Work Plan
The approach proposed here is to investigate the hydrological cycle for tropical islands using PR as the control scenario. The approach will consist in integrating all possible positive and negative contributors including synoptic and orthographic cloud formations, soil storage, surface runoff, local anthropogenic and vegetative consumption, evapotranspiration, and groundwater flow into a common system at a regional level. The investigation will include a statistical analysis of the monthly and annual water balance in the island based on reported observations by the USGS, the National Weather Service, and local water distribution companies. A second major task will be to develop a meso-scale atmosphere/land/groundwater model that simulates the hydrological balance of the island. This model effort will be validated with results from the analysis and will be used to predict future events in which local and large-scale events will be present. The effort of investigating in detail the hydrological cycle in tropical islands could be used for other islands different from PR. An interdisciplinary team that includes climatologists, hydrologists, remote sensing, and computational fluid dynamics experts has been assembled to address this challenging but yet interesting problem.
Specific models that will be coupled include:
o RAMS – Atmospheric processes
o LEAF-2/TOPMODEL – Near surface processes (i.e., soil moisture, runoff, evapotranspiration, subsurface hillslope moisture transport )
o MODFLOW – Groundwater flow
Figure 5 shows how the various models will work together.
Figure 5. Flowchart of numerical models used to simulate Puerto Rico water balance.
RAMS will be the simulation tool for the atmospheric component. It is a highly versatile numerical code developed for simulating and forecasting meteorological phenomena. It consists of three major components, 1) a data analysis component, 2) an atmospheric model, and 3) a post-processing component. The data analysis component prepares the data for model initialization and nudging from observed meteorological data. The atmospheric model is built around the full set of non-hydrostatic, dynamical equations that governs atmospheric dynamics and thermodynamics, plus conservation equations for scalar quantities like mass, water vapor, liquid and ice hydrometeor mixing ratios. These equations are complemented by a large selection of parameterizations available in the model. Pielke et al. (1992) describes the data analysis technique available in RAMS. The data analysis for the initial and boundary conditions is as follows. An isentropic analysis interpolates the pressure data in the vertical direction to specified isentropic levels, and horizontally interpolates this data to the higher resolution grid to be used in the simulation. Then the vertical isentropic data set is interpolated to the model, to obtain a full set of prognostic fields for model integration. The transient data introduced in the model is the pressure level data provided by the National Center of Environment Prediction (NCEP) at 2.5 degree of resolution. In this study the model will include three nodes of 20, 5 and 1 km to avoid possibly instabilities.
We plan to make full use of remotely sensed (RS) data, which will be obtained from NOAA and NASA via the Internet, and from the Direct Broadcast (DB) station at UPRM. RS products will be used for both model calibration and validation in situations where there are either sparse or no appropriate in-situ data. The local DB station is called the Space Information Laboratory (SIL) and is a component of the NASA-funded Tropical Center for Earth and Space Studies (TCESS). One of SIL’s main functions is to support projects at UPRM that require RS data over the Caribbean region. The SIL collects AVHRR, SeaWiFS, MODIS, Radarsat, and Landsat 7 data. SIL is able to process AVHRR data to level 3, and SeaWiFS and MODIS to level 2. At present SIL cannot process Landsat and Radarsat data beyond level 0. This situation could change by 2003 if licensing negotiations are successful.
The major tasks associated with configuring LEAF-2 and TOPMODEL model include:
o Delineation of Land Cover. This work will utilize a variety of sources of data including remotely-sensed data for PR. Remotely-sensed data may be obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS), the Landsat Multispectral Scanner (MSS), the Landsat Thematic Mapper (TM), and/or the Calibrated Airborne Multispectral Scanner (CAMS). Although air photos are generally too detailed for the scale of this modeling project, they will be used in some cases where remote sensing techniques are inadequate.
o Delineation of Soil Type. A soil GIS has been previously developed for PR and will be used in this project. The GIS database will provide soil hydraulic properties (e.g., layer texture, permeability, bulk densities, water holding capacity, layer thicknesses, etc.)
o Delineation of Watershed Boundaries and Stream Network. Watershed boundaries and the stream network will be delineated using USGS Digital Line Graph (DLG) files.
o Topography. Development of slope directions will be obtained from DEM (Digital Elevation Model). The DEM for PR will be used for elevation data needed in the TOPMODEL subsurface flow model. Slope directions will be determined within ARC/INFO/ArcView.
