Re: warming up SWAT

[willem vervoort]

I am not sure, I probably should have a look at SWAT again, been busy
with other research for the last year or so. Check what NSKIP exactly
does, do you just skip the output? Then it would be sufficient as a
warm-up. It might mean the model does not need much of a warm-up.
Willem

On Fri, Mar 13, 2009 at 4:33 AM, lukinha <lluk...@gmail.com> wrote:
> Hello WILLEM,
>
>
> I need your help again.
>
> I am in doubt about how to warm up SWAT.
>
> I did this:
>
> - first I runned the model using interface
> - than I changed de NSKIP (3 years ) in file.cio (for ICLB =0)
> - after that I compared the values from the simulations above and nothing
> happened.
>
> The way I did is correct to warm up the model?
>
> Or I should do the change in NSKIP only  when doing the calibration
> (ICLB=2)?
>
> Could you help me in this?
>
>
> thank you in advance
>
> lukinha

--
http://blogs.usyd.edu.au/waterhydrosu/

[Jim Almendinger]

Right -- NSKIP refers to the warm-up period at the beginning of the run.  The model is running for the full time you specify in the model-set-up dialog box, but the output is summarized only for the years after NSKIP.  

A few years (maybe 1; 2 is safer; I tend to use 5) is probably sufficient to get the system hydrology to warm-up.  I don't know how the scale of your watershed and climate affect this -- perhaps larger and drier watersheds need longer warm-up periods.  (My experience is with a 700 km2 watershed, with annual rainfall about 800mm.)  

For water quality, be aware that some model parameters (I'm thinking of soil phosphorus, for example) could take considerably longer to equilibrate -- perhaps decades.  In this case, a few years of model "warm-up" are technically not sufficient, not at least for your model to reach a quasi-steady state condition (a "stationary" condition without directional change).  However, you may never notice the effect if the parameter changes so slowly that the changes in output are small over the time period of the model run.  

I'd think this would be a publishable idea and not too time-consuming to explore, if any of you out there want to tackle it.  (OK, a small publication for a limited audience -- but I'd like to see it done.)  

Cheers,

-- Jim

Dr. James E. Almendinger
St. Croix Watershed Research Station
Science Museum of Minnesota
16910 152nd St N
Marine on St. Croix, MN  55047
tel: 651-433-5953 ext 19

--
BEGIN-ANTISPAM-VOTING-LINKS
------------------------------------------------------

Teach CanIt if this mail (ID 5789932) is spam:
Spam:        https://canit.smm.org/canit/b.php?i=5789932&m=b430b1a7677a&c=s
Not spam:    https://canit.smm.org/canit/b.php?i=5789932&m=b430b1a7677a&c=n
Forget vote: https://canit.smm.org/canit/b.php?i=5789932&m=b430b1a7677a&c=f
------------------------------------------------------
END-ANTISPAM-VOTING-LINKS

[Andrew Fang]

Why is a "warm-up" period necessary? I would think if you get your
initial conditions (hydrology, water quality, and soil chemistry etc)
correct at the beginning of the model setup, the simulation results
should be acceptable over the entire period of simulation. Also, "warm-
up" can be costly if one has only a precious few years of monitoring
data to work with.

Any enlightenment will be appreciated.

Andrew Fang
Oklahoma DEQ

[Jim Almendinger]

Andrew et al. --

Well -- this is a good question, and makes me realize that I've just made some assumptions that I haven't fully checked out.

(1) For hydrology, I think it depends on where SWAT initializes the various hydrologic pools.  Do soil moisture and shallow aquifer storage start out at zero, or some estimated value?  I've never really looked.  As a user, I haven't explicitly set such values -- though perhaps there's an option I've missed.  But unless you can set all these values to known amounts -- or SWAT does a very good job of estimating initial values based on soil properties (for example) -- then I'd think a warm-up period would be useful, if not necessary.  In any case, hydrologic warm-up doesn't take much time -- a year or two may be plenty.  Again, this may vary with watershed size, land cover, and annual precipitation.  

      And, if you've only got a few years of monitoring data, couldn't you still run the data for a few years in advance of the monitoring period?  Even if you don't have climate data for that warm-up period, you could just duplicate a few years of climate data (as representative values) and apply it to the warm-up period --- you're not going to use the output from that period anyway.  It'd be useful to pick a single year of climate data and duplicate it for 10-20 years; then run the model and see when (how many years is takes) annual runoff reached a pseudo-steady state.  

(2) Soil chemistry is perhaps a slightly different animal.  If you know your soil chemistry and nutrient application rates, then I think you're quite right -- you don't really need a warm up period.  I tend to think in terms of long-term average results under pseudo-steady state conditions -- but in many (if not most) cases the soil chemistry may not really be in equilibrium with nutrient applications and removal.  Soils may still be accumulating phosphorus, for example.  A warm-up period could, in theory, alter your conclusions -- i.e., where your soil ends up with regard to nutrient content might depend on the length of warm-up period you choose.  This is a problem in theory -- it would be interesting to see how big an effect this could have. 

Cheers,

-- Jim

Dr. James E. Almendinger, Senior Scientist

St. Croix Watershed Research Station

Science Museum of Minnesota

16910  152nd St. N

Marine on St. Croix, MN  55047

tel: 651-433-5953 X 19

fax: 651-433-5924

email: din...@smm.org

web: www.smm.org/SCWRS/

[willem vervoort]

Jim et al.

as a modeller I think there is no such thing as

>if you get your
>initial conditions (hydrology, water quality, and soil chemistry etc)
>correct at the beginning of the model setup

Given the highly variable nature and complexity of catchments and the
translation into "effective parameters" in SWAT it is impossible to
get it correct. By doing a warm-up you are assured that your pools
have at least stabilised. Again that does not mean that they are
correct. I think this is inherent in complex models representing
complex nature. I can spend another hour talking about issues with
calibration and validation and representation of catchments. But I
won't bore you. Please read the the latest Kirchner 2009 paper (Water
Resources Research 45 W02429 doi:10/1029/2008WR006912) which probably
would get you started.

How long your warm-up period has to be is easily tested using some
extreme values to start and plotting some of the variables you are
interested in (for example flow or soil moisture) to see how long you
need to wait for the effect of the initial conditions to dissapate.

Willem

--
http://blogs.usyd.edu.au/waterhydrosu/

[UMUBYEYI SAFARI Prosper]

Hello everyone, Happy new year

Can anyone help me to sort out the problem of my model?

I am using SWAT CUP to calibrate and validate my model (SWAT 2012) with SUFI2 Algorithm 

I made a model of 11 years period (1999-2009) later on I realise I only have river discharge of 4 years (2005-2008), however, I did not make any change in the model.

I want to calibrate using two years and validate using another two years 

I made SWAT CUP model and run it but the thing is when I want to see the output of calibration I get the following warning message telling me that the file does not exist.

Can anyone tell me why? is it because I used less calibration data or there is something missing in my model (SWAT CUP)?

I followed the SWAT CUP user manual as well as your video, but I cannot find the output of calibration

Best regards