|Simulations of Regional Climatic Patterns which Impact the Florida Bay Water Cycle|
Craig A. Mattocks
University of Miami/CIMAS
Hurricane Research Division, NOAA-AOML
4301 Rickenbacker Causeway
Miami, Florida 33149 -USA-
Paul Trimble, Matthew Hinton, Beheen Trimble, Marie Pietrucha
South Florida Water Management District
3301 Gun Club Road
West Palm Beach, FL 33406 -USA-
This project directly addresses the first central question articulated in the Strategic Plan of the Interagency Florida Bay Science Program, namely:
"How, and at what rates, do persistent and/or catastrophic storms alter freshwater input (via their associated local evaporation/precipitation patterns), thereby inducing changes in the circulation and salinity/nutrient content of Fl orida Bay?"
The approach is to employ the Center for Analysis and Prediction of Storms' (CAPS) Advanced Regional Prediction System (ARPS) cloud-/mesoscale atmospheric numerical weather prediction model, coupled to the SFWMD's hydrologic models, to simulate persistent , locally-forced weather regimes (land/lake/urban heat island breeze circulations) which generate thunderstorm complexes over the Everglades and coastal areas that account for roughly one-third of Florida's annual rainfall (Burpee and Lahiff, 1984) (Figure 1). Besides providing precipitation and high-resolution surface winds for use as boundary conditions/forcing in bay and ocean circulation models, the atmospheric model's surface energy parameterization allows the predictio n of evaporation - at the ground surface, from the fraction of foliage covered by intercepted rainfall, and from transpiration by leaves.
Numerous meteorological modeling studies (Pielke et al., 1998; Bougeault et al., 1991; Pielke, 1974; Gannon, 1978; McCumber and Pielke, 1981) have demonstrated that landscape discontinuities (including anthropogenic modifications due to comm
ercial development, agriculture, and water management practices) can generate strong mesoscale circulations which impact the spatial and temporal distribution of regional scale rainfall under certain synoptic scale meteorological conditions. Small scale h
eterogeneities in the soil moisture, surface albedo and thermal inertia are the most dominant controlling factors. These surface variations can alter the speed and intensity of the sea breeze convergence zone and shift the location of the attendant thunde
rstorm formation over the Florida peninsula.
Horizontal variations in vegetation profoundly impact evapotranspiration. To the atmosphere, plants act like "pipes" which pull underground water out of the soil reservoir. The upward moisture transfer cools the surface and moistens the air in the planeta ry boundary layer (PBL). This evaporation absorbs energy and buffers the surface against further solar heating, damping the strength of the thermal/solenoidal circulations. Intricate concave/convex geographical patterns in the surface moisture gradients c an generate mesoscale winds as strong as sea breezes. According to Lynn et al. (1998), when the wet/dry soil patch length is similar in size to the local radius of deformation in the presence of large background wind shear, the mesoscale circulatio ns generated by landscape heterogeneities can trigger deep moist convection. In a complex feedback process involving the soil hydrology, the moist convection creates new rainfall "footprints", the largest CAPE (convective available potential energy) evolv es above the wet patches, and the most intense rainfall occurs along the sea-breeze-like fronts induced by the soil moisture gradient in the internal boundary layer. The transient nature of the soil moisture patterns makes the prediction of subsequent con vection difficult. If substantial areas of land remain flooded, convective suppression and shallow "outflow" layers can spawn circulations analogous to lake/river breezes.
Until recently, such numerical studies were severely limited by computer capacity, so the simulations could not include these feedback mechanisms between the hydrologic and atmospheric systems. With the advent of powerful desktop computer workstations, th e simulation of important coupled system processes is now possible. An exact, transparent coupling of the ARPS mesoscale meteorological model with SFWMD's regional scale hydrologic models would involve an extensive redevelopment effort to match their comp lex surface energy budget formulations. This is beyond the scope of the current project. Nevertheless, it is possible to reconfigure the models so they have consistent grid domains and run them in a "loosely coupled" mode - exchanging boundary conditions and fields off-line, rather than in exact lockstep sequence - while integrating the models over a period of a few months. It is then possible to run land use sensitivity experiments (such as present day development vs. pre-colonized natural system conditi ons) and examine periods which exhibit correlations between differences in ponded water and the spatial and temporal distribution of rainfall. One of the most intriguing potential applications of this coupled model approach is the prediction of the enviro nment's response to "What if ...?" anthropogenic impact assessment scenarios, such as the development/urbanization of pristine areas or the restoration of natural habitats.
