RMSE on RETC

[infoleon]

Hi,

I am calculating a mean for Van Genuchten parameters, from 5 samples of the same soil using RETC.
I want to compare the RMSE from some Hydrus simulations (inverse modeling) with RMSE from RETC. I think RETC do not provide RMSE directly.

So I ask:

Should I use the sum of squares to do it?
This RMSE from weighted value the best way to compare with the RMSE provided from Hydrus output? (the Hydrus inverse modeling was made just with water content values. No pressure values)

As example, I have this:



Should I do this?

RMSE (Ret) = (1/500 * (0.05232))^0.5 = 0.01
RMSE (Cond) = (1/500 * (0.01584))^0.5 = 0.005
RMSE (All) = (1/1000 * (0.06816))^0.5 = 0.008

(500 data points input on retention data and more 500 on conductivity data.)

And use the RMSE (All)?


Thanks a lot.

[Jirka]

What you suggest makes sense for retention and conductivity data (if you use unweighted SSQ). I do not think that it make sense for "all" data, since each has different units. This is how these values are calculated:

SSQ1=0.0
SSQ2=0.0
SSQW1=0.0
SSQW2=0.0
do 84 i=1,NWC
R(i)=Y(i)-F(i)
SSQ1=SSQ1+R(i)**2 ! unweighted retention data
SSQW1=SSQW1+(R(i)*W(i))**2 ! weighted retention data
84 continue

do 88 i=NWC1,NOB
R(i)=Y(i)-F(i)
SSQ2=SSQ2+R(i)**2 ! unweighted conductivity data
SSQW2=SSQW2+(R(i)*W(i))**2 ! weighted conductivity data
88 continue

SSQ=SSQ1+SSQ2 ! unweighted all data
SSQW=SSQW1+SSQW2 ! weighted all data

I do not think that HYDRUS and RETC are using weights in the same way and thus I would not try to compare weighted SSQ.

J.

[infoleon]

Thanks a lot for the answer. It helped me a lot.