Urban-PLUMBER

A multi-site model evaluation project for urban areas

Phase 2 Phase 1 Download project protocol
Mathew Lipson (UNSW), Sue Grimmond (Reading), Martin Best (Met Office),
with observational and modelling participants.

Questions and Answers

This page records questions or clarifications for the project.

Spinup

Q: Is spinup compulsory?

A: The protocol recommends 10yr spinup to allow model soil state to equilibrate with local climate conditions, however this may not be a) useful or b) feasible for some models. If this is the case you may choose a reduced spinup. The submission checking system (on modelevaluation.org) will still expect a netCDF over the full forcing period (spinup + analysis, so for Preston starting 1993-01-01). If some/all spinup is not undertaken, submitted netcdf should still begin at start of spinup period, with non-simulated intervals being filled with the missing_float value (set as -9999. in the provided script). This will satisfy the automatic check on submission, and will record for us whether models undertook spinup. The primary analysis will still begin at “time_analysis_start” (2003-08-12 for Preston).


Q: Should automatic (repeating) spinup be used?

A: Repeated spinup is not necessary. The spinup and analysis period should be run together without repetition.


Site data parameters

Q: what is “impervious_area_fraction” (site parameter 5)

A: impervious_area_fraction is the total plan area of surfaces which are “hard”, or not water permeable. Roof + road + footpaths = impervious_area_fraction


Q: For models without explicit urban representation, what should “impervious_area_fraction” be?

A: Use whatever land type the model would normally apply to urban land points in a typical regional simulation (some use bare soil, some use grassland etc.). In the “detailed” experiment, surface albedo could be altered to match more closely with “average_albedo_at_midday”.


Q: What type of tree is “tree_area_fraction”?

A: Baseline: Use default tree type, or plant functional type (PFT) for that geographic location (if your modelling system has that capability). Detailed: Use tree type(s) you think best represents the location.


Q: what is “footpath_area_fraction” (site parameter 12)

A: footpath_area_fraction could have been better named as “paved_area_fraction” or similar, as it is intended to include paths, courts, plazas, parking areas and any other hard ground coverings that are not roads.


Q: What are the assumptions I should use for the baseline?

A: Where parameters are not provided in the baseline case (e.g. building height), models should use their default values, or values which modellers believe represent a generic, or average urban surface. These default values will be used in all baseline cases in the multi-site stage.


Requested variables

Q: What is ACond: Aerodynamic Conductance?

A: This is a “mandatory” output from the ALMA protocol, so is included in this protocol. ALMA describes ACond as:

“Aerodynamic conductance to vapor transport, averaged over all vegetation types within the grid cell.”

Aerodynamic conductance is the inverse of aerodynamic resistance. As ALMA discusses “all vegetation types”, the Urban-PLUMBER protocol asks for vegetation canopy aerodynamic conductance only (i.e. not conductances over urbanised surfaces). The vegetation canopy aerodynamic conductance should be the area-weighted average if multiple vegetation types exist as subgrids.


Submission

Q: In create_netcdf.py many of variables are not generated by my model. Should I delete them from .py script or left empty?

A: Yes, you can delete unused variables from they .py script (delete from both set_netcdf_data() and create_empty_netcdf() sections), but please include the following variables as a minimum: SWnet, LWnet, Qle, Qh, Qanth, Qstor.


Q: Is possible to have the raw observations to compare the outputs of our model?

A: The analysis component of observations are not directly distributed as part of the agreement with data owners, however you are able to get comparison with observations if you upload your output to modelevaluation.org. On upload you will see plots comparing your output with observations, along with some basic error metrics. You are able to re-upload/replace your output at any time up until November.