SUMMA Model Guide#

Overview#

SUMMA (Structure for Unifying Multiple Modeling Alternatives) is a physically-based hydrological model designed to systematically evaluate modeling decisions. SUMMA provides a common framework to test multiple conceptual approaches for representing dominant hydrological processes.

Key Capabilities:

  • Distributed, physically-based snow and hydrological modeling

  • Multiple options for each physical process (200+ decision combinations)

  • Detailed snow physics with multiple layering schemes

  • Sophisticated energy balance calculations

  • Flexible spatial discretization (HRUs, GRUs)

  • Glacier modeling support (with glacier mode)

  • Native parallel execution support (MPI/OpenMP)

  • NetCDF-based input/output

Typical Applications:

  • Snow hydrology and snowmelt processes

  • Mountain/alpine hydrology

  • Energy balance studies

  • Process uncertainty quantification

  • Hydrological model structural uncertainty analysis

  • Glacier mass balance and hydrology

Spatial Scales: Point to regional (tested up to continental scale)

Temporal Resolution: Sub-hourly to daily (typically hourly)

Model Physics and Structure#

Mathematical Foundation#

SUMMA solves the mass and energy conservation equations for:

  1. Energy Balance:

    • Surface energy balance (shortwave/longwave radiation, sensible/latent heat)

    • Snowpack energy evolution

    • Soil heat transfer

  2. Mass Balance:

    • Snow accumulation and melt

    • Canopy interception and throughfall

    • Infiltration and saturation excess runoff

    • Unsaturated zone drainage

    • Saturated zone baseflow

    • Evapotranspiration

  3. State Variables:

    • Snow water equivalent and density (per layer)

    • Soil moisture (per layer)

    • Soil temperature (per layer)

    • Canopy water storage

    • Aquifer storage

Spatial Discretization#

SUMMA uses a two-level hierarchy:

Grouped Response Units (GRUs):

  • Collections of HRUs with similar characteristics

  • Used for parallel execution (can run GRUs independently)

  • Typically represent sub-basins or computational domains

Hydrologic Response Units (HRUs):

  • Finest spatial element

  • Homogeneous land surface characteristics

  • Can be defined by elevation bands, land use, soil type, aspect, etc.

  • Downslope connectivity between HRUs (if SETTINGS_SUMMA_CONNECT_HRUS = true)

Decision Options#

SUMMA’s unique feature is the ability to specify different model physics through decision options:

SUMMA_DECISION_OPTIONS:
  snowLayers: CLM_2010              # Multi-layer snow model
  vegeParTran: CLM_NOAHMP           # Vegetation-canopy parameterization
  stomResist: BallBerry             # Stomatal resistance
  soilStress: NoahType              # Soil moisture stress function
  spatial_gw: localColumn           # Groundwater parameterization

Over 30 decision categories with 2-5 options each allows for flexible model configuration.

Configuration in SYMFLUENCE#

Model Selection#

To use SUMMA in your configuration:

HYDROLOGICAL_MODEL: SUMMA

Key Configuration Parameters#

Installation and Execution#

Parameter

Default

Description

SUMMA_INSTALL_PATH

default

Path to SUMMA installation directory

SUMMA_EXE

summa_sundials.exe

SUMMA executable name

SUMMA_TIMEOUT

7200

Model execution timeout (seconds)

SETTINGS_SUMMA_USE_PARALLEL_SUMMA

false

Enable parallel SUMMA execution (using specified backend)

SETTINGS_SUMMA_PARALLEL_BACKEND

slurm

Parallel backend (slurm or local)

SETTINGS_SUMMA_PARALLEL_EXE

summa_actors.exe

Parallel SUMMA executable name

SETTINGS_SUMMA_PARALLEL_PATH

default

Path to parallel SUMMA executable

File Management#

Parameter

Default

Description

SETTINGS_SUMMA_FILEMANAGER

fileManager.txt

Master file manager

SETTINGS_SUMMA_ATTRIBUTES

attributes.nc

HRU attribute file

SETTINGS_SUMMA_FORCING_LIST

forcingFileList.txt

List of forcing files

SETTINGS_SUMMA_COLDSTATE

coldState.nc

Initial conditions file

SETTINGS_SUMMA_TRIALPARAMS

trialParams.nc

Parameter file for calibration

SETTINGS_SUMMA_OUTPUT

outputControl.txt

Output variable selection

Calibration Parameters#

Parameter

Default

Description

PARAMS_TO_CALIBRATE

null

Comma-separated list of parameters to calibrate

BASIN_PARAMS_TO_CALIBRATE

null

Basin-scale parameters to calibrate

CALIBRATE_DEPTH

false

Enable soil depth calibration

DEPTH_TOTAL_MULT_BOUNDS

null

Bounds for total depth multiplier [min, max]

