Introduction

This tutorial and guide was adapted from standard operating procedures used when developing HEC-HMS models for use within Corps Water Management System (CWMS) implementations.

Snow Data Sources, Collection, and Development

The temperature index snow model within HEC-HMS requires temperature data at each timestep, as well as initial snow conditions at the first timestep of the model. Precipitation data is not required at every timestep, but it is recommended. Snow observations are useful for calibration of the snow meteorological model. These data may come from multiple sources and may require development prior to incorporation into the models.

There are numerous sources of meteorological and snow data available, both point and gridded data. Options for obtaining and developing the data depend on several factors including the data type required, location, and date. In some instances, gridded data may need to be developed from point data by using spatial interpolation methods (e.g. GageInterp). In other cases, where point snow data is limited, point data can be extracted from gridded data at locations where other meteorological data (e.g. rain and temperature) are available. This section describes several sources of data and tools available to process the data into the correct format for the model.

Snow Water Equivalent and Other Snow Data

SWE represents the volume of liquid water available in the snowpack. Changes in SWE are modeled within HEC-HMS. If the simulation begins when there is an existing snowpack, an estimate of initial SWE at the start of the simulation is required as an initial condition; otherwise, an initial value of zero can be used. Additionally, SWE observations are useful for calibration of the snow model. Both point observations and gridded SWE data are available from numerous sources.

National Operational Hydrologic Remote Sensing Center Snow Data Assimilation System

The National Oceanic and Atmospheric Administration’s (NOAA) National Operational Hydrologic Remote Sensing Center (NOHRSC) provides modeled snow data output from the SNOw Data Assimilation System (SNODAS) model. SNODAS integrates modeled snow estimates with satellite, airborne and ground observations to create 1 kilometer-resolution gridded datasets of snow covered area, SWE, and other snow variables for the continental United States. These data are available daily from October 2003 through the present (http://www.nohrsc.noaa.gov/).

Natural Resources Conservation Service Snow Telemetry

The Natural Resources Conservation Service (NRCS) owns and maintains a network of snow observation sites, SNOw TELemetry (SNOTEL). The NRCS also regularly conducts manual snow surveys throughout the mountainous regions of the western United States (http://www.wcc.nrcs.usda.gov/snow/). Each SNOTEL site is an automated, near real-time data collection station that collects hydroclimatic data including SWE, snow depth, precipitation, and temperature observations. SNOTEL data collection began in 1978 at some stations and extends through the present. Snow courses measure snow depth and water content along a designated transect, typically around the first of each month during the winter season. The federal snow course program began in 1934 to help with water supply forecasting.

United States Army Corps of Engineers Snow Surveys

Several districts within the United States Army Corps of Engineers (USACE) conduct regular snow surveys during the winter, typically at or near project sites. Surveys are generally done on a weekly or monthly basis and follow a standard procedure, similar to NRCS. Volunteer snow survey programs have also been established in a limited number of locations. USACE districts provide SWE measurement equipment to volunteers, who collect SWE information and provide the data to USACE at various times throughout the snow season.

Remote Sensing

Satellite observations can provide spatially distributed estimates of snow covered area (SCA) and SWE at a regular temporal resolution (daily). HEC-HMS does not currently provide a means to incorporate SCA data within the model, but it can be useful for comparison to the model output of distributed snow. Visible and near-infrared measurements provide fairly high-resolution estimates of SCA, though the data are limited during periods of cloud cover or at night. The National Aeronautics and Space Administration’s (NASA) MOD10A1 daily snow product is an example of remotely-sensed snow cover data (http://modis-snow-ice.gsfc.nasa.gov/?c=MOD10A1). A filtering technique developed at the Cold Regions Research and Engineering Laboratory (CRREL) can help eliminate clouds from the data.

Passive Microwave Measurements

Passive microwave measurements are available at a coarser resolution and can provide SWE estimates even during cloudy or night-time conditions. These data are archived at the National Snow and Ice Data Center (NSIDC) (http://nsidc.org). The passive microwave data are limited however by deep snowpacks and heavy vegetation, particularly in mountainous regions. The signal is also affected by wet snow. Locations within the U.S. where this data may be useful are in large basins in the southern Rocky Mountains or in the Northern and Central Plains regions.

