How to best model within canyon air temperature

General
1

Hi,

Would it possible to get some guidance for how to setup the model to model intra-urban Ta variations. We (me and Janka) see very small changes in T2 when we change various vegetation parameters. We suspect that it has to do with physics settings i.e. profile scheme etc. Would it be possible to get some info on best practice for this? /Fredrik and Janka

Question on SUEWS sensitivity to vegetation parameters
2

Good question, Fredrik — and yes, the physics settings matter a lot here. The small T2 sensitivity is likely not a bug but a consequence of how SUEWS constructs the temperature profile. Let me walk through what’s happening and how to get better results.

How T2 is computed

T2 in SUEWS is not a direct output of the energy balance. Instead, the model builds a vertical temperature profile from the surface up and then interpolates to 2 m. The profile shape depends on T_* (the temperature scale from Q_H), u_*, and the Obukhov length L, with stability corrections applied via \psi_h(z/L).

The key question is which profile scheme is used — and this is where rslmethod matters.

rslmethod and its effect on T2

With rslmethod: 2 (which Janka’s config uses), SUEWS auto-selects between two schemes:

  • RSL (roughness sub-layer) — a two-layer profile with an exponential within-canopy decay and MOST + RSL corrections above canopy. T2 is interpolated at 2 m above ground level
  • MOST (Monin-Obukhov similarity) — a simpler log-linear profile. Here, T2 is interpolated at 2 m + z_d + z_0 above ground — i.e. displaced upward

The auto-selection logic requires all of: plant area index 0.1 < \mathrm{PAI} < 0.68, and canopy height z_H > 2\,\text{m}. If these conditions aren’t met, SUEWS falls back to MOST.

Why this causes low T2 sensitivity

If your grids are falling back to MOST mode, the effective “2 m” measurement height becomes 2 + z_d + z_0, which in a dense urban area can be 6–10 m above ground. At that height, the temperature profile is well-mixed and barely responds to surface-level changes in vegetation parameters. Changing tree cover or conductance alters Q_E and Q_H, but by the time you interpolate the profile at ~8 m effective height, the signal is heavily damped.

Even in RSL mode, the coupling pathway is indirect:

\text{vegetation} \;\to\; Q_E/Q_H \;\text{partition} \;\to\; Q_H \;\to\; T_* = \frac{-Q_H}{\rho c_p \, u_*} \;\to\; \text{profile shape} \;\to\; T_\text{2m}

The profile shape scales with T_*, so T2 only responds if Q_H changes meaningfully. Vegetation parameters primarily affect Q_E — and while Q_H = Q^* - Q_S - Q_E does respond, the response is diluted by the other terms.

Practical recommendations

  1. Check whether RSL is activating — look at the output PAI and canopy height for each grid. If PAI falls outside 0.1–0.68 or z_H \leq 2\,\text{m}, you’re in MOST mode and T2 will be insensitive. You can force RSL with rslmethod: 1, but only if the canopy geometry supports it

  2. Compare \Delta Q_H first — before looking at T2, check whether Q_H actually differs between your runs. If \Delta Q_H is small, \Delta T_2 will be small regardless of the profile scheme

  3. Consider \Delta T_s (surface temperature) — for cooling efficiency studies, the surface temperature or the canyon air temperature difference relative to a reference grid may be more diagnostic than T2, which is inherently smoothed by the profile interpolation

  4. RSL level setting — rsllevel: 1 in Janka’s config means diagnostic output at a single level. If you want the full vertical profile for analysis, check the RSL output options

For the cooling efficiency study specifically, working with \Delta Q_H or \Delta T_s may be more physically meaningful than \Delta T_2 — the profile interpolation inherently dampens near-surface contrasts, especially in MOST mode.