DEPTH_SHAPE_FACTOR_BOUNDS

null

Bounds for depth shape factor [min, max]

Common calibration parameters:

  • routingGammaShape: Gamma distribution shape for routing (1.0 - 5.0)

  • routingGammaScale: Gamma distribution scale for routing (1.0 - 500.0)

  • summerLAI: Summer leaf area index (0.5 - 10.0)

  • winterLAI: Winter leaf area index (0.01 - 5.0)

  • aquiferBaseflowExp: Baseflow exponent (1.0 - 10.0)

  • aquiferBaseflowRate: Baseflow rate (0.001 - 1.0 mm/day)

Glacier Mode#

Parameter

Default

Description

SETTINGS_SUMMA_GLACIER_MODE

false

Enable glacier preprocessing

SETTINGS_SUMMA_GLACIER_ATTRIBUTES

attributes_glac.nc

Glacier-specific attributes file

SETTINGS_SUMMA_GLACIER_COLDSTATE

coldState_glac.nc

Glacier initial conditions file

Parallel Execution#

For large-scale applications:

Parameter

Default

Description

SETTINGS_SUMMA_GRU_COUNT

85

Total number of GRUs

SETTINGS_SUMMA_GRU_PER_JOB

5

GRUs processed per parallel job

SETTINGS_SUMMA_CPUS_PER_TASK

32

CPUs allocated per task (SLURM)

SETTINGS_SUMMA_MEM

5

Memory per task (GB)

SETTINGS_SUMMA_TIME_LIMIT

01:00:00

Wall time limit (HH:MM:SS)

Decision Options#

Customize model physics:

SUMMA_DECISION_OPTIONS:
  # Snow model structure
  snowLayers: CLM_2010           # Options: jrdn1991, CLM_2010

  # Vegetation and canopy
  vegeParTran: CLM_NOAHMP        # Options: Noah_BATS, CLM_NOAHMP, SSiB
  stomResist: BallBerry          # Options: BallBerry, Jarvis, simpleResistance

  # Soil moisture and stress
  num_method: itertive           # Options: iterative, non_iterative
  fDerivMeth: analytic           # Options: analytic, numerical

  # Runoff generation
  spatial_gw: localColumn        # Options: localColumn, singleBasin
  subRouting: timeDlay           # Options: timeDlay, qInstant

See SUMMA documentation for complete list of 30+ decision options.

Input File Specifications#

SUMMA requires several input files, automatically generated by SYMFLUENCE:

Forcing Data (NetCDF)#

File pattern: <domain>_<year><month>.nc

Required variables:

time        : Time coordinate (CF-compliant)
hru         : HRU dimension

airtemp     : Air temperature [K]
pptrate     : Precipitation rate [kg m-2 s-1]
windspd     : Wind speed [m s-1]
SWRadAtm    : Downward shortwave radiation [W m-2]
LWRadAtm    : Downward longwave radiation [W m-2]
spechum     : Specific humidity [kg kg-1]
airpres     : Air pressure [Pa]

Attributes File (attributes.nc)#

Contains HRU-specific spatial attributes:

hruId         : HRU identifier
gruId         : GRU identifier
hru2gruId     : HRU to GRU mapping
downHRUindex  : Downslope HRU connectivity
elevation     : Mean elevation [m]
latitude      : Latitude [degrees]
longitude     : Longitude [degrees]
HRUarea       : HRU area [m²]
tan_slope     : Tangent of topographic slope [-]
contourLength : Contour length [m]
slopeTypeIndex: Slope type (1=flat, 2=sloped)
soilTypeIndex : Soil class index
vegTypeIndex  : Vegetation class index
mHeight       : Measurement height [m]

Trial Parameters (trialParams.nc)#

Calibration parameters with dimensions (hru, variables):

  • Default parameter values

  • Updated during calibration

  • Can be HRU-specific or uniform

Output File Specifications#

SUMMA outputs NetCDF files with user-defined variables.

