Downscaling climatic maps

[AliceUK]

I want to downscale some climatic maps accurately for the 70's and
80's-how does one go about it?
Thanks

Alice

[Kevin Buffington]

Here's the method I've used, maybe someone else has a better/more
accurate way...

--In ArcMap, create a grid of points such that there is one point in
the center of each pixel of your climate maps.
--Use the Sample tool in Spatial Analysis to assign each point the
value of the pixel.
--Run a Spline Interpolation through those points--you can choose a
new grid size.

My study area is very small, so this works pretty quickly. I
downscaled 1km pixels to 30m, if you have much larger pixels, say
10km, then a more complex method that takes into account other factors
like terrain would be needed.

I'd appreciate any comments on my methodology

Kevin

[Julian Ramirez]

Kevin and Alice,

The method you propose is not actual downscaling, but smoothing of surfaces. Downscaling would require a more strong statistical procedure. Moreover, spline interpolation in ArcGIS is fairly basic and does not allow to use co-variates, for example.

You can fit smoothed splines with co-variates. Google "ANUSPLIN", then you will find something useful. But use whatever is useful for your objectives.

Julian

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To post to this group, send email to max...@googlegroups.com.
To unsubscribe from this group, send email to maxent+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/maxent?hl=en.

[ah3881]

What did was, get my low resolution rasters from my second (i.e. 2100)
timeframe, and with the same dataset (so same resolution-i.e. IMAGE)
get the current timeframe, 2000.

Using raster calculator I minus current from future to find the
temperature difference.

I then turned the calculation into point format, and then Krieged the
points to the cell size my preferred size.

I then add this to the appropriate parameter (i.e maximum
precipitation) in the current timeframe with my preferred resolution.

I repeat this for temperature, and add it to the appropriate current
temperature bioclims

How would you rank my method?

Thanks so much!

Alice

On Jan 22, 7:59 pm, Julian Ramirez <dawnpatrolmusta...@gmail.com>
wrote:

> Kevin and Alice,
>
> The method you propose is not actual downscaling, but smoothing of surfaces.
> Downscaling would require a more strong statistical procedure. Moreover,
> spline interpolation in ArcGIS is fairly basic and does not allow to use
> co-variates, for example.
>
> You can fit smoothed splines with co-variates. Google "ANUSPLIN", then you
> will find something useful. But use whatever is useful for your objectives.
>
> Julian
>

> > maxent+un...@googlegroups.com<maxent%2Bunsu...@googlegroups.com>

[Julian Ramirez]

That's the so-called "delta method", which should work fine, but you might be aware of the assumptions it makes:

1. There's no micro-scale variability in changes (anomalies). This might be true in some cases (i.e. very homogeneous landscapes such as the Sahara), but definitely not accurate in heterogeneous landscapes (i.e. the Andes).

2. Relationships between variables in present climates do not change significantly towards the future. Forecasting these changes is highly uncertain even with more climatologically robust methods (i.e. RCM)

You should also take into account the differences in baselines, especially for precipitation, which could lead to negative values in your future datasets. I suggest you to use relative differences for calculating future precipitations after interpolating. Using absolute differences for temperature should work ok though.

Best,

Julian

To unsubscribe from this group, send email to maxent+un...@googlegroups.com.

For more options, visit this group at http://groups.google.com/group/maxent?hl=en.

--
Julian Ramirez-Villegas
Research Assistant
International Centre for Tropical Agriculture (CIAT)
Cali, Colombia
http://gisweb.ciat.cgiar.org/dapablogs/