«Mapping the Climate of Puerto Rico, Vieques and Culebra CHRISTOPHER DALY, 1* E. H. HELMER, 2 AND MAYA QUIÑONES 2 1 Spatial Climate Analysis Service, ...»
INTERNATIONAL JOURNAL OF CLIMATOLOGY
Int. J. Climatol. 23: 1359-1381 (2003)
Published online 6 August 2003 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/joc.937
Mapping the Climate of Puerto Rico, Vieques and Culebra
CHRISTOPHER DALY, 1* E. H. HELMER, 2 AND MAYA QUIÑONES 2
Spatial Climate Analysis Service, Department of Geosciences, 316 Strand Agricultural Hall, Oregon State University, Corvallis, OR
International Institute of Tropical Forestry, USDA Forest Service, P.O. Box 25000, Río Piedras, 00928-5000, Puerto Rico Received 10 August 2002 Revised 19 May 2003 Accepted 22 May 2003
Both PRISM and HYPS performed well for July maximum temperature, but HYPS performed relatively poorly for January minimum temperature, due primarily to lack of a spatially varying temperature/elevation slope, vertical atmospheric layer definition, and coastal proximity guidance.
Mean monthly precipitation varied significantly throughout the year, reflecting seasonally differing moisture trajectories. Spatial precipitation patterns were associated most strongly with elevation, upslope exposure to predominant moisture-bearing winds, and proximity to the ocean.
IDW performed poorly compared to PRISM, due largely to the lack of elevation and moisture availability information. Overall, the full PRISM approach resulted in greatly improved performance over simpler methods for precipitation and January minimum temperature, but only a small improvement for July maximum temperature. Comparisons of PRISM mean annual temperature and precipitation maps to previously-published, hand-drawn maps showed similar overall patterns and magnitudes, but the PRISM maps provided much more spatial detail.
Key Words: Puerto Rico; Culebra; Vieques; PRISM; climate mapping; temperature; precipitation; interpolation Correspondence to: Christopher Daly, Spatial Climate Analysis Service, Department of Geosciences, 326 Strand Agriculture Hall, Oregon State University, Corvallis OR 97331-2204, USA; email: firstname.lastname@example.org
Many spatially explicit ecological and hydrological models require spatially continuous climate data. As scientists and managers using these models work at finer spatial and temporal resolution, higher resolution climate data become more relevant. In the tropics and elsewhere, spatially continuous climate data help predict woody vegetation (Ohmann and Spies, 1998, Foster et al., 1998, Bongers, 1999, Ohmann and Gregory, 2002) and vertebrate species distributions (Woinarski et al., 1999, Joseph and Stockwell, 2000, Stockwell and Peters, 1999). Seasonal variability and climate extremes can be more important to distributions of woody plant species than annual means (Ohmann and Spies, 1998), which illustrates the importance of climate data with temporal resolution finer than annual. Many spatially explicit terrestrial ecosystem models, like those that predict vegetation primary production and related parameters, also rely on climate data, including at daily resolution (e.g. Running and Gower, 1991, Running and Hunt, 1993). These models can also use monthly climate data (Tian et al., 2000, Wang et al., 2001), and include examples of predicting nitrogen trace gas emissions from soils (Liu et al., 2000).
Climate data have long been important to mapping potential vegetation or ecological zones (Holdridge et al., 1971, Ewel and Whitmore, 1973), and such mapping has become more sophisticated or increased in spatial resolution recently (Neilson, 1995, Host et al., 1996, Lugo et al., 2000, Isaac and Bourque, 2001). Ecological zone maps or spatially continuous climate data can also be important for mapping vegetation type with satellite imagery (e.g. He et al., 1998). Such data are crucial to satellite image based mapping of land cover, vegetation type or forest successional stage in complex tropical areas (Helmer et al., 2000, Helmer et al., 2002). Ecological zones and illumination angles change rapidly in such regions, which leads to spectral confusion in which varied vegetation communities have similar spectral signatures. In addition, the ranges of forest successional stages produced by human disturbance compound spectral confusion between vegetation types. Puerto Rico, which is the focus of this study, is an especially complex Caribbean island. In the Caribbean, ecological zones change rapidly over small areas due to complex topography, climate and soils (Beard, 1949). Conservation of the many endemic species found in these complex tropical islands requires maps with finer spatial and class resolution than existing global- or sub continental-scale ecological zone or satellite image based maps provide (Helmer et al., 2002).