Standard Output Files#

File pattern: <domain>_<GRU>_<run_suffix>_timestep.nc

Common output variables:

# Snow variables
scalarSWE              : Snow water equivalent [kg m-2]
scalarSnowDepth        : Snow depth [m]
scalarSnowAlbedo       : Snow albedo [-]
mLayerTemp             : Snow layer temperature [K]
mLayerDepth            : Snow layer depth [m]

# Runoff variables
scalarSurfaceRunoff    : Surface runoff [m s-1]
scalarRainPlusMelt     : Rain plus melt reaching ground [kg m-2 s-1]
averageInstantRunoff   : Instantaneous runoff [m s-1]
averageRoutedRunoff    : Routed runoff [m s-1]

# Soil variables
mLayerVolFracWat       : Volumetric soil moisture [-]
mLayerMatricHead       : Matric head [m]
scalarAquiferStorage   : Aquifer storage [m]

# Energy balance
scalarLatHeatTotal     : Total latent heat flux [W m-2]
scalarSenHeatTotal     : Total sensible heat flux [W m-2]
scalarNetRadiation     : Net radiation [W m-2]

Control Output Variables#

Specify in outputControl.txt or set all outputs:

# Output all variables at every timestep
! varName       | outFreq | instant | sum | avg | var
*               | 1       | 1       | 0   | 0   | 0

Model-Specific Workflows#

Basic SUMMA Workflow#

# config.yaml
DOMAIN_NAME: my_basin
HYDROLOGICAL_MODEL: SUMMA

# Spatial discretization
DOMAIN_DEFINITION_METHOD: semidistributed
POUR_POINT_COORDS: [-115.0, 51.0]
DISCRETIZATION:
  elevation_bands:
    n_bands: 5

# Forcing data
FORCING_DATASET: ERA5
FORCING_START_YEAR: 2015
FORCING_END_YEAR: 2020

# SUMMA configuration
SUMMA_INSTALL_PATH: /path/to/summa
SUMMA_EXE: summa_sundials.exe

SUMMA_DECISION_OPTIONS:
  snowLayers: CLM_2010
  vegeParTran: CLM_NOAHMP
  spatial_gw: localColumn

Run the workflow:

# Full workflow
symfluence workflow run --config config.yaml

# Or step-by-step
symfluence workflow steps setup_project calibrate_model --config config.yaml

Snow-Focused Application#

For detailed snow modeling:

SUMMA_DECISION_OPTIONS:
  snowLayers: CLM_2010          # Multi-layer snow (up to 5 layers)
  snowDenNew: hedAndPom         # Fresh snow density parameterization
  snowRMelt: variableLiquidHold # Variable liquid water holding
  snowCompactMethod: anderson76 # Snow compaction

# Calibrate snow parameters
PARAMS_TO_CALIBRATE: "Fcapil,upperBoundHead,aquiferBaseflowRate"

# Output detailed snow variables
# (Configure in outputControl.txt or use all variables)

Glacier Mode Workflow#

For glacier-dominated basins:

# Enable glacier preprocessing
SETTINGS_SUMMA_GLACIER_MODE: true

# Glacier-specific attributes will be generated
SETTINGS_SUMMA_GLACIER_ATTRIBUTES: attributes_glac.nc
SETTINGS_SUMMA_GLACIER_COLDSTATE: coldState_glac.nc

# Use glacier-suitable decision options
SUMMA_DECISION_OPTIONS:
  snowLayers: CLM_2010
  alb_method: varDecay      # Variable decay for glacier ice albedo

Parallel SUMMA for Large Domains#

For continental-scale or high-resolution applications, SUMMA can run GRU subsets in parallel through a specified backend. The slurm backend submits job arrays, while the local backend launches multiple SUMMA subprocesses on the current machine.

SLURM backend:

# Enable parallel SUMMA
SETTINGS_SUMMA_USE_PARALLEL_SUMMA: true
SETTINGS_SUMMA_PARALLEL_BACKEND: slurm
SETTINGS_SUMMA_PARALLEL_EXE: summa_actors.exe
SETTINGS_SUMMA_PARALLEL_PATH: /path/to/parallel/summa

# Configure parallel execution
SETTINGS_SUMMA_GRU_COUNT: 1000      # Total GRUs
SETTINGS_SUMMA_GRU_PER_JOB: 10      # GRUs per job
SETTINGS_SUMMA_CPUS_PER_TASK: 32    # CPUs per task
SETTINGS_SUMMA_MEM: 16              # GB memory per task
SETTINGS_SUMMA_TIME_LIMIT: "04:00:00"  # 4 hours