Daymet

Daymet is a collection of algorithms and computer software designed to interpolate and extrapolate from daily meteorological observations to produce gridded estimates of daily weather parameters for North America. This includes daily continuous surfaces of minimum and maximum temperature, precipitation occurrence and amount, humidity, shortwave radiation, snow water equivalent, and day length. The model produces a gridded output at a spatial resolution of 1 kilometer and a temporal resolution of 1 day. The required model inputs include a digital elevation model and observations of maximum temperature, minimum temperature, and precipitation from ground-based meteorological stations. The model method is based on the spatial convolution of a truncated Gaussian weighting filter with the set of station locations. Sensitivity to the typical heterogeneous distribution of stations in complex terrain is accomplished with an iterative station density algorithm. The model output includes minimum and maximum temperature, precipitation, water vapor pressure, shortwave radiation, and SWE. Daymet is supported by funding from the National Aeronautics and Space Administration (NASA) and the U.S. Department of Energy’s Office of Science (https://daymet.ornl.gov/). The Daymet model results are available from 1980 to the present, for a period of record of 37 years. Limitations include lack of ground-truthing and limited testing. The reliability and limitations of this dataset are currently being investigated.

Other Sources

Certain National Weather Service (NWS) offices collect snow depth and SWE data for NWS stations (https://www.ncdc.noaa.gov/snow-and-ice/). Another source of snow data is the Community Collaborative Rain, Hail & Snow network (CoCoRaHS), a volunteer network of weather observers, that can collect snow depth information and sometimes measure SWE (http://www.cocorahs.org). Finally, several state environmental or dam safety agencies collect their own manual snow surveys and may be a source of information.

Precipitation

Precipitation data are used in each timestep during an HEC-HMS simulation, but it is not required in all timesteps. Precipitation in the form of rainfall or snowfall is determined by the HEC-HMS model based on the precipitation discrimination (PX) temperature. These data are typically obtained from the NWS at point observation sites or in gridded format for the current and forecast time periods. Additional gridded options are available during the historical period.

National Weather Service

In addition to observational station data available from NOAA’s National Center for Environmental Information (NCEI) formerly the National Climatic Data Center (NCDC) (https://www.ncdc.noaa.gov/cdo-web/), each NWS River Forecasting Center (RFC) collects gridded NEXRAD radar data on a daily basis to generate Multisensor Precipitation Estimator (MPE) data. These data are available for download or can be requested through individual RFC’s (http://dipper.nws.noaa.gov/hdsb/data/nexrad/nexrad.html). Additional manipulation is required to convert data to DSS for use in HEC-HMS. Stage III precipitation data are available from the mid-1990s through the early 2000s. MPE data are available from the early 2000s to the present.

Parameter-elevation Relationships on Independent Slopes Model

The Parameter-elevation Relationships on Independent Slopes Model (PRISM) Climate Group of Oregon State University provides estimates of multiple spatial climate data sets derived from modeled results and climate station networks (http://www.prism.oregonstate.edu/). The precipitation data sets are available in gridded format and in either daily or monthly temporal distributions from 1981 through the present. PRISM precipitation grids include rain and melted snow (http://www.prism.oregonstate.edu/documents/PRISM_datasets.pdf).  An example application that utilizes PRISM precipitation data can be found here: Importing and Using Parameter-elevation Regressions on Independent Slopes Model (PRISM) Precipitation Data.

Reanalysis Data

Reanalysis is a method of producing gridded estimates of meteorological data sets for a historical period through a combination of modeling and observational data assimilation. These data are typically generated for climate research on a global scale and may be less accurate for regional studies, depending on the availability of observational data. Some examples of reanalysis data include, NASA Modern Era Retrospective-Analysis for Research Applications (MERRA, http://gmao.gsfc.nasa.gov/reanalysis/merra/), the European Centre for Medium-Range Weather Forecasts (ECMWF, http://www.ecmwf.int/en/research/climate-reanalysis), as well as NOAA’s National Centers for Environmental Prediction - North American Regional Reanalysis (NCEP-NARR, http://www.esrl.noaa.gov/psd/data/gridded/data.narr.html).

