>> prediction (NWP) model output
>> (2) ingesting the decoded data into GRASS to calculate basin average 
>> precipitation & temperature (separate grids) for each subbasin
>> (3) writing out all calculated basin average values for each grid to 
>> separate files (one file contains one time step of data for all 
>> subbasins) for both temperature & precipitation
>> (4) for data management reasons, the files from (3) are written to a 
>> PostgreSQL database
>> (5) once all time steps of gridded precipitation and temperature 
>> field data are written to the PostgreSQL database, another process 
>> collects the data and generates individual ascii time series files 
>> for each subbasin for both temperature & precipitation
>> (6) once (5) is completed a hydrologic model runs using the 
>> temperature & precipitation time series as input and hydrologic 
>> forecasts are generated.
>>
>> This is suppose to be a *real-time* process. The problem I am having 
>> is a matter of scale. What I did not say is that there are 12 
>> different sets of NWP output covering a period of 168 hours at 6-hour 
>> time steps for both temperature & precipitation. So, this means I 
>> must process 12*2*(168/6) = 672 grids. Also, I need the mean areal 
>> values for 686 subbasins within the domain for each of the grids.
>>
>> Steps (2-4) take about 20 seconds total for each of the grids… which 
>> is ~7.5 hours total
>> Step-5 also takes about 20 seconds for each time series file… which 
>> is ~7.5 hours total
>>
>> So, about 15 hours total. Now, I can cut this time in half by running 
>> the processing of the temperature & precipitation grids and 
>> generating their separate time series files in parallel, rather than 
>> sequentially. So, I can get to about 7 hours fairly easily — what I 
>> am shooting for is to get the processing time from 7 hours to about 3 
>> hours (or less).
>>
>> I need to more efficiently generate the many basin average time 
>> series files from the numerous grids. Can starspan help by reducing 
>> the time to calculate the the basin average values faster?
>>
>> I would also appreciate any/all suggestions on how to efficiently go 
>> from 'starspan generated basin average values' to my time series 
>> files. Realize, of course, the the individual grids are only a slice 
>> in time, so I have to track the grids and their resulting individual 
>> basin values (in time) to generate the time series files.
>>
>> To compound the problem, very soon, I need to add model grids from an 
>> additional 21 models, bringing the total from 12 to 33!
>>
>> Regards,
>> Tom
>>
>


-- 
Thomas E Adams
National Weather Service
Ohio River Forecast Center
1901 South State Route 134
Wilmington, OH 45177

EMAIL:	thomas.adams at noaa.gov

VOICE:	937-383-0528
FAX:	937-383-0033

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ohrfc.subbasins.gif
Type: image/gif
Size: 33720 bytes
Desc: not available
Url : http://lists.osgeo.org/pipermail/grass-user/attachments/20080306/b6f51252/ohrfc.subbasins-0001.gif