Local backend:

# Enable parallel SUMMA with local backend
SETTINGS_SUMMA_USE_PARALLEL_SUMMA: true
SETTINGS_SUMMA_PARALLEL_BACKEND: local

# Configure parallel execution
SETTINGS_SUMMA_CPUS_PER_TASK: 8

Calibration Strategies#

Parameter Bounds#

Typical ranges for common parameters:

# Routing parameters
routingGammaShape:  [1.0, 5.0]
routingGammaScale:  [1.0, 500.0]

# Vegetation parameters
summerLAI:          [0.5, 10.0]
winterLAI:          [0.01, 5.0]
heightCanopyTop:    [0.5, 30.0]  # meters

# Soil and groundwater
aquiferBaseflowRate: [0.001, 1.0]  # mm/day
aquiferBaseflowExp:  [1.0, 10.0]
fieldCapacity:       [0.05, 0.5]   # fraction
Fcapil:              [0.01, 1.0]

Multi-Objective Calibration#

For comprehensive performance:

OPTIMIZATION_ALGORITHM: NSGA2  # Multi-objective algorithm

OPTIMIZATION_METRICS:
  - KGE
  - NSE_log
  - RMSE

OPTIMIZATION_MAX_ITERATIONS: 500
OPTIMIZATION_POPULATION: 100

Known Limitations#

  1. Computational Cost:

    • Detailed physics = slower execution

    • Multi-layer snow increases compute time 2-3x

    • Large domains (>1000 HRUs) benefit from parallel SUMMA

  2. Initial Conditions:

    • Cold start requires spinup (typically 1-2 years)

    • Soil moisture initialization critical for arid regions

    • Snow states need careful initialization in seasonal snow zones

  3. Data Requirements:

    • Requires 7 meteorological variables

    • Hourly forcing recommended for snow applications

    • Missing radiation data degrades energy balance accuracy

  4. Model Stability:

    • Very small HRUs (<0.01 km²) can cause numerical issues

    • Extreme parameter values may cause non-convergence

    • Some decision option combinations untested

  5. Output Size:

    • Full output can be large (GB per year for 100 HRUs)

    • Use selective output variables for storage efficiency

Troubleshooting#

Common Issues#

Error: “SUMMA executable not found”

# Solution: Verify installation path and executable name
SUMMA_INSTALL_PATH: /absolute/path/to/summa/bin
SUMMA_EXE: summa_sundials.exe

Error: “Simulation failed - NaN values”

Causes and solutions:

  1. Extreme parameter values → Check calibration bounds

  2. Insufficient spinup → Increase SIMULATION_START_YEAR

  3. Bad forcing data → Validate forcing with ncview or ncdump

  4. HRU connectivity issues → Set SETTINGS_SUMMA_CONNECT_HRUS: false

Error: “NetCDF dimension mismatch”

# Check forcing file structure
ncdump -h forcing_file.nc

# Verify HRU dimension matches attributes file
ncdump -h attributes.nc | grep "hru ="

Slow execution

Solutions:

  1. Enable parallel SUMMA for >100 HRUs

  2. Reduce output frequency

  3. Use fewer snow layers (snowLayers: jrdn1991 instead of CLM_2010)

  4. Disable HRU connectivity if not needed

Missing output variables

# Check outputControl.txt
cat <project_dir>/settings/SUMMA/outputControl.txt

# Use wildcard for all variables
echo "* | 1 | 1 | 0 | 0 | 0" > outputControl.txt

Performance Optimization#

For faster calibration:

  1. Use simple decision options during calibration:

    SUMMA_DECISION_OPTIONS:
  snowLayers: jrdn1991      # Single-layer snow (faster)
  num_method: non_iterative # Explicit scheme

  2. Reduce output variables during calibration

  3. Use coarser temporal resolution for large-scale studies

For accurate physics:

  1. Use detailed options:

    SUMMA_DECISION_OPTIONS:
  snowLayers: CLM_2010      # Multi-layer snow
  num_method: iterative     # Implicit scheme

  2. Hourly forcing data

  3. Full energy balance variables in output

Additional Resources#

SUMMA Documentation:

SYMFLUENCE-specific:

Example Configurations:

# View example SUMMA configuration
cat $SYMFLUENCE_ROOT/resources/config_templates/examples/config_basin_notebook.yaml

Publications: