Calibration Methods for Streamflow


Introduction

These notes contain basic methods for calibration of HFAM for water balance, including soil moistures, actual evapotranspiration and streamflows. These notes are almost all applicable to HSPF parameter calibration as well, since HFAM and HSPF share many algorithms and parameters.

HFAM uses input meteorologic time series of precipitation and potential evapotranspiration. For snowmelt, time series for air temperature, solar radiation, dewpoint and wind velocity are added (see "Calibration Methods for Snowmelt").

The following are the key hydrologic processes:

1) Infiltration - the movement of water during rainfall into the soil. A portion of this water becomes interflow and base or groundwater flow in streams. Actual evapotranspiration occurs from water that has infiltrated and is held in the soil.

2) Actual Evapotranspiration - the combination of evaporation from soils, and transpiration from vegetation. By definition actual evapotranspiration is less than or equal to the current potential evapotranspiration, where potential evapotranspiration is the rate at which water would evaporate from a free (lake) water surface. Actual evapotranspiration varies with soil properties, soil moisture and vegetation.

3) Soil Moisture - the storage or capture of water in soils for later evaporation, transpiration, or movement into perched or deep groundwater aquifers. Soil moisture variations control infiltration rate variations at the watershed surface.

4) Surface Runoff, Interflow and Groundwater Flow - Three "paths" for water movement from the land surface into streams are modeled. Surface runoff is modeled as flow on an inclined plane, where the slope, length and roughness of the plane are model parameters. Surface runoff occurs when rainfall rates exceed infiltration rates at the surface of a land segment.

Interflow is water that infiltrates as flows in the soil along 'preferred pathways'. Preferred pathways originate as animal and insect burrows, as shrinkage cracks, as roots of plants, and as buried microscale flowpaths were flowing water has deposited relatively coarse or permeable sediments. Tillage of agricultural lands may also create preferred pathways.

Interflow pathways are not necessarily continuous. Water may flow subsurface for a short distance, and then return to the surface and become surface runoff. Water moving on interflow pathways may also infiltrate into surrounding soils.

Groundwater flow is subsurface, saturated zone flow that contributes to stream channels. Groundwater flow creates the 'base flow' in streams, weeks or months after the precipitation or snowmelt has occurred in a watershed. If the saturated zone in a watershed is below the stream channels and the channels are influent rather than effluent, 'deep percolation' is occurring and the amount of water on this flowpath is represented by a model parameter.

Modeling three hydrologic flowpaths is fictional. The actual number of flowpaths that will exist in a watershed is unlimited. The three flowpaths are a convenience for modeling that allow continuously varying processes to be represented by a small number of parameters.

HFAM model parameters control how the hydrologic processes operate. The effect or importance of any given parameter or process will vary from one watershed to another.

The data that are typically available for calibration are:

  1. Precipitation, Class-A pan evaporation, and streamflow records
  2. Lake levels or contents, groundwater levels
  3. Records of changing land use such as urbanization, forestry or agricultural land use. Records of changes in the channel system, storm drain construction, construction of detention basins or reservoirs, may also be available.

Selecting HFAM Land Surface Segments

HFAM uses lands surface segments that are 'homogeneous' insofar as they experience the same time series of precipitation and potential evapotranspiration, and they have the same mean soil properties, vegetation, and topography. Algorithms cause key soil properties like infiltration capacity to vary within a land segment. The number of land surface segments that can be used is not limited by HFAM. Segments need not be contiguous. The land surface segment boundaries can be based on GIS overlays of precipitation, soils, vegetation, and topography, but the selection of criteria for creating segments is subjective.

In watersheds that include snow accumulation and melt, similarity of elevation, temperatures, and aspect are considered in addition to the factors above.

Calibration of Hydrologic Parameters

A summary of the principal HFAM parameters for hydrologic processes follows. Each parameter is listed under the hydrologic process that it primarily affects. There is interaction among parameters, especially for parameters affecting streamflow volumes. This interaction occurs in nature, and also occurs in modeling.

 

Streamflow Volumes, Watershed Water Balance: LZSN and UZSN; Nominal Soil Moisture Storages in the Lower and Upper Zones

Soil Moisture Storage, Actual Evapotranspirtation: INFILT; Infiltration rate (median) when the soil moisture storage ratio, LZS/LZSN = 1

Actual Evapotranspiration: LZETP; Evapotranspiration parameter from the Lower Zone Storage

Seasonal and Low Flows: AGWRC; Active Groundwater Recession Constant

Hydrograph Shape, Peak Flows: INTFW; Interflow (volume) parameter, IRC: Interflow Recession Rate

 

Watershed Water Balance

The soil profile may be viewed as 'water holding capacity' limited, or as 'infiltration capacity' limited. The soil profile can become saturated so that it will hold no additional water, or it might have such low permeability that very little water can enter, even though water holding capacity is present.

