PyCatch

Copyright Derek Karssenberg & Noemí Lana-Renault Monreal

User manual (draft)

Copyright of this software: Derek Karssenberg & Noemi Lana-Renault Monreal Contact: [email protected] or [email protected]

Installation

Create a conda environment for PCRaster (for instance by running pcraster_pycatch.yaml)

To run the model with 1 h timestep, run main.py

To run the model with 1 week timestep, run main_weekly.py (this model includes erosion, not documented yet)

To run the model with 10 second timestep, run main_seconds.py

Configuring the models

To remove all output, run clean.sh, this does NOT remove inputs, so there is no major risk, but check what is in clean.sh first

Settings are in configuration.py (1 h model) and configuration_weekly.py (1 week timestep model)

Instructions for hybrid modelling

Use the main_weekly.py model. Settings can be configured in configuration_weekly.py.

Realizations of the model are executed parallel (forked). Each realisation is written to a subdirectory. The number of realizations can be defined in configuration_weekly.py. Each realisation of the model will be different as there is some tiny random noise included which results in different paths of how the system evolves.

The model calculates grazing pressure inside the dynamic method. It currently gradually increases grazing pressure then decreases it. Output is written to 1/grazing.npy, 2/grazing.npy,.. etc (all the same as there are no differences between realizations.)

The model calculates and reports soil depth and biomass. These are written as txt files in the folders 1, 2, 3, etc again

The number of timesteps (time step duration is one week) is given in configuration_weekly.py

For displaying some map outputs, use setOfVariablesToReport = 'some' in configuration_weekly.py, it writes biomass and grazing pressure for a defined interval. These can be displayed using the PCRaster Aguila visualisation tool.

To do's PyCatch hourly model

• Use correct geographical coordinates
• Check time alignment (precip)
• Consider using solar inclination and topography (it now only uses shading effect)
• Why is Rq max value greater than 100? (on output maps) This is the same in original model (more or less)
• Ground water
  • Check the ldd's for both soil and ground water layer
  • Consider using slope in water for lateral flow (for the second layer)
  • Check whether model with 1 cm groundwater gives same output as original model
• Erosion is too high. This is partly due to too high rain drop impact, without raindrop impact however (only wash detachment) it is still too high. Transport capacity also may be too high. Need to check parameters etc. Spatial pattern however makes sense. Or consider calling the erosion component once in the so many days

Groundwater layer

All new things in files with extension _gg

Geometry:

• The first x m is always first layer ('soil'), x is about 1 m
• Below x, the groundwater layer starts up to 'bed rock'
• If the regolith thickness is below x, the groundwater layer is made very thin (0.01 m?)

Lateral flow:

• Fixed local drain direction (same as surface), i.e. not depending on water levels
• Gradient equal to slope of surface or equal to slope of groundwater surface (to be decided)
• Use saturated conductivity (same as first layer)
• Note that the first layer will still have lateral flow as well (otherwise there would not be 'interflow' on hillslopes

Upward seepage:

• Seepage is all the water above the maximum storage capacity of the layer
• Pass it on as 'recharge' of the top layer (i.e. like rain)

Percolation from top layer:

• Amount is up to field capacity
• See van Beek (2012), van Beek (2009). Van Beek 2009 eq 5, but use instead of 'residual moisture content
• Add to the groundwater layer (always - it will be removed again through upward seepage)

Capillary rise (it could also be left out for now?, it will be zero with gw depth > 5 m or so):

• See van Beek (2009), eq. 9, but use instead of 'residual moisture content the field capacity (this is the same?)', no capillary rise if gw depth > 5 m (or better use linear function, between zero and five m depth to groundwater

Evapotranspiration:

• Only occurs from the top layer

Data needed for PyCatch hourly model

Refer to paper for background.

If the surface water erosion is added some more inputs are needed (most of them uniform values most likely).

The inputs directory gives these maps for the small catchments in Spain.

Maps needed in addition to DEM:

• mergeFieldCapacityFractionFS.map # derive from geology
• mergeLimitingPointFractionFS.map # derive from geology
• mergePorosityFractionFS.map # derive from geology
• mergeStream.map # pixels with a stream (can also be calculated from the upstream area)
• mergeVegAlbedoFS.map # derive from vegetation classes map or remote sensing products, fixed over time (assumption)
• mergeVegHeightFS.map # idem albedo
• mergeVegLAIFS.map # idem albedo
• mergeVegStomatalFS.map # idem albedo
• mergeWiltingPointFractionFS.map # derive from geology

Timeseries needed:

Could be for each variable a single timeseries for the whole catchmenlt (to start with). Consider using https://chelsa-climate.org, currently a state of the art reanalysis data set. We need at least 2 years.

• airTemperatureArnaJulAugSep0506.tss
• incomingShortwaveRadiationArnasJulAugSep0506.tss
• rainfallFluxTwoCatchsJulAugSep0506.tss
• relativeHumidityArnasJulAugSep0506.tss
• windVelocityArnasJulAugSep0506.tss

Output file name conventions

first letter:

biomass or exchange vars starting with       X (Xst, biomass)
precipitation files starting with            P
interception files starting with             V (from vegetation)
surface store files starting with            S
infiltration files staring with              I
evapotranspiration files starting with       E
subsurface store files starting with         G (from groundwater)
runoff files starting with                   R
shading files starting with                  M (from shading due to Mountains )
budgets files starting with                  B
soil wash files starting with                W
regolith files starting with                 A (Ast, regolith depth (=soil depth))
bedrock weathering files starting with       C
base level files starting with               L
creep files starting with                    D

second letter:

store (s, unit m)

actual flux in (i, unit m/h, for geomorphology m/year)
potential flux in (j, unit m/h, for geomorphology m/year)

actual flux out (o, unit m/h, for geomorphology m/year)
potential flux out (p, unit m/h, for geomorphology m/year)

actual flux in (in (positive) or out(negative)) (c, of change, unit m/h, for geomorphology m/year)
another flux (x, unit m/h, for geomorphology m/year)

lateral flux (q, cubic metre per hour)

two extra letters: other values, eg Ecl, cloud factor

see respective classes for details !! E.g. actual flux out is not always all out fluxes but sometimes only one of two.

For particle filtering (not yet tested)

To create observed discharge, run createObservedDischarge.sh, it creates maps in the folder 'observations'. This is only required for particle filtering

todo and changes are in changes.txt at the bottom of changes is also the names of files used

file name conventions

# reports as numpy arrays

Got.npy   self.totalActualAbstractionInUpstreamAreaCubicMetrePerHour, from subsurface
Gxt.npy   self.totalUpwardSeepageInUpstreamAreaCubicMetrePerHour,from subsurface
Vot.npy   self.totalActualAbstractionInUpstreamAreaCubicMetrePerHour, from canopy
Rq.npy    discharge m3/h
Rqs.npy   discharge m3/h, averaged over 2 hours

RPic.npy  self.maximumInterceptionCapacityPerLAI,
RPks.npy  self.ksat,
RPmm.npy  self.multiplierMaxStomatalConductance
RPrt.npy  self.regolithThicknessHomogeneous,
RPsc.npy  self.saturatedConductivityMetrePerDay,


Removed code
------------

Early warning signals

Code to calculate statistics in the dynamic section. Removed as it was hard to get GSTAT running. Note also that it considerably slows down the model.