Temperature

Temperature data are required for each timestep of an HEC-HMS snowmelt simulation. Temperature data are used as an index for all of the energy fluxes into the snowpack and they are used to determine whether precipitation falls as rain or snow, whether snow melts, and at what rate melting occurs. Data are available at point locations from the NWS. Gridded temperature data can be obtained from several sources or created through interpolation of point measurements. When creating gridded temperature data from point data, it is important to account for the effects of elevation. Within GageInterp, a lapse rate, or rate at which temperature drops with increasing elevation can be incorporated. An average lapse rate is 3.6 ^oF/1000 feet (6.5 ^oC/1000 meter), though this value can be calculated from available data as shown within the following figure. The slope of the lines represent the lapse rate.

National Weather Service

The primary source of historical temperature data is from NWS observation stations, which can be accessed through the NCEI (https://www.ncdc.noaa.gov/cdo-web/). Point observations can be converted to gridded data using a calculated lapse rate and HEC’s GageInterp program.

Parameter-elevation Relationships on Independent Slopes Model

The PRISM Climate Group of Oregon State University provides estimates of multiple spatial climate data sets derived from modeled results and climate station networks (http://www.prism.oregonstate.edu/). The temperature data sets are available in gridded format and in either daily or monthly temporal distributions. The daily temperature data comes in the form of Max Temperature, Min Temperature, and Mean Temperature for a given day. For snowmelt models within HEC-HMS, further data development would be required to create gridded temperature data for an hourly timestep. Contact HEC for assistance.

Gridded Data Development

If gridded data are not available at the location or for the time period of interest, it can be developed through interpolation of point data. The HEC software package GageInterp can convert a time series of point data into DSS gridded objects, and can use a bias grid or lapse rate to account for elevation effects. The HEC point of contact and the local district system administrator should work together to produce this product.  Examples or commonly used gridded snow data can be found here: Gridded Data Sources.

Snow HEC-HMS Meteorologic Model Creation and Calibration

It is recommended that the HEC-HMS snowmelt parameters be developed and calibrated separately from the basin runoff calibration using ground observations or independent SWE estimates (e.g. SNODAS). Snowmelt parameter calibration reduces the number of tuning parameters affecting the hydrological results during hydrologic model runoff calibration. Once the snow parameters have been calibrated to match observed snow accumulation and melt, selected snowmelt runoff events can be simulated within the HEC-HMS model. The hydrological parameters calibrated for rainfall events can be further adjusted to match snowmelt events.

Initial Snowmelt HEC-HMS Meteorological Parameters

Within the HEC-HMS snow meteorological model there are several initial snow conditions that must be set in addition to the snow model parameter values. This section describes both initial snow conditions and snow model parameters.

Initial Conditions

The initial conditions of the model represent the physical conditions of the snowpack at the start of the simulation. This section describes those parameters and explains how to determine their values for the start of the model.

Initial Snow Water Equivalent

The Initial SWE is the SWE, in inches, that exists at the beginning of the simulation. This information can be obtained from ground measurements of SWE or available gridded estimates (e.g. SNODAS). If there is no snow, this value can be set to zero.

Initial Cold Content

The Cold Content (cc) represents the heat required to raise the temperature of the snow pack to 32^o F and is expressed in inches of water. This value can be calculated with the following SNODAS equation if the initial SWE and snowpack temperature are known:

$//<![CDATA[ \begin{array}{l}\mathrm{CC}=\frac{S W E * C_{p} * \Delta T_{S}}{L_{f}}\end{array} //]]>$

where ΔT_s = T_savg – T _base ; T_savg= Snow pack average temperature; T_base= Base temperature (melting temperature of snow); L_f = latent heat of fusion of water (334,000 J/kg); and C_p = specific heat of ice (2100 J/kg-ºC).  If this value is not known at the start of simulation or if there is no snow, it can be set to zero. The error in doing this when the value is unknown may be small for relatively shallow ephemeral snow covers, but it may cause melt to begin too early for deep, seasonal snowpacks.

