[Qgis-user] nodata and raster calc

[Nicolas Cadieux]

Hi,
In 99 % of what I do, treating nodata in any other way than a data mask would be severely detrimental to my results.  This would include OR and AND. (The scientific equivalence of “Do you want it eat pizza or ‘I don’t know’? or Do you want to eat pizza or nothing?” are two different valid questions.

Data mask should be the default for simplicity with user being able to select other behaviours implemented.
Nicolas

> Le 11 févr. 2019 à 05:02, <st_kiefer at web.de> <st_kiefer at web.de> a écrit :
> 
> In fact I would always expect nodata. nodata is what it says: no data. How can one calculate or generate new values with one unknown term. Especially in the case of boolean.
> Lets take an example from real life. You have the information of wet or not wet area. In the case of wet you have to reduce irrigation if not wet you must increase irrigation. In either cases you make a mistake, if you do the opposite. Without knowledge of the inputdata (and respecting the scale), someone getting the result could do a big mistake. Particularly if the action is highly cost intensive. Nodata need special treatment. One must be aware of the lack of data, and a model has rather to fail than false simulate the existing data. What if you process hundreds of raster sets and get always valid results. If some don't like nodata, he has to actively process the data in one way, either simulate valid data (and of course knows that he has to communicate that fact) or collect more precise data. Anything else is scientifically bad behaviour (to use a buzzword: fake data).
> 
> Cheers.
> 
> Stefan
> 
>> Havard Tveite <havard.tveite at nmbu.no> hat am 11. Februar 2019 um 10:20 geschrieben:
>> 
>> 
>> nodata + 1 = nodata
>> nodata * 1 = nodata
>> nodata AND 1 = nodata
>> nodata OR 1 = 1
>> 
>> nodata + 0 = nodata
>> nodata * 0 = nodata
>> nodata AND 0 = 0
>> nodata OR 0 = nodata
>> 
>> Håvard
>> 
>>> On 11. feb. 2019 09:47, st_kiefer at web.de wrote:
>>> I absolutely agree.
>>> 
>>> For the Discussion. nodata is a valuable information, and is getting lost when dealing as desired value. Although the OR part needs more Discussion. For consistency that has to be nodata as well.  What about nodata OR 0? In my opinion nodata always needs a discrete treatment to avoid generating fallcious information.
>>> 
>>> Cheers
>>> 
>>> Stefan
>>> 
>>>> Havard Tveite <havard.tveite at nmbu.no> hat am 11. Februar 2019 um 09:38 geschrieben:
>>>> 
>>>> 
>>>> nodata + 1 = nodata
>>>> nodata * 1 = nodata
>>>> nodata AND 1 = nodata
>>>> nodata OR 1 = 1
>>>> 
>>>> Håvard
>>>> 
>>>>> On 11. feb. 2019 03:11, Nyall Dawson wrote:
>>>>> Hey all,
>>>>> 
>>>>> I'm seeking some feedback on whether our current handling of nodata
>>>>> pixels in the raster calculator is correct, or (my suspicion) annoying
>>>>> and limiting.
>>>>> 
>>>>> Current