Micro Scale Process
Micro-scale processes are the atmospheric processes small enough that individual obstacles matter individually — roughly metres to a kilometre horizontally, seconds to an hour in time. The defining feature is not simply that they are small, but that at this scale the surface stops being describable by aggregate parameters. A specific building casts a specific shadow and generates a specific wake, and neither is recoverable from the mean building height of the neighbourhood.
That is what separates micro-scale from meso-scale work. A meso-scale model represents a city through effective roughness, albedo, and heat capacity — statistical descriptions of a surface it does not resolve. A micro-scale model resolves the geometry and lets the flow respond to it.
Where the distinction has consequences
Thermal exposure. Within a single street, sunlit and shaded pavement can differ by more than 20 K in surface temperature, and the mean radiant temperature governing human heat stress varies over a few metres. Any grid cell large enough to contain the street reports one number, and that number describes no location a person could stand in. For heat-stress questions this is disqualifying rather than merely imprecise.
Air quality near sources. Pollutant concentrations fall off sharply from a road, and street-canyon circulation can hold emissions against one facade while the opposite side stays comparatively clean. Concentrations at breathing height depend on canyon aspect ratio and wind direction relative to the street axis — geometry that only exists in a resolved representation.
Wind comfort and building wakes. Downwash around tall buildings produces pedestrian-level wind speeds that can exceed the approach flow. This is entirely a resolved-geometry effect and has no meso-scale analogue.
Intervention design. This is where it becomes practical. Questions like where to plant trees, which facade to shade, and whether a courtyard ventilates are all micro-scale by construction. A meso-scale model can say a city is warm; it cannot say which intervention would help, because the intervention is smaller than its grid cell.
The cost
Resolving this scale means large-eddy simulation at metre resolution over domains of at most a few square kilometres, with time steps well under a second. That is expensive enough that micro-scale runs are episodes, not climatologies — a few characteristic days, not a decade. PALM-4U is the tool I have used for this.
The consequence is a coupling problem rather than a choice between scales. The micro-scale domain needs boundary conditions from something larger, and the selection of which episodes to run determines what the results represent. Getting that selection wrong is a subtler error than getting the physics wrong, and harder to notice.
See also: urban atmosphere for the canopy-layer context, urban climate for the patterns micro-scale processes aggregate into, and dynamical downscaling for the scale-bridging problem in general.