Initial Liquid Water Content

The Initial Liquid Water Content is the liquid water, in inches, held within the snowpack. For any melt or precipitation to be released from the snowpack, the liquid water holding capacity of the snow must first be satisfied. Liquid water can persist in the snow only if the snowpack temperature is at 32 ^oF; at which point the cold content is zero. A snowpack with liquid water is said to be “ripe.” Generally, this value is not known at the start of the simulation unless there is no snow, in which case it can be set to zero. Initial liquid water content can also be set to zero if the snow is assumed to be below freezing at the start of the simulation. If the simulation begins during the melt period, the error in setting the initial value to zero may cause a delay in melt runoff.

Antecedent Temperature Index for Cold Content

The Antecedent Temperature Index for Cold Content (ATICC) is an index used to represent the snow temperature near the surface of the snowpack. It is calculated assuming an approximation to the transient heat flow equations. This value is used to estimate the cold content of the snow. It should be set to the approximate snowpack temperature, if known (often this is not known or documented, but this information may be included in a snow survey). If unknown, it can be set to 32 ^oF.

Antecedent Temperature Index for Meltrate

The seasonal variation of the meltrate is indexed by the Antecedent Temperature Index for Meltrate (ATIMR). The initial melt antecedent temperature index (ATI) should be thought of as similar to “the accumulated thawing degree days.” This antecedent temperature function allows the meltrate to change over the course of the spring melt period. If there is no snow on the ground at the start of the simulation, or if the simulation begins before the onset of spring melt this term can be set to zero. Once the spring melt period begins, the antecedent temperature index should be calculated based on observed air temperature data above the base temperature since the onset of melt.

Snow Model Parameters

The snow model parameters influence how a snowpack changes over time, and those parameters are described in this section.

Discrimination Temperature

The Discrimination Temperature (PX Temperature) is the threshold between precipitation falling as rain or snow. When the air temperature is less than the specified PX Temperature, any precipitation is assumed to be snow. When the air temperature is above the specified PX Temperature, any precipitation is assumed to be rain. Typical values range from 32 ºF to 35 ºF.

Base Temperature

The Base Temperature is the temperature at which snow melts. The difference between the base temperature and the air temperature defines the temperature index used in calculating snowmelt. If the air temperature is less than the base temperature, then the amount of melt is zero. Typically, the base temperature is set to 32 ^oF, or a value close to it.

Rain Rate Limit

The Rain Rate Limit is the discrimination rain rate which determines whether the dry meltrate (snowmelt during warm periods with no precipitation) or wet meltrate (snowmelt that is precipitation induced) is used. The wet meltrate is typically greater than dry meltrate primarily due to the condensation of water vapor from the rain inside the snowpack. If the rain rate is less than the rain rate limit, the meltrate is computed as if there were no precipitation. Suggested values of the rain rate limit range from 0.25 to 1.0 inches/day.

Wet Meltrate:

The Wet Meltrate is used during periods of rainfall precipitation when the precipitation is falling at rates greater than the rain rate limit. The wet meltrate is applied as the meltrate when it is raining at rates greater than the rain rate limit and the dry meltrate is applied when the rate of rain is less than the rain rate limit (as if there was no precipitation). Typical values set at the high end of the dry meltrate range from 0.08 to 0.15 inches/^oF-days.

Antecedent Temperature Index Meltrate Coefficient

The ATI meltrate coefficient is the model die-away coefficient used to adjust the meltrate ATI calculated during the previous timestep. It should be set to 0.98.

Antecedent Temperature Index Meltrate Function

The ATI meltrate function allows the meltrate to change as snowpack matures and ages as a function of accumulated thawing degree-days. Typically, a deep snowpack (e.g. 10-20 inches SWE) with a long melt period will have a meltrate that changes seasonally to account for increased solar radiation. A shallow snowpack that melts quickly can typically be modeled reasonably well with a constant meltrate function. A table of meltrate versus ATI values is required, which adjusts the dry meltrate based on the antecedent temperature index for meltrates (ATIMR). The function should define appropriate meltrates to use over the range of meltrate index values that will be encountered during a simulation. Meltrates typically range from 0.015 to 0.15 inches/^oF-days. See section 3.4 for additional information.

