Might seem trivial but whence slope is moved to log10, zeros are a issue. Machine learning might not care per se but slope, like area is a multiplicative process in hydrology, so log-transform is appropriate in additive models. I have worked on the low-slope problem for TxDOT. As slope vanishes, water still flows and watersheds still retain characteristic time. This is because the gravitational forces drawn out (pun intended) but pressure differential comes to dominate. Pressure differential small when slopes are "steep."
So this is my reset but not additive as in a log offset (X + 0.0005).
Where does 0.0005 come from? TxDOT Research Report 0-6382-1 (see sections 5.2.5 and 5.2.6). In Texas practice, we are encouraging the recognition of a lower bounds, though it was easier to treat it as a small log offset as oppose to my reset shown above.
Might seem trivial but whence slope is moved to log10, zeros are a issue. Machine learning might not care per se but slope, like area is a multiplicative process in hydrology, so log-transform is appropriate in additive models. I have worked on the low-slope problem for TxDOT. As slope vanishes, water still flows and watersheds still retain characteristic time. This is because the gravitational forces drawn out (pun intended) but pressure differential comes to dominate. Pressure differential small when slopes are "steep."
So this is my reset but not additive as in a log offset (X + 0.0005).
spCOV$acc_basin_slope[spCOV$acc_basin_slope == 0] <- 0.0005
Where does 0.0005 come from? TxDOT Research Report 0-6382-1 (see sections 5.2.5 and 5.2.6). In Texas practice, we are encouraging the recognition of a lower bounds, though it was easier to treat it as a small log offset as oppose to my reset shown above.
REFERENCE (I ghost wrote numerous pieces of this and was a funded member of the team.) Cleveland, T.G., Thompson, D.B., and Li, Ming-Han, 2012, Establish effective lower bounds of watershed slope for traditional hydrologic methods: Texas Department of Transportation Research Report 0-6382-1, Texas Tech University Multidisciplinary Research in Transportation (TechMRT), https://library.ctr.utexas.edu/Presto/content/Detail.aspx?ctID=UHVibGljYXRpb25fMTE2MTA=&rID=MTU3NjA=&ssid=c2NyZWVuSURfMTQ2MDk=