[GRASS-user] Interpolating rainfall data across area from points

[Thomas Adams]

Hello all,

I think there is sufficient evidence to recommend using elevation as a
secondary variable for spatial interpolation of precipitation using, for
instance, Universal Kriging or spatial drift, or a method such as PRISM in
the USA (http://www.prism.oregonstate.edu/) -- PRISM estimates are used
widely in the US by the National Weather Service and others. See:

ftp://140.115.123.47/download/%A6h%C5%DC%B6q%A6a%BD%E8%B2%CE%ADp%BE%C7/01_Geostatistical%20approaches%20for%20incorporating%20elevation%20into%20the%20spatial%20interpolation%20of%20rainfall.pdf

https://www.sciencedirect.com/science/article/pii/S0022169418303548

https://www.hydrol-earth-syst-sci.net/10/197/2006/hess-10-197-2006.pdf

But, there is this, e.g., which supports Micha's statement:

ftp://140.115.123.47/download/%A6h%C5%DC%B6q%A6a%BD%E8%B2%CE%ADp%BE%C7/01_Geostatistical%20approaches%20for%20incorporating%20elevation%20into%20the%20spatial%20interpolation%20of%20rainfall.pdf

Best,
Tom

On Mon, Sep 24, 2018 at 2:19 PM Micha Silver <tsvibar at gmail.com> wrote:

> Hi Rich
>
>
> On 09/24/2018 06:32 PM, Rich Shepard wrote:
>
>   I want to determine whether GRASS or R is best suited to
> interpolating/extrapolating annual mean precipitation data from 58
> reporting
> stations (unevenly distributed within the county) across the county. Some
> flavor of kriging would be applied to these data to illustrate a general
> long-term pattern of rainfall.
>
>   Some years ago there was both a static display of a chemical constituent
> in a river reach, a "heat map", (and an automation of temporal changes, if
> I
> correctly recall) and I'm not finding this in the web site galleries.
>
>   While elevation could be included as an explanatory variable using
> regression kriging my purpose is to illustrate county-wide mean annual
> rainfall
> distribution over a 13 year period, not to interpolate values for
> specific,
> unsampled locations.
>
>   Please provide some thoughts on the work flow to do this within GRASS.
> I'm
> digging into the gstat docs to get a sense of how to do this within R (and
> I
> have the rgrass7 package working well; it imported the GRASS county
> boundary
> map which I converted to a SpatialPolygonDataFrame.)
>
>
> The guidelines that I follow include:
>
>    - Rainfall interpolation (of any kind) should be done only for long
>    time periods = at least a full season. since you are looking at 13 years of
>    data then this requirement is fulfilled.
>    - The rules of thumb for "how many points" for kriging interpolation
>    usually says > 30. So again you are fine.
>    - The GRASS modules offer only ordinary kriging AFAIK, which might be
>    appropriate in this case. But if you want to use Kriging with External
>    Drift with the elevation as the secondary "trend" variable, I think you'll
>    need to go with with R. Having said that, I think that "the jury is still
>    out" on whether elevation improves the interpolation or not.
>
> Regards, Micha
>
>
> Rich
>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-user
>
>
> --
> Micha Silver
> Ben Gurion Univ.
> Sde Boker, Remote Sensing Lab
> cell: +972-523-665918
>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-user



-- 
Thomas E Adams, III
1724 Sage Lane
Blacksburg, VA 24060
tea3rd at gmail.com (personal)
tea at terrapredictions.org (work)

1 (513) 739-9512 (cell)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20180925/e792c86c/attachment.html>