Meltrate Pattern

The Meltrate Pattern is used to adjust the dry snow meltrate computed from the index meltrate function. Changes in a snowpack albedo and/or incoming solar radiation can be captured through the use of a meltrate pattern. A paired dataset defines the meltrate pattern as a percentage adjustment as a function of the time of year.

Cold Limit

The cold content of an existing snowpack is reset when a sufficient amount of new snowfall precipitation accumulates. When the precipitation rate exceeds the specified cold limit, the cold content index is set to the air temperature at the time of the precipitation if the air temperature is below base temperature. If the temperature is above base temperature, the cold content index is set to base temperature. If the precipitation rate is less than the cold limit, the cold content index is computed as an antecedent index. The suggested value for the cold limit is 0.2 inches/day.

Antecedent Temperature Index Coldrate Coefficient

The ATI Coldrate Coefficient (ATICC) represents the influence of air temperature on the internal temperature of the snowpack. Values can range from 0 to 1.0. Higher values will more closely track the observed air temperature. For instance, if this coefficient is set to 1.0, the snow temperature will remain close to the air temperature. Thus, values closer to 1.0 are more appropriate for shallow snow packs and values closer to 0.0 are more appropriate for deep snow packs.

Antecedent Temperature Index Coldrate Function

The ATI Coldrate Function is a table of values that adjust the coldrate based on the ATICC. Values typically range from 0.01-0.028 inches/^oF-days, with a reasonable constant value of 0.02 inches/^oF-days recommended.  

Liquid Water Capacity

The Liquid Water Capacity is the maximum amount of liquid water that can be held in the snowpack before runoff occurs. Liquid water can persist in the snow only if the snowpack temperature is at 32 ^oF (0 ^oC). This is calculated for every timestep as SWE decreases. Suggested values range from 3 to 5 percent.

Groundmelt

Groundmelt is the rate at which snowmelt occurs due to heat from the ground entering the snowpack. The value is typically set to 0 inches/day.

Point Snowmelt and Accumulation Model

The primary purpose of this step is to accurately simulate SWE accumulation and melt over the snow season at a location within the basin with SWE, temperature, and precipitation observations. This process allows the modeler to converge on temperature index parameters that will be used in the HEC-HMS model.

It is recommended that ground observations of SWE be used to develop initial estimates of the snow model parameters. This can be done using HEC-HMS to simulate the snow accumulation and melt at one or several point locations and comparing the modeled SWE to observed SWE. In HEC-HMS a simple basin model can be set to run the temperature index snowmelt (not gridded) meteorologic model with observed precipitation and temperature data. A typical location would be at a SNOTEL location where observed SWE, temperature, and precipitation data can be gathered.

The following steps are a general workflow for this process. The following figure provides an example basin point model.

1. Create subbasins for each point location that has observed temperature, precipitation and SWE data. Enter 1 square mile for the subbasin areas. Change all the modeling methods to “None.”
2. Create precipitation and temperature gages with all of the observed data. The temperature gages will require an elevation.
3. Create a meteorologic model using specified hyetograph and temperature index methods for the Precipitation and Snowmelt components, respectively. Within the meteorologic model, assign one elevation band to each subbasin using the elevation of the SWE gage. If the SWE and temperature data come from different locations, then a lapse rate will be required for the model to estimate the temperature at the elevation of the SWE observations.
4. Enter initial temperature index parameters and assign all precipitation and temperature gages to their respective subbasins.
5. Create a control specification to run a desired length of time, which typically includes the complete snow accumulation and melt cycle for a given year. If data is available, running multiple years with snowpack is recommended. The timestep used will be dependent on the observed data. Most SNOTEL sites collect daily data.
6. Create a simulation run and compare modeled and observed SWE at each point location. Figure 3-2 shows example results of the HEC-HMS model run at a point location compared to observations.
7. Make changes to parameters and continue calibrating simulations to match the observed results.