SUMMARY OF RESULTS TO DATE
In earlier coarse-mesh ARPS model simulations of an August 1975 Florida Area Cumulus Experiment (FACE) sea breeze case, the atmosphere's response to incorporating realistic, modern-day horizontal gradients in the land use were striking. Enhanced diurnal h
eating over heavily developed areas, such as Naples-Fort Meyers and Tampa, induced strong urban heat island circulations which doubled the amount of simulated rainfall over these heavily populated regions. Rapid evaporation/drainage and heating of the por
ous, cultivated land south of Lake Okeechobee caused abrupt divergent deflections of the surface winds over the lake (Figure 2) and generated a thunderstorm complex similar to the convective cells diagnosed in the real data
(Figure 3). The arc-shaped band of maximum rainfall, associated with the lake breeze, shifted from east of Lake Okeechobee to a more realistic location at its southern shore.
The evaporation pattern (Figure 4) correlated well with the strongest divergent and initially driest surface wind fields, in the vicinity of the greatest surface-to-air temperature/humidity differences. Moisture was picked u p by the atmospheric flow over Lake Okeechobee and by the organized offshore downdrafts associated with the west coast sea breeze circulation, while the Florida Everglades "muck" soil in the interior of the state tended to resist evaporation due to its hi gh water retention and strong capillary forces. Thus, a significant north-south mesoscale gradient in evaporation was simulated across Florida Bay. Based on these results, a conceptual model has been developed (Figure 5) whi ch will serve as a working hypothesis to guide future research.
DETAILS OF METHODOLOGY
1. Terrain Generation
As investigators seek to diagnose the essential dynamical processes of coastal phenomena in greater detail by increasing the resolution of their numerical models, the accuracy of the representation of the topography in these models becomes more crucial. M
ost of the techniques commonly used for numerical terrain generation were designed 15 years ago for global or synoptic-scale simulations (with grid mesh sizes larger than 100 km) and involve processing datasets with relatively coarse (global, 5-minute) re
solution. Blindly extending these techniques to high-resolution mesoscale models may generate numerical noise or suppress key features in simulations.
Gridbox binning and averaging, for example, creates oversmoothed model terrain that frequently "slops" offshore, creating erroneous regions of elevated water. Ad hoc truncation of height values to preserve coastal geometry (usually through applicat ion of a land/water bitmask) can generate 2 delta-x grid noise. Depending on the model formulation, this can have a deleterious effect on all terms in the equations of motion which involve the calculation of the terrain gradient (the coordinate transforma tion Jacobian, pressure gradient force, mountain wave drag, turbulent mixing, etc.). The silhouette ("maximum profile") terrain enhancement, used in NMC/NCEP's eta forecast model (Mesinger, 1984), is a subsampling technique which suffers the effects of al iasing. A high-frequency "spike" (perhaps as far as one-half grid box off-center) will be selected and misrepresented as a low-frequency component for an entire grid box, which generates excessive numerical noise. Bilinear interpolation also results in sp urious numerical noise when applied to high-resolution datasets. The Nyquist criterion dictates that at least 2 samples per cycle are needed to resolve a wave properly when data is interpolated, that is, the data must be band-pass filtered so that the hig hest wavenumber remaining in the terrain database has a wavelength longer than 2 grid mesh widths. Otherwise, interpolation results in an undersampling of the data, the high-frequency information is lost, and a fictitious low frequency component is genera ted through aliasing. High-frequency information becomes "folded over" into low-frequency information, making it impossible to reconstruct the original data.
To remedy these defects, a new hybrid topographic representation, produced by combining an envelope "valley filling" scheme (Dell 'Osso, 1984) with an optimum interpolation (OI) objective analysis algorithm, has been developed. Dubbed "envoila" (envelope + OI = en-"Voilą!", being French for "Behold, see, there it is!"), it incorporates the best aspects of all available terrain generation techniques: minimal smoothing, negligible aliasing and numerical noise, physically-based scaling, enhanced low-level b locking and flow-splitting, steep terrain gradients near coastlines, the retention of significant gaps/breaks/channels in complex topography, and reasonable computational efficiency. Terrain height measurements from the 100-meter resolution Defense Mappin g Agency (DMA) datasets are blended with GIS data from the South Florida Water Management District using the envoila scheme to produce a superior rendition of the elevation for the ARPS model grid domain (Figure 6). Fine-sca le features, such as the coral ridge south of Miami and the Florida Keys, are well resolved.
2. Soil and Land Use/Land Cover
In the first phase of development of a hydro-meteorological coupled model, the ARPS model grid has been completely reconfigured to closely match the grid of the Natural System Model (NSM) and the South Florida Water Management Model (SFWMM). The horizonta
l resolution of ARPS has been tripled, from 9 km to 3.22 km (2 miles) on a Mercator projection. High-resolution GIS soil/vegetation, land cover/use surface characteristics, and terrain elevation data from SFWMD has been reprocessed using GIS ARC-INFO "fis
hnet" area-weighted polygon interpolation techniques and incorporated into ARPS. This became a labor-intensive process because substantial data gaps had to be filled in with 200-meter resolution USGS land use data and 100-meter resolution Defense Mapping
Agency (DMA) terrain elevation data.
