Dynamical downscaling

Dynamical downscaling is a technique for producing high-resolution regional climate information by nesting a limited-area atmospheric model within coarser-resolution global data — either a reanalysis product or a general circulation model (GCM). The regional model solves the full equations of atmospheric motion at a finer grid spacing, allowing it to resolve meso-scale processes that the driving data cannot represent.

Methodology

The approach involves three core steps:

1. Lateral boundary conditions: The global dataset provides time-varying atmospheric fields (temperature, humidity, wind, pressure) at the boundaries of the regional domain. These fields constrain the large-scale circulation within the nested model.

2. Physics parameterization: The regional model applies its own suite of physical parameterizations — cumulus convection, planetary boundary layer, microphysics, radiation, and land surface schemes — tuned to the resolution and climate regime of the target domain.

3. Nesting: Most setups use one-way nesting, where the global model drives the regional model but receives no feedback. Multiple nesting levels (parent-child domains) progressively increase resolution. Typical nesting ratios are 3:1 or 5:1 per level.

The WRF model

The Weather Research and Forecasting (WRF) model, most commonly used with the ARW dynamical core, is one of the most widely used tools for dynamical downscaling. It is a fully compressible, nonhydrostatic model that supports terrain-following vertical coordinates, Arakawa C-grid staggering, and a wide range of physics options.

The choice of physics configuration strongly influences downscaling quality. For tropical island environments like the Galapagos, the tropical WRF setup uses the New Tiedtke cumulus scheme, WSM-6 microphysics, RRTMG radiation, YSU PBL, and the Noah land surface model.

Added value

Dynamical downscaling adds value over the driving data primarily through:

• Orographic effects: Resolving terrain-forced Precipitation, valley winds, and slope flows that coarse grids cannot represent
• Land-sea contrasts: Capturing coastal circulations, sea breezes, and island effects
• Mesoscale convective organization: Representing organized convection at grid spacings approaching the convection-permitting scale (~3 km)
• Surface heterogeneity: Better representing urban-rural contrasts, vegetation gradients, and soil moisture feedbacks

However, dynamical downscaling inherits systematic biases from the driving data and can introduce additional errors through imperfect physics parameterizations, domain boundary effects, and spin-up artifacts.

Applications in my research

The DARWIN project (Schmidt et al., 2025, Int. J. Climatol.) applied dynamical downscaling of ERA5 with WRF to produce the Galapagos refined analysis. That work resolved the vertical Precipitation gradients, Garua fog dynamics, and ENSO-driven rainfall variability that are invisible in coarser global products.

Earlier work on the Qaidam Basin (Wang, Schmidt et al., 2021, JGR-Atmospheres) used dynamical downscaling to study water balance sensitivity under Mid-Pliocene climate conditions, demonstrating the method’s applicability to paleoclimate questions.

The Central Europe Refined analysis version 2 (CER v2) applied the same methodology to produce three decades of high-resolution Precipitation data for the Berlin-Brandenburg metropolitan region.

See also: ERA5 reanalysis, Regional climate modeling, WRF