For water resource applications, the snow melt period is generally the most important. To model the melt period from the peak SWE, a similar process is followed; however a simulation should be created for each year analyzed. Within the meteorologic model, the initial SWE will need to be defined for each subbasin which can be obtained from the observational data. This process is more time consuming, but it allows the modeler to refine the temperature index model parameters, specifically the ATI Meltrate Function, before moving on to the gridded HEC-HMS model. Engineering judgment should be used to determine a set of recommended meteorological model parameters to be used for initial basin snowmelt and accumulation model analysis (step 1.4.3); however, it is worth noting, that these parameters will undergo further refinement so significant time should not be spent getting a “perfect” parameter set.

Gridded Basin Snowmelt and Accumulation Model

Once the development of a gridded HEC-HMS model is completed (e.g. Geospatial Hydrologic Modeling Extension (GeoHMS) delineated subbasins have been imported into HEC-HMS with associated grid cell file), the meteorologic parameters in the snowmelt model developed through point analysis can be further calibrated with comparison to available gridded snow data. The default modeled snow output in HEC-HMS are time series of basin-averaged snow variables. To evaluate the gridded snow model results, the basin-averaged SWE from HEC-HMS can be compared to other gridded SWE estimates averaged over the same basins. A typical application would be to extract basin-averaged SWE from an observed gridded dataset such as SNODAS and to use HEC-MetVue to compute a basin-averaged SWE hyetographs for each subbasin. A comparison can then be made between the two basin averaged SWE hyetographs for multiple subbasins. Snowmelt meteorologic parameters can then be adjusted to better match observed data.

The following figures show an example of the basin-averaged SWE results from a gridded HEC-HMS model compared to basin-averaged SNODAS data. Initial SWE and cold content grids were obtained from SNODAS data, and initial ATICC, ATIMR, and liquid water content were set to default values (see section 3.1.1). During this process, the modeler can refine meteorological parameters (i.e. ATI Meltrate Function) to better fit SNODAS results. Because only one set of gridded temperature index parameters can be used for the entire basin model, it is important to perform comparisons for multiple subbasins to verify that the parameters are representing the snowmelt process adequately.

An alternative method for evaluating the performance of the gridded HEC-HMS snowmelt model is to compare the spatial extent of the gridded model output to observed snow covered area (SCA) observations. Within the HEC-HMS Program Settings, the option is available to “store gridded state variable results.” With this option checked the model will output modeled data storage system (DSS) gridded estimates of the snow variables which can be viewed in HEC-MetVue or exported to geographic information systems (GIS) and compared spatially to observed data. Sources of distributed snow extent data include satellite imagery from Moderate Resolution Imaging Spectroradiometer (MODIS) or Visible Infrared Imaging Radiometer Suite (VIIRS), as well as the SNODAS model data.

Several standard statistical measures are appropriate for evaluating the performance of the HEC-HMS snow model compared to SWE observations. If the entire winter season is modeled, then the difference between observed and modeled peak SWE is used to evaluate the total volume of SWE calculated. The difference between the modeled and observed post-melt snow-free day of year will provide information on the timing performance of the model. Several additional statistics can be used to assess the model performance throughout the winter or snowmelt season (Moriasi et al., 2007). These include mean bias, Pearson correlation coefficient (R), the normalized root mean squared error (RMSE), and the Nash-Sutcliffe model efficiency (NSE). Ultimately a recommended snowmelt parameter set should be identified for use for runoff modeling. If a single parameter set is not applicable and it has been determined that various stimuli result in the need for multiple parameter sets, recommendations should be provided to determine when each parameter set should be selected for runoff modeling.

It is recognized that the temperature index model has limitations. For example, it does not account for all hydrometeorological phenomenon that naturally occur in a basin (e.g. hillshading, aspect, sub-grid variability, etc.), and only one meteorologic model can be used per basin model. Some situations may warrant the need to break down the greater basin into smaller basins which could be modeled with separate meteorologic models. Consultation with the project delivery team (PDT) and advisors should be sought before deciding to go this route.