Data from hydrologic models previously developed in PR (e.g., Cruise and Miller, 1993 and 1994; Miller and Cruise, 1995; Mashriqui and Cruise, 1997; Boyington, 1998; PRWRERI, 2002; Vélez-Rodrigüez, 2002; Pérez-Alegría, 2002) will be used to assist in configuring LEAF-2 and TOPMODEL.
Modifications will be made to the version of TOPMODEL currently being used in the RAMS/LEAF-2/TOPMODEL modeling system. In those areas of the island where deep vertical seepage occurs (e.g., coastal alluvial and karstic limestone aquifers), aquifer recharge will be estimated and passed to the MODFLOW groundwater flow model. MODFLOW will only simulate groundwater flow in selected areas of the island. In those areas of the island where significant aquifer systems are not present (e.g., the interior mountain area/volcanic rock), TOPMODEL will be used as is to simulate lateral downslope transport of water within saturated regions of the soil. Modifications to the computer codes will be made using the Lahey FORTRAN 95 and/or Digital Visual Fortran compilers, depending upon which compiler was used for developing the original code.
The groundwater flow model will be configured using data from numerous USGS groundwater resources studies conducted in PR (e.g., Puig and Rodríguez-Martínez, 1993; Rodríguez-Martínez, 1996; Pérez-Blair and Carrasquillo-Nieves, 1994; Pérez-Blair, 1997; Graves, 1991; Rodríguez-Martínez and Richards, 2000; Rodríguez-Martínez, 2001; Ramos-Ginés, 1994). Data will also be obtained from previously developed groundwater flow models in PR (e.g., Torres-González, 1985; Quiñones-Aponte, 1986; Graves, 1989; Tucci and Martínez, 1995; Quiñones-Aponte et al., 1996; Sepúlveda, 1999; and Kipp, 1987). The locations of these studies are summaries in Figure 6. The GIS-based user interface GMS (Groundwater Modeling System; Brigham Young University, 1997) will be used to manipulate input and output databases for the groundwater flow model. GMS was developed under the direction of the U.S. Army Corps of Engineers and involved support from the Department of Defense, the Department of Energy, and the Environmental Protection Agency. Tools are provided for site characterization, model conceptualization, finite-difference grid generation, geostatistics, telescopic model refinement, and output post-processing.
Figure 6. Locations of Groundwater Resource and Modeling Studies in Puerto Rico
Calibrate the models for regional/local conditions.
RAMS/ LEAF-2/TOPMODEL
RAMS/LEAF-2/TOPMODEL will be calibrated by adjusting model parameters until simulated calibration variables correspond reasonably close to actual measured data. Calibration variables will include rainfall, near surface temperature and relative humidity, reference evapotranspiration, stream base flow and storm discharge. The calibration will be conducted during a one-month period during the dry season (e.g., February) and a one-month during period during the wet season (e.g., November). Surface water data is collected at forty stream flow stations throughout the island. Weather data (rainfall and air temperature) is collected at over seventy stations. Daily information will be used to perform a transient calibration. We will perform the RAMS/LEAF-2/TOPMODEL calibration with the assistance of a commercially available nonlinear optimization program such as PEST (Doherty, 1994). PEST is able to "take control" of a complex, multi-dimensional, transient model, running it as many times as it needs to while adjusting its parameters until the discrepancies between selected model outputs and a complementary set of field measurements is reduced to a minimum in the weighted least squares sense. PEST implements a particularly robust variant of the Gauss-Marquardt-Levenberg method of nonlinear parameter estimation. The program can run within a Windows or UNIX operating environment.
MODFLOW
Where sufficient data exist, steady-state and transient model calibrations will be performed. Many of the areas being modeled will be based on data from previously calibrated groundwater flow models. However, for various reasons, it may be necessary to configure the larger regional-scale model differently from the smaller local-scale models (e.g., because different grid spacing may be used), and these differences may have an effect on the simulated groundwater levels.
The groundwater flow model will initially be calibrated for long-term average steady-state conditions. Calibration will be achieved by adjusting aquifer properties within reasonable limits in order to match observed average groundwater levels and discharge rates. Discharges will include base flow to rivers and discharges to the ocean. These data will be obtained from published reports. We will perform the MODFLOW calibration with the assistance of a commercially available nonlinear optimization program such as PEST (Doherty, 1994). In addition, a one-year transient model calibration will be performed in aquifers where synoptic groundwater level and discharge data exist.
Model Validation
We propose validating the model in two ways.
