Question about why erosion volumes are bigger in unthresholded versus thresholded DoDs

[joewheaton]

From a user:

My question is about estimating erosion with the help of DOD. When I set a threshold (LOD), a lot of my data is deleted and the amount of erosion is very, very small compared to soil losses. But when I do not set the threshold, the erosion number becomes very large and I am confused. Do you suggest a solution? I do my experiments in a laboratory flume with an area of 6 square meters and the surface of the soil is rough with a range of 20 cm and is not smooth. Also, the pixel size is 2 mm.

I also need help to calculate the “ t ” factor in the Ucrit index formula, and please tell me about t and it’s formula. I do not know how to calculate it. I have 2 DEMs before and after a rain event.

[joewheaton]

My question is about estimating erosion with the help of DOD. When I set a threshold (LOD), a lot of my data is deleted and the amount of erosion is very, very small compared to soil losses.

The data is not "deleted", it is not included in the reported results because for those cells, your calculated changes are below that minimum level of detection threshold. This is known as "subtraction", and other methods exist.

But when I do not set the threshold, the erosion number becomes very large and I am confused.

This is the paradox of change detection with LoD. The total volume is greater when not threhsold or lower LoD is used, but so is +/- uncertainty! A higher LoD "throws away" greater volumes, but tends to lower your threshold. See: http://gcd.riverscapes.xyz/Concepts/error-propagation.html

Do you suggest a solution? I do my experiments in a laboratory flume with an area of 6 square meters and the surface of the soil is rough with a range of 20 cm and is not smooth. Also, the pixel size is 2 mm.

Well, you might consider building a spatially variable error model, but you really need to focus on calibrating the error models to the method you are using (presumably ground-based LIDAR or SFM). See Kasprak et al. (2015) for example from a flume:

Kasprak A, Wheaton JM, Ashmore PE, Hensleigh JW, Peirce S. 2015. The relationship between particle travel distance and channel morphology: Results from physical models of braided rivers. Journal of Geophysical Research: Earth Surface 120 : 55–74. DOI: 10.1002/2014JF003310

I also need help to calculate the “ t ” factor in the Ucrit index formula, and please tell me about t and it’s formula. I do not know how to calculate it. I have 2 DEMs before and after a rain event.

See both Kasprak et al. 2015 and http://gcd.riverscapes.xyz/Concepts/error-propagation.html.