[GRASS-dev] how to reproject a raster map with absolute numbers without losing data

[Moritz Lennert]

On 01/12/16 15:13, Markus Metz wrote:
>
>
> On Thu, Dec 1, 2016 at 12:01 PM, Moritz Lennert
> <mlennert at club.worldonline.be <mailto:mlennert at club.worldonline.be>> wrote:
>>
>> On 30/11/16 09:40, Paulo van Breugel wrote:
>>>
>>>
>>>> On Tue, Nov 29, 2016 at 5:44 PM, Moritz Lennert
>>
>>
>>>>
>>>> > An area() function in r.mapcalc would be nice...
> [...]
>>
>>> |r.mapcalc "rcs = (sin(y()+$nsres/2) - sin(y()-$nsres/2)) * ($ewres *
>>> $PI/180) * float($a)^2";|
>>>
>>
>> I tested this approach as well. I.e.
>>
>> area = above formula
>> pop_density = pop / area
>>
>> reproject pop_density
>>
>> pop = pop_density * (ewres*nsres)
>>
>> It is faster than the r.in.xyz <http://r.in.xyz> approach. But it does
> not seem to be as precise.
>>
>> I still lost about 100.000 inhabitants, and more when I smooth more
> (the "loss" increases from nearest neighbor to bilinear to bicubic).
>>
>> In order to avoid precision issues, I tried with both density by m2
> and density by km2, but results are similar.
>>
>> I don't know which part of the error comes from the use of spheroid
> approximation and which part comes from the resampling at reprojection.
>>
>> But I do think that if it is important to maintain exact totals, then
> the r.to.vect - v.proj - v.out.ascii - r.in.xyz <http://r.in.xyz>
> solution seems to be more reliable.
>
> But with this approach you get spatial artefacts. Assume there are 9
> cell in source with
>
> 10 10 10
> 10 10 10
> 10 10 10
>
> they can become in target e.g.
>
> 10  0 20
> 10  0 20
> 10 10 10
>
> Total count is maintained, but some target cells might receive no source
> cell whereas other target cells might receive more than one source cell.

Right. I didn't think about that...

Moritz