search

SoilWatch / soil-erosion-watch

GEE App to explore the state of the world's (degraded) soils
MIT License
60 stars 23 forks source link

V factor should be accumulated by median #2

Closed shahab271069 closed 3 years ago

shahab271069 commented 3 years ago

Hey

I checked your code and the V factor in Rusle.Sfactor calculation should be accumulated by median.

Please check it. It should be median because the Sfactor are extremely underestimated.

Best Shahab

wouellette commented 3 years ago

Hi Shahab,

The V factor is a model of a natural phenomenon, which is the impact of annual vegetation cover on soil erosion. The more often and more dense vegetation is, the higher the L*V factor should be.

In order to truly embody the temporal dynamics of vegetation cover on soil erosion, calculating the integral of the FCover curve would probably be the most accurate way to model it, and taking the mean of that sequence may be an over-simplification (so would taking the median in that sense).

What we could do is use a trapezoidal approximation to calculate the area under the curve and better model the temporal character of the V factor, similar to what is done here for the AUC: https://gis.stackexchange.com/questions/265623/how-to-access-accuracy-of-binary-probability-in-google-earth-engine If you agree, I can create a pull request to update it.

Regards,

William

shahab271069 commented 3 years ago

Hi William

Thank you for your answer. Yes, you are right, It should be calculated by integral form. I agree, because this may provide more realistic V factor. Please create a pull request to elevate this part.

Thanks Shahab

wouellette commented 3 years ago

The pull request for review: https://github.com/SoilWatch/soil-erosion-watch/pull/3

shahab271069 commented 3 years ago

Hi William

Thank you for your pull. I checked the values, which I think are comparable with median. However, I will check the results deeply and inform you ASAP.

Regards Shahab

shahab271069 commented 3 years ago

Hi William

I checked the new version. It has improved the V factor calculation. Therefore, I want to thank you for your magnificent work. I checked it with other available datasets. If you prefer, later I can provide some validation sets of this work.

Best Shahab

wouellette commented 3 years ago

With pleasure! In the meantime, I will proceed with the merge.