Translation tables were derived to convert the detailed, multi-level USGS and SFWMD land use categories (Figure 7)into the less precise ARPS vegetation types in a consistent manner (Figure 8). After several iterations back-and-forth between state and US government agencies, an error-free, high-resolution soil database compatible with the ARPS model grid projection was delivered by SFWMD. Still, this dataset had to be supplemented with data valu es from 1993 LANDSAT imagery to fill in data-void regions. The soil categories were then thinned to match the less descriptive soil types used in ARPS (Figure 9). These improvements in horizontal resolution, the specificatio n of detailed surface characteristics, and the inclusion of feedback mechanisms between the hydrologic and atmospheric systems should improve the simulation of local weather regimes driven by micro-scale features in the surface properties.
The first step is to rerun the climatologically representative sea breeze simulation at the higher resolution, incorporating the SFWMD's hydrologic model soil, vegetation, land cover and terrain elevation datasets. The results will be compared/calibrated against composites of the real data measurements to assess model performance. This realistic "present day" atmospheric response will then be contrasted against a simulation initialized with surface property databases that portray the more homogeneous pre- colonized "natural system". Through direct comparisons of the simulations and by selectively reverting isolated areas of urbanization and drainage to their natural state, any distinctive microclimates (urban heat islands, associated shifts in the horizont al rainfall distribution, and modulations in the amount of rainfall) which have emerged over the past century can be identified.
OUTLOOK FOR REMAINING WORK
These improvements in horizontal resolution, the specification of surface characteristics, and the inclusion of feedback mechanisms between the hydrologic and atmospheric systems should help to resolve details in the shape of thunderstorm convective cells , rectify previous underpredictions of rainfall over the Florida peninsula, and provide more reliable estimates of total freshwater input from the atmosphere into the ground surface/bay/ocean below. It will then be possible to quantify the effects that he avy or persistent rain episodes have on the salinity/nutrient composition of Florida Bay, determine the extent of sewage system overflows, and assess the degree of eutrophication by fertilizer/pesticide/contaminant runoff from agricultural and industrial areas. This work also lays a foundation for the development of more closely coupled versions of a hydro-meteorological model in the near future.
Bougeault, P., B. Bret, P. Lacarrere and J. Noilhan, 1991: An experiment with an advanced surface parameterization in a mesobeta-scale model. Part II: The June 1986 simulation. Monthly Weather Review, 119, 2374-2392.
Burpee, R.W. and L.N. Lahiff, 1984: Area average rainfall variations on sea breeze days in south Florida, Monthly Weather Review, 112, 520-534.
Dell 'Osso, L., 1984: High-resolution experiments with the ECMWF model: A case study. Monthly Weather Review, 112, 1853-1883.
Gannon, P.T., 1978: Influence of earth surface and cloud properties on the south Florida sea breeze, NOAA Technical Report ERL 402-NHEML, 2, 91 pp.
Lynn, B.H., W-K Tao and P.J. Wetzel, 1998: A study of landscape-generated deep moist convection. Monthly Weather Review, 126, 928-942.
McCumber, M.C. and R.A. Pielke, 1981: Simulation of the effects of surface fluxes of heat and moisture in a mesoscale numerical model. 1. Soil layer. Journal of Geophysical Research, 86, 9929-9938.
Mesinger, F., 1984: A blocking technique for representation of mountains in atmospheric models. Riv. Meteor. Aeronautica, 44, 195-202.
Pielke, R.A., 1974: A three-dimensional numerical model of the sea breezes over south Florida, Monthly Weather Review, 102, 115-139.
Pielke, R.A., 1974: A comparison of three-dimensional and two-dimensional numerical predictions of sea breeze, Journal of the Atmospheric Sciences, 31, 1577-1585.
Pielke, R.A., 1984: Mesoscale Meteorological Modeling. New York, Academic Press, 612 pp.
Pielke, R.A., R.L. Walko, L. Steyaert, P.L. Vidale, G.E. Liston and W.A. Lyons, 1998: The influence of anthropogenic landscape changes on weather in South Florida, Monthly Weather Review, submitted.
Ralph, F.M., L. Armi, J.M. Bane, C. Dorman, W.D. Neff, P.J. Neiman, W.Nuss and P.O.G. Persson, 1998: Observations and analysis of the 10-11 June 1994 coastally trapped disturbance. Monthly Weather Review, 126, 2435-2 465.
Thompson, W.T., T. Haack, J.D. Doyle and S.D. Burk, 1997: A nonhydrostatic mesoscale simulation of the 10-11 June 1994 coastally trapped wind reversal. Monthly Weather Review, 125, 3211-3230.