Calibration Overview

Section for FAQs regarding i-Tree Hydro

Moderators: i-Tree Support, i-Tree Team

Post Reply
rcoville
i-Tree Team
Posts: 64
Joined: Fri May 08, 2015 9:29 am
Location: Syracuse, NY

Calibration Overview

Post by rcoville » Tue Aug 04, 2015 7:59 am

General questions about calibration in i-Tree Hydro are hopefully answered by the following. More specific questions about troubleshooting auto-calibration can be addressed case by case. Additional information about this subject can be found in:
> What's the role of calibration in i-Tree Hydro?
In short:
Calibration is about getting an effective hydrological parameter set for a project. This is done automatically by adjusting parameters and running the model many times until a 'best fit' outcome is reached where observed base case streamflow and predicted base case streamflow match most closely. Auto-calibration requires observed streamflow which corresponds directly with your project area and inputs, but the model does not otherwise need observed streamflow to run.
Calibration is a necessary step for a small group of Hydro users, including some researchers and those required by specific regulations to match the base case of their model to observed stream flows. For most users, especially in non-watershed areas, the default Hydrological Parameter set is recommended as it simplifies the modeling process while providing effective comparative results. Knowledge of the site (i.e. soil testing) could dictate the adjustment of several of the suggested default parameters, but users should test and compare these changes in a methodical manner.

A full explanation:
i-Tree Hydro is designed to assess the effects that changes in vegetation and impervious cover have on water flows and qualities in a given area. To make it a relatively easy model to use, we have simplified some of the complexities of hydrology (for example, by providing a default hydrological parameter set for Step 3 of i-Tree Hydro). As a result, Hydro is more accessible and better suited for use by diverse audiences who want to assess various scenarios comparatively and simulate the hydrological effects of vegetation and impervious cover. Also due to such simplifications, Hydro is not as well suited for highly accurate streamflow predictions (which is not what it was designed for in the first place).

In some instances, due to strict research criteria or regulatory requirements, it's necessary for a model's predicted base case flow to closely match observed base case flow. It is possible to increase the accuracy of predictions by reducing assumptions and fine-tuning hydrological parameters, thus coming closer to a match between predicted and observed base case results. To enable this - enhancing the functionality of i-Tree Hydro without sacrificing usability - we've included the option to manually edit or auto-calibrate complex hydrological parameters. We recognize some of those parameters are very technical and may be out of the comfort zone of some users, and so we offer a default parameter set which serves well when using the model for comparative assessments.

Calibration is designed to find a parameter set which gets predicted flow as close to observed flow as possible. To do this, observed flow data is required. Auto-calibration, based on PEST parameter estimation software, is designed to adjust parameters and test results until a good fit between predicted and observed base case results are found. The resulting parameter set may include values for certain parameters that you would not find in the field, and that is why scientists refer to such parameter sets as effective parameter sets: the parameter values are not informed by actual measurements, they're informed by finding what values align predictions with provided observations. Calibration reduces the assumptions of the model, which theoretically means the analysis done in i-Tree Hydro are then that much finer approximations of reality. Again, this higher-order approximation is not necessary for most assessments comparing scenarios and expressing the impact of changes in vegetation and impervious cover on water quantities and qualities.

Generally, environmental modeling requiring higher-order approximations is expected to require much more in terms of data inputs, time, and technical expertise. We're striving to enable more technical users to fine-tune this model to suit their needs while also keeping this robust environmental model accessible for use by diverse audiences interested in water resources. We continue to work on improving this balance as we develop i-Tree, putting Forest Service science into people's hands.


> What does uncalibrated versus calibrated mean for the model's accuracy?
If the model is uncalibrated:
Model results should be considered qualitative. This qualitative assessment i-Tree Hydro provides does serve most purposes i-Tree Hydro is designed for.

If the model is calibrated:
It is not appropriate to expect Hydro to quantitatively predict how many cubic meters of water will flow in a stream if a designed scenario becomes reality.
It is appropriate to use and expect Hydro to quantitatively predict how a designed scenario would change how many cubic meters of water will flow in a stream if it becomes reality.
Hydro is designed for measuring the effects of changes in vegetation and impervious cover on water quantity and quality.


> What type of project areas can be calibrated, and how?
In terms of project areas, it is recommended that calibration only be done if modeling a watershed (which would be direct calibration) or a municipal area which largely falls within a watershed (which would be indirect calibration, if any). To explain more about direct vs. indirect calibration:

Watershed areas can use direct calibration. This can be done using the auto-calibration feature in the Hydrological Parmaters step of i-Tree Hydro if observed stream flow data is available and formatted for use in i-Tree Hydro. This can also be done manually based on local knowledge of the project area's average soil & hydrological properties.

Certain non-watershed areas can use indirect calibration if the project area is likely to have hydrological parameters very similar to a local drainage area (e.g. a municipal area which largely falls within a watershed): calibration can be performed with that local watershed, and the Hydrological Parameters from that calibration can be exported (as “param.dat”) and applied to a new, non-watershed area project. To use a calibrated parameter set in a separate project: within i-Tree Hydro, select File > Build New with Hydrological Parameters, and from the file explorer, then select the calibrated param.dat file.

Again, i-Tree Hydro does not require model calibration for comparative assessments of base case versus alternative cases, and therefore using the default Hydrological Parameter set is suggested (especially for non-watershed areas where direct calibration is inherently impossible).
A member of the i-Tree Team

Please provide feedback on our technical support services by clicking here to take our 5 minute survey
Post Reply