Recently, much attention has been focused on the land- atmosphere interface in the context of numerical weather prediction and climate modeling. Models representing the land surface in atmospheric models are typically known as Land Surface Parameterizations (LSPs) or Soil-Vegetation-Atmosphere Transfer Schemes (SVATS). Most LSPs used in atmospheric models were developed from the perspective of their host models to represent the "essential" processes of momentum, heat and mass (i.e. water) transfer at the land-atmosphere interface. Typically, these LSPs are applied at the spatial and temporal scales of their host models without consideration of the inherent spatial and temporal scales of hydrological processes.
This spatial and temporal scale disparity is largely an issue of convenience in coupling the LSP with the host model. Due to heterogeneities of the land surface (topography, soil, vegetation), and the time scale of moisture transport, the natural spatial scale for hydrology is 100 meters or less, while the temporal scale is much slower, being minutes to hours. Meterological meso-scales are quite different, on the order of 10's of kms in the horizontal together with a time-scale on the order of 10's of seconds. Thus, physically, there is no reason to apply a hydrological model at a 10 km or larger grid resolution and a 60 second or smaller temporal resolution. In addition, the fundamental control on water flow is gravity, which leads to the fundamental unit of hydrology as the catchment (or watershed). These concepts are missing from the current generation of LSPs, and therefore modelers are unable to verify their water budgets by taking advantage of the best available data for that purpose-- streamflow data.
There are two critical issues which must be addressed when redesigning LSPs from a hydrological perspective:
1. The LSP must be efficient in its computations and operate at its own spatial and temporal scales; and
2. The coupling between the LSP and host model must consider the problem of disaggregation and aggregation of fluxes of momentum, heat and mass.
To address issue 1. above, we will show that by taking advantage of similarities in hydrological behavior, a new version of the TOPLATS hydrological model (SPARSE-TOPLATS) can be applied with computational performance approaching that of a much simpler LSP while retaining the complex soil-topographical-hydrological details of the original TOPLATS.
To address issue 2. above, we have designed a model-coupling interface approach which consists of: (1) a drop-in MM5V2 module that provides selective direct access variable-by-variable MM5V2 outputs; (2) a drop-in MM5V2 module which read TOPLATS fluxes and aggregates them for use by MM5V2; (3) drop-in TOPLATS modules that perform mirror functions (output to MM5V2 and disaggregation of MM5 data for use by TOPLATS); (4) a PVM-based interface that allows the two models to coordinate with each other and exchange data, while retaining their own fundamental spatio-temporal physical and computational characteristics.
Combined, we are using these tools to couple TOPLATS, which simulates the land-surface at the fundamental hydrological space-time scales, to be coupled to the MM5V2 mesoscale model. We will first describe intermediate applications of the building blocks -- SPARSE-TOPLATS, MCPL(), TCPL() and the PVM extension to our I/O API, and then discuss early results from the coupled model system, including 1-D column simulations of the 1994 Little Washita Watershed IOP. Finally, we will describe how these building blocks are extended to data-assimilation mode, whereby GOES solar radiation and NEXRAD Stage IV precipitation are used to drive the coupled model system at the land-atmosphere interface, and how this type of approach can enable modelers to better reconcile modeled water-budget calculations with observations.