How to consistently adapt soil parameters to express urban growth in physically based precipitation modeling ?

Atmospheric limited-area models are superb tools built by atmospheric scientists, and can also be used by scientists from other disciplines. As hydrologists interested in urban rainfall hazard, we want to study possible changes in local-scale precipitation intensities and patterns under urban growth scenarios.

Unfortunately, the parameterization of ground properties appears scattered in many datasets. These differ by their spatial resolution, computational type (exclusive categories expressed as integers, categories expressed as percentages in the patchwork/tile approach, continuous parameters as real numbers, month-dependent real numbers), and of course by their semantic (land use/land cover, radiative properties such as LAI according to one or another sensor, orography, soil type according to one or another research institute).

From the above, the basic way to deal with expected land use changes in impact simulation changes would involve reading the scientific literature exhaustively - literally: to the point of exhaustion - to establish which parameter must be changed, and to hope that no inconsistencies will be introduced in the individual values or in their interdependence.

We propose another, easier, and above all safer strategy. The first step is to recognize the "ground properties" are not a list of individual parameters, but a compound object where many parameters are related in a hierarchy of aspects  : parameters related to land use, parameters related to orography, etc. The determination of this hierarchy is quite easy using multivariate statistics, individuals being locations sampled in the domain of interest and data being the parameters values at these locations. This approach helps to establish the list of parameters connected to the intended change.

Armed with this list, a "geographic cut-and-paste" strategy can be safely adopted to express intended land use change: the relevant parameter values of a representative (donor) location will be used at the target (modified) location, while leaving all other local parameters untouched.

We illustrate this approach with the specific case of prescribing variable levels of urban development for the city of Querétaro, Mexico, in the technical context of using WRF's UEMS distribution (89 datasets distributed as 25633 files distributed in 219 directories).