Antecedent Temperature Index-Meltrate Function Development

Two of the most influential parameters affecting snowmelt are temperature and meltrate, which is defined by ATI-Meltrate functions. In a general sense, the meltrate at a particular timestep is equivalent to computed degree days of the timestep (instantaneous, not accumulated degree days) multiplied by the meltrate, which is determined from the meltrate function. This function largely dictates the ability of the model to simulate correctly the rate at which snow melts as well as the timing of the snow-free date. This section provides additional information about the development of the meltrate function to help in model calibration.

The rate of snowmelt typically increases as the snowmelt season progresses due to increased solar radiation input and metamorphic changes within the snowpack. In the HEC-HMS temperature index model, the ATIMR function relates the change in meltrate throughout the season to the accumulated thawing degree days. A deep snowpack with a long melt period will likely have a meltrate that changes seasonally to account for increased solar radiation and more mature snowpack. Observed data within the basin can be used to develop the ATI-Meltrate relationship that best fits across multiple locations and years. The following figure is a plot of the SWE versus the ATI at a station during one season. The absolute values of the slopes of the lines represent meltrate and can be determined using a linear trendline for the segments of approximate constant slope. A steeper slope as the ATI increases corresponds to a faster meltrate. Breaks in slope indicate a potential need to use an ATIMR function that varies with ATI. The following table provides a sample ATIMR function representative of the observed data shown in the following figure for that particular gage and year.

Typically, a shallow snowpack that melts quickly (e.g. plains) can be modeled reasonably well with a constant meltrate function. Data can be developed for a point or basin such as in the following figure to verify this assumption.

Sample Antecedent Temperature Index Meltrate Function

In order to demonstrate meltrate in HEC-HMS, Figure 3-6 illustrates a plot of SWE versus ATI from output of HEC-HMS when a constant ATIMR function is used in the meteorologic model. Note how the absolute value of the slope of the line (0.0975 inches/°F-days) is approximately the indicated meltrate (0.09 inches/°F-days). The small difference is likely due to the higher wet meltrate having some influence during the melt and causing the average meltrate to be higher than the dry meltrate.

Snow meltrates are a function of the physical characteristics of a region, including forest cover, topography, and average weather conditions. A heavily forested basin where incoming solar radiation and wind are reduced will typically have lower meltrates than an open area. Basins with large variations in topography (i.e. slope and aspect) will generally have lower average meltrates across the basin than flat areas. Basins with heavy forest cover or generally cloudy or humid conditions have meltrates that vary less during the melt season than open, dry locations. When developing the ATI-Meltrate table it is important to understand how the basin characteristics affect the meltrates and analyze the available ground measurements.

Snowmelt Runoff Model Calibration and Verification

Although required input is different, calibration and verification of a HEC-HMS snowmelt model is similar to that of a rainfall runoff model. The purpose of calibration is to match the peak (flow and timing) and volume of historic events within the basin. This section describes event selection, calibration, verification, and determination of preferred parameter sets for snowmelt runoff modeling using HEC-HMS.

Identify Snow Events of Interest

The type of snow data available often depends on the time period of interest. During this step, various historical snowmelt events should be identified with particular interest paid to recent years where data are most readily available. Types of snowmelt events that should be considered include years with a significant peak snowpack, years when the melt occurred late in the season (thus possibly resulting in a rapid melt-out and high flows), and events with combined rainfall precipitation and snowmelt contributing to runoff. Often it is easiest to select the same years which were used to calibrate the gridded snowmelt parameters, as many of the required inputs have already been developed

Calibration of HEC-HMS Model

To simulate a snowmelt runoff calibration event in HEC-HMS, observed streamflow, precipitation, and temperature data are required over the entire time period, as well as initial snow conditions at the start of the simulation. The basin model calibrated for rainfall runoff events should initially be used in the snowmelt runoff calibration. It is expected that most of the hydrologic parameters should remain the same regardless of the time of year. However, the initial loss and baseflow conditions will likely be different depending on the antecedent conditions. In addition, the process of seasonal snowmelt runoff is generally better represented by a multiple layer baseflow method that accounts for percolation and basin response, versus a single recession curve relationship which is better suited for event-based runoff. It is recommended that a linear reservoir baseflow method be used in basins where seasonal volumetric runoff is an important consideration in addition to peak flows.

