For a three layer soil model, I would like to define each layer based on texture. Using eight texture classes, this leads to 8^3 soil profiles (many more if layer depth is allowed to vary). Is there a simpler way to make the same classification? For instance, in the .rvh, is it possible to forego the SOIL_PROFILE specification and instead directly define the classification and depths of SOIL_LAYER1, SOIL_LAYER2, and SOIL_LAYER3? Are there any recommended best practices for characterising soil in RAVEN?
Thanks,
Kai
Characterising soil profiles
-
rchlumsk
- Posts: 156
- Joined: Fri Sep 16, 2016 3:18 pm
Re: Characterising soil profiles
Hi Kai
In terms of defining the soil profile based on texture, if I am understanding you correctly then what you likely want to do is to keep the three soil classes defined, which simply sets up 3 layers in the model. From there, you can define the 8 different soil profiles based on the texture of the soils, which would all have different soil parameters to reflect the differences in the soil textures. If you mean to create 24 different soil profiles, based on the various combinations of 8 soil textures and 3 soil layers/classes, this is also doable but would likely become too cumbersome.
The Raven manual for v2.8 on page 6 has a discussion on the creation of soil profiles, although there are no strict rules on how these can be configured. My recommendation, however, would be to amalgamate some of the soil textures into fewer classes to create fewer profiles, otherwise the parameter assignment and calibration becomes parameterized and tedious to calibrate. Typically the soils are characterized on a larger scale, i.e. based on the dominant soil types in a larger watershed area, which leads to a handful of soil profiles and is more manageable in the model. The answer may be to increase the scale of your soil characterization and ignore some of the smaller variations in soils in your area, although that is up to you to determine in the model development.
For calibration specifically, an option to reduce the number of parameters would be to create tied parameters in the calibration, i.e. parameters that are dependent on others, for example, infiltration rate B that is X*infiltration rate A. This would reduce the number of effective parameters, but this may be further down the line from what you are considering now.
In short the soil class definition defines the three layers, which are basically the soil storage units in your Raven model. From there it is up to you to create soil profiles, which can have one or many more profiles to reflect the variation in soil types in your landscape, which includes variations in depths of each layer. As mentioned, the recommendation is likely to merge the soil textures into fewer groupings and/or look at larger scale soil variability to reduce the number of profiles you need.
Hope that helps and please post back if you have further questions or if that was unclear.
Cheers,
Rob
In terms of defining the soil profile based on texture, if I am understanding you correctly then what you likely want to do is to keep the three soil classes defined, which simply sets up 3 layers in the model. From there, you can define the 8 different soil profiles based on the texture of the soils, which would all have different soil parameters to reflect the differences in the soil textures. If you mean to create 24 different soil profiles, based on the various combinations of 8 soil textures and 3 soil layers/classes, this is also doable but would likely become too cumbersome.
The Raven manual for v2.8 on page 6 has a discussion on the creation of soil profiles, although there are no strict rules on how these can be configured. My recommendation, however, would be to amalgamate some of the soil textures into fewer classes to create fewer profiles, otherwise the parameter assignment and calibration becomes parameterized and tedious to calibrate. Typically the soils are characterized on a larger scale, i.e. based on the dominant soil types in a larger watershed area, which leads to a handful of soil profiles and is more manageable in the model. The answer may be to increase the scale of your soil characterization and ignore some of the smaller variations in soils in your area, although that is up to you to determine in the model development.
For calibration specifically, an option to reduce the number of parameters would be to create tied parameters in the calibration, i.e. parameters that are dependent on others, for example, infiltration rate B that is X*infiltration rate A. This would reduce the number of effective parameters, but this may be further down the line from what you are considering now.
In short the soil class definition defines the three layers, which are basically the soil storage units in your Raven model. From there it is up to you to create soil profiles, which can have one or many more profiles to reflect the variation in soil types in your landscape, which includes variations in depths of each layer. As mentioned, the recommendation is likely to merge the soil textures into fewer groupings and/or look at larger scale soil variability to reduce the number of profiles you need.
Hope that helps and please post back if you have further questions or if that was unclear.
Cheers,
Rob
Robert Chlumsky
rchlumsk@uwaterloo.ca
rchlumsk@uwaterloo.ca