- Compare model estimates with ground-based historical data; and
- Compare the model-estimated island-wide water balance with a water balance obtained from ground-based and remotely-sensed data.
Validation Step 1
The data used for Validation Step 1 will be of the same form as was used in the model calibrations (i.e., from data collection stations), except that the data will be selected from different years. For example, if the transient calibration data for the groundwater flow model were from 1994, the validation data set would be from some other year, preferably a year with significantly different conditions (e.g., more wet or more dry).
Validation Step 2
The monthly water balance for the island, over a period of one year, will be calculated using the following simple equation:
DP = P + ET + RO + BF - DS (1)
where DP is deep percolation or aquifer recharge, P is precipitation, ET is evapotranspiration, RO is surface runoff, BF is river base flow, and DS is change in moisture storage. Each component of equation 1 is a function of space and time. On average DS is negligible for long periods (e.g., one year), however, it will be important for shorter periods (e.g., one month).
The components on the right-hand-side of equation 1 will be estimated using ground-based and remotely-sensed data. Soil moisture will be estimated using the coupled hydrologic/radiobrightness model (Laymon et al., 2002) with data from the Advanced Microwave Scanning Radiometer-EOS (AMSR-E). Daily soil moisture content will also be estimated using a simplified water budget approach. The GIS-based water budget procedure is as follows:
1. Infiltration will be estimated by subtracting surface runoff from rainfall. Runoff will be estimated using the curve number (CN) approach. Soils data (e.g., CN and soil moisture holding capacity) currently exist in GIS form for PR.
2. If water initially within the soil profile plus the infiltrating water does not exceed the soil water holding capacity, then soil moisture content is equal to the initial volume plus infiltration.
3. If water initially within the soil profile plus the infiltrating water exceeds the soil water holding capacity, then the excess water will be considered percolation and the soil moisture content will be adjusted to the value of the soil moisture holding capacity.
Remotely-sensed evapotranspiration will be obtained from the Aqua/MODIS system. Evapotranspiration will also be determined (within a GIS) using simplified procedures for estimating average monthly climate data for PR described by Harmsen et al., 2002.
Microwave, Surface and Precipitation Products (MSPPS) suite of products, which includes rain rate, land surface temperature, and land-surface emissivity will be obtained from NOAA/NESDIS. These are proven hydrological data products produced from the NOAA polar orbiters and are updated globally every four hours. These products have coarse spatial resolution (16-48 km) and will be downscaled. The most useful NASA products will be those generated from MODIS data. MODIS data can now be obtained from both the Terra (AM) and Aqua (PM) satellites. They have good temporal (1-2 days) and spatial (1-km) resolution. Products include land-surface temperature, land-cover type, vegetation indices, and leaf area index.
Conduct simulation studies of the hydrological cycle on Puerto Rico.
A series of short-term simulations on the order of days and months will be conducted to determine the sources and sinks of the precipitation across the island and parameters that could influence the hydrological balance. The hydrological sources and sinks will be stratified into evapotranspiration, runoff, soil storage, aquifer recharge, precipitation from convective clouds, frontal systems, and easterly waves, etc. These simulations will be configured for the following scenarios:
o Doubling CO2. Doubling the atmospheric carbon dioxide concentration represents a realistic condition that may exist in the future if concentrations continue to increase at present rates. This scenario was considered, for example, by Bouraoui et al. (1999) in a study that evaluated the impact of climate change on water storage and groundwater recharge at the watershed scale. Doubling the CO2 is expected to increase the mean air temperatures by 1 to 5 oC.
o Land-Use. PR is currently undergoing dramatic changes in land-use patterns, which can be expected to affect the components of the hydrologic cycle. We will run simulations that consider probable changes that are expected in land-use in PR in next 25 to 50 years. Land-use changes will include conversion of agricultural land to residential and urban land, as well as increase industrialization.
o Hurricanes. Capiel and Calvesbert (1976) have suggested (qualitatively) that hurricanes are not important to PR’s water balance. We would like to investigate this assertion on a quantitative basis.
o ENSO Index. Someone else needs to write this!!
Questions that we will attempt to answer, relative to the above scenarios include:
o How will river flows be affected?
o How might reservoir levels drop or rise due to changing surface water evaporation rates?
o How might groundwater levels drop or rise due to changing aquifer recharge rates?
o If groundwater levels drop, owing to a reduction in aquifer recharge rates, how might saltwater intrusion increase in the coastal areas?
o What will be the water requirements by agriculture, and how might competition between water users (agriculture, urban and industrial) increase?