Phase H5b — temperature through the LID layer stack. More...

#include <cstddef>

Include dependency graph for HeatLid.hpp:

This graph shows which files directly or indirectly include this file:

Namespaces
namespace	openswmm

namespace	openswmm::lid

namespace	openswmm::transport

Functions
void	openswmm::transport::initLidLayerTemperature (SimulationContext &ctx, const lid::LIDSolver &solver)

void	openswmm::transport::setLidInflowTemperature (SimulationContext &ctx, int type_index, int unit, int subcatch, double rain_rate, double q_imperv, double q_perv, double q_runon, double lid_area)

void	openswmm::transport::routeLidLayerTemperature (SimulationContext &ctx, const lid::LIDSolver &solver, double dt)

double	openswmm::transport::lidLayerHeatCapacity (const SimulationContext &ctx, const lid::LIDGroupSoA &g, std::size_t unit, int layer) noexcept
	Heat capacity of one layer, J/m²/K over the layer's footprint.

Detailed Description

Phase H5b — temperature through the LID layer stack.

A4 built LidLayerSpeciesState with a species stride for exactly this. Temperature is row 1; nothing about the layer indexing changes. What differs from WaterAgeLid is the same three things HeatWatershed differs from WaterAgeWatershed by — no aging term, no zero floor, and a dry layer takes the D-H5c policy rather than A4's = 0 — plus one that is new here: the column exchanges heat with itself.

The thermal step is COLUMN-COUPLED, and that is the call to challenge: Conduction couples adjacent layers, so it is a second operator on the same state. Lesson 80 says an operator applied sequentially after a relaxation does not compose — that is the defect D-H5e just fixed between SurfaceExchange and RadiativeExchange, and applying conduction as its own pass would reproduce it inside one phase.

So the whole column is stepped at once: atmospheric flux on the exposed layer, conduction on every adjacent pair, solved implicitly as one tridiagonal system.

Two alternatives were measured and rejected:

Per-layer relaxation with frozen neighbours is stable, but the energy leaving a layer is not the energy its neighbour receives, so the column leaks. On a thin surface film relaxing fully toward its equilibrium with the soil in one step, the leak is the whole exchange.
Explicit conduction conserves energy exactly but is stiff where it matters. A 1e-4 m surface film against a 0.3 m soil layer gives k·dt = 1.91 at a 60 s step — the same regime that produced H5a's NaN. Not "probably fine": computed, with the module's own capacity and conductivity expressions. The same column with the soil layer alone gives k·dt = 5.3e-4, which is why conduction looks harmless until a film appears.

Adjacency comes from the DONOR MAP, not a second table: WaterAgeLid's donorsFor already encodes which layer feeds which, per LID type, and for a vertical stack that IS the physical adjacency. Writing a second stacking table would be a third copy of the same topology, which is the shape of lesson 81.

See also: plans/transport/HEAT_TRANSPORT_PLAN.md §6 H5b, §6.1 D-H5b/D-H5c; src/engine/transport/components/WaterAgeModule/WaterAgeLid.hpp

Author: Caleb Buahin caleb.nosp@m..bua.nosp@m.hin@g.nosp@m.mail.nosp@m..com

License\n Apache-2.0

Namespaces

Functions

Detailed Description