If calibrating for numerous events, multiple basin models should be created to track parameters during calibration. Only events for which gridded precipitation/temperature data and streamflow records are available or can be developed should be used for calibration. If historical flow and meteorological data has not already been collected, it can be obtained and processed from the sources described above. Calibration and verification of snowmelt runoff is similar to the process for the rainfall model. The modeler may choose to use optimization trials or to calibrate manually. The process below describes using optimization trials but is applicable to manual calibration as well.

Initial Basin Parameters

Initial basin parameters should be developed according to guidance provided here: Applying Loss Methods within HEC-HMS, Estimating Clark Unit Hydrograph Parameters, Applying Baseflow Methods in HEC-HMS, Applying Reach Routing Methods within HEC-HMS. Additional guidance on method and parameter selection can be found in the HEC-HMS Technical Reference Manual. Additional helpful references for developing the Linear Reservoir parameters are Fleming (2002) and Bennett and Peters (2000).

Creation of Optimization Trials

Once historic data has been collected and entered into the HEC-HMS project, a new simulation run must be created for each historic event for which calibration will be conducted. Using the simulation run and observed streamflow data as inputs, a new optimization trial can be created from the Compute menu in HEC-HMS. When creating the optimization trial, you must specify the HEC-HMS element that you will be optimizing for; only elements with observed flow can be selected. Only parameters at or upstream of the selected element will be evaluated and adjusted during the optimization trial. An optimization trial should be created for each gaged location within the watershed and not just at the outlet.

Parameter Search Method

The optimization trial can either adjust one parameter at a time while holding other parameters constant (univariate gradient method) or evaluate all parameters simultaneously and determine which parameters to adjust (Nelder-Mead method); for most applications, the Nelder-Mead method is recommended. The optimization method can be specified by clicking on the optimization trial under the compute tab of the Watershed Explorer and selecting the appropriate method in the component editor. The tolerances and number of iterations should remain at the default values.

Objective Function

The Objective Function measures the goodness of fit between the computed outflow and the observed streamflow at the selected element. Several functions are available for measuring goodness of fit. The objective function used in the optimization trial can be specified in the component editor for the objective function.

Specify Parameters

Parameters that will be adjusted should be added to the optimization trial by right clicking the optimization trial in the Watershed Explorer and selecting Add Parameter. Select the watershed element and associated parameter to be evaluated during the optimization process. Leave the default values for minimum and maximum. Add as many parameters for as many watershed elements as needed to calibrate the model.

Optimization Trial Simulation

Run the optimization trial and view results and recommended optimization parameters for each trial under the results tab of the Watershed Explorer. Iteratively continue adjustments to watershed parameters until all targets are met; if any locations exist where calibration targets cannot be achieved, then reevaluate the model configuration, boundary conditions, basic characteristics, and reasonability of calibration targets. For additional discussion on model calibration refer to the HEC-HMS User’s Manual and Technical Reference.

Verification of HEC-HMS Model

Once the model has been calibrated to historic events, model adequacy will be verified by running other snow events that were not used in the calibration. The verification events should be entered into the HEC-HMS project as observed data (streamflow, precipitation, temperature, and initial snow) and a new control specification and simulation created for each event. Calibration targets will be set appropriately by the team lead based on schedule and funding limits, and will be targeted to provide best calibration results within given team schedule. It is expected that the district will further calibrate and validate the models.

Development of Final Suggested Parameter Set

The HEC-HMS modeler should compare output from the verification simulation runs to the observed streamflow data. If the simulation does not adequately match the observed data, further adjustment to model parameters may be needed, using the same procedure described for the calibration process. Once adjustments to watershed parameters have been made for the verification events, the user will re-run all events used in both the calibration and verification processes. This may be an iterative process until reasonable calibration targets are met for the events.

Documentation of Calibrated/Validated HEC-HMS Model

Documentation describing model development and the calibration process should be prepared to supplement the model deliverables once model development and calibration is complete. This documentation should be included in the final watershed report.