Soil profiles usually become water holding capacity limited more often than they become infiltration capacity limited.

Other parameters in HFAM that may affect watershed water balance, in addition to the parameters listed above under Streamflow Volumes, are FOREST , PETMAX, PETMIN and DEEPFR. Water balance is relatively insensitive to FOREST , PETMAX and PETMIN. However, non-zero values of DEEPFR have a significant effect.

Soil Moisture Storage, Actual Evapotranspiration

Soil moisture storage is critical, and is poorly approximated in traditional hydrologic analysis. It controls the division between subsurface and surface flows, and it affects runoff timing (floods) and water balance. Heterogeneity, or areal variability in infiltration/soil moisture storage is important as is variability in time.

There is interaction of soil moisture storage parameters, LZSN and UZSN, and the infiltration rate parameter INFILT. This can usually be resolved since INFILT also affects seasonal and low flows.

Seasonal and Low Flows

Seasonal and low flows are relatively easy to model since these processes are well damped in most watersheds. The AGWRC models a time invariant groundwater recession.

In some watersheds, where there are unusually large amounts of data on soils and soil depths, it may be possible to estimate the total volume of water in the saturated zone that contributes to streamflows.

Hydrograph Shape, Peak Flows

Hydrograph shape is affected by the relative magnitudes of surface runoff and interflow. This occurs among watersheds, and plays a role in creating unique watershed hydrologic regimes. It also occurs within a watershed and alters hydrograph shapes from storm to storm.

Channel cross-sections and slope are also important in hydrograph shape and peak flows on all but the smallest watersheds. On watersheds where channel flow timing is 10 minutes or less, the overland flow parameters, LSUR, SLSUR, and NSUR are important for hydrograph shape and peak flows.

Summary of the Calibration Procedure

1. The first goal in calibration is to correctly reproduce the long term water balance in the watershed. The water balance equation;

Precipitation - Actual ET Storage Change = Runoff

applies to any time interval. When HFAM is run using several years of meteorologic data, i. e. 3 to 10 years, the cumulative terms in the water balance equation (precipitation, actual evapotranspiration, and runoff) all increase but the storage change does not increase, and it must become small relative to the other terms. Storage change is the difference in total storage in the watershed over the time interval.

If the watershed precipitation and runoff are accurately established, and if the long term simulated runoff exceeds the recorded runoff then the simulated 'Actual ET' must be less than the watershed's Actual ET. To increase the simulated Actual ET more water must be retained in the soil profile for later evapotranspiration, provided that the simulated Actual ET is less than the Potential ET. Actual ET is increased by increasing LZSN, and possibly UZSN and INFILT.

It is important to ignore runoff timing in this first calibration step.

Difficulties with other terms in the water balance equation that are present will be apparent. If for example watershed precipitation rates are over or under estimated it may not be possible to reproduce the recorded watershed runoff.

This step is completed when the long term water balance, simulated and observed, matches reasonably closely.

2. Seasonal and low flows are calibrated by increasing or decreasing the volume of water flowing on the groundwater flowpath, or by changing the 'recession rate' for flows on the groundwater flowpath. The volume of water moving on the groundwater flowpath is sensitive to infiltration rates, the INFILT parameter. Increasing INFILT increases groundwater flows.

The timing of groundwater flows may also be calibrated. Observed groundwater recession rates can be found from streamflow records. AGWRC is equal to

(groundwater flow today/groundwater flow n days ago)1/n  

The AGWETP parameter (which creates ET losses from groundwater storage), and the BASETP parameter (which creates ET losses from the outflow from groundwater storage) can be used to refine simulated low and seasonal flows. These parameters are used together with field surveys of flood plain and streambank vegetation.

This step is completed when the seasonal flows and low flows, simulated and observed, match reasonably closely. The volume of water moving in the watershed on subsurface flowpaths in now established.

3. Hydrograph shape and peak streamflows will now be calibrated. Since the total runoff simulated and observed was matched in Step 1, and since the groundwater flows, simulated and observed, were calibrated in step 2, the total simulated surface runoff plus simulated interflow volume must be match the observed. HFAM provides one parameter, INTFW, to divide water between the surface runoff flowpath and the interflow flowpath. Increasing INTFW increases interflow and decreases surface runoff, and decreasing INTFW decreases interflow and increases surface runoff. This tradeoff does not affect water balance or seasonal and low flows.



Back to Index

Hydrocomp, Inc.