Furthermore, efficient water resource management is critical to Caribbean islands.
Periodic droughts in Caribbean islands cause economic losses and hardships because high population densities combine with limited water resources and storage capacity (Granger, 1983, Larsen, 2001). Spatially continuous data on climate and watershed characteristics improve accuracy of water discharge predictions in Puerto Rico and elsewhere (GarcíaMartinó et al., 1996a), which should contribute to water resources management.
Developing spatially continuous rainfall data, for example, contributed to estimating low stream flow in Northeastern Puerto Rico (García-Martinó et al., 1996a, b). Rainfall is also related to erosion, so spatially explicit models that simulate erosion under various land use configurations (López et al., 1998) can point towards landscape management that minimizes sedimentation flow to reservoirs. Finally, biogeographical and other modeling that relies on climate data can predict potential patterns of change in vegetation Copyright © 2003 Royal Meteorological Society Int. J. Climatol. 23: 1359-1381 (2003) and runoff induced by climate change (Nelson and Marks, 1994, Iverson and Prasad, 1998). These modeling efforts are especially important for Caribbean islands, where vegetation and water resources may be particularly vulnerable to global climate change.
Scatena (1998) predicted that increases in coastal plain temperature of 1 to 2° C, or decreases in rainfall of 11 to 33%, would adversely impact water supplies, as well as impact the distribution of forest types in the Luquillo Mountains of Puerto Rico, potentially affecting endemic species.
To support mapping and spatially explicit analysis of vegetation composition, structure and other attributes, species distributions, ecosystem modeling, and water resources management in Puerto Rico, we had the objective of mapping annual and monthly rainfall and temperature at a relatively fine spatial resolution. The Parameter-elevation Regressions on Independent Slopes Model (PRISM) (Daly and Neilson, 1992, Daly et al., 1994, 1997, in press) is especially suited to mapping climate in a complex landscape like that of Puerto Rico. This study describes application of PRISM to the particular landscapes of Puerto Rico, Vieques and Culebra, under alternative parameterizations that simulate simpler interpolation methods as well as the full PRISM model. The effectiveness of each method is assessed, and compared and contrasted with other methods. All simulations are made at 15 arc-seconds (~450 m) as a spatial resolution that would capture the island's complexity and improve on current data for satellite image based mapping.
The regression-based PRISM uses point data, a digital elevation model (DEM), other spatial data sets, a knowledge base, and expert interaction to generate repeatable estimates of annual, monthly, daily, and event-based climatic elements. These estimates are interpolated to a regular grid, making them GIS-compatible. Recent mapping efforts include peer-reviewed, official USDA precipitation and temperature maps for all 50 states and Pacific Islands (Bishop et al., 1998, USDA-NRCS, 1998, Daly and Johnson, 1999, Vogel et al., 1999, Daly et al., 2001); a new official climate atlas for the United States (Plantico et al., 2000); a 103-year series of monthly temperature and precipitation maps for the conterminous 48 states (Daly et al., 2000b); precipitation and temperature maps for Canada, China and Mongolia (Daly et al., 2000a), and the first comprehensive precipitation maps for the European Alps region (Schwarb et al., 2001a, 2001b).
2.1. Study Area The main island of Puerto Rico lies between about 17°45’ N and 18°30’ N, and its longitude ranges from about 65°45’ W to 67°15’ W (Figure 1). With the main island occupying about 8740 km2, it is the smallest and easternmost of the Greater Antilles.
Vieques and Culebra Islands lie a short distance to the east of the main island, with areas of 125 km2 and 25 km2, respectively. Vieques and Culebra are the westernmost of the Lesser Antilles, which extend in a southeasterly arc from Puerto Rico to the northern coast of South America. Vegetation zones on the main island range from dry, semi deciduous forests in patches and bands on the north and east coasts and in the southwest part of the island, to moist forests that cover the major proportion of the island, to wet and rain forests, including cloud forests, at higher elevations. Elevations range from sea level to 1338 m at Cerro de Punta, in the Cordillera Central on the main island (Figure 1).
Copyright © 2003 Royal Meteorological Society Int. J. Climatol. 23: 1359-1381 (2003) Puerto Rico, Vieques, and Culebra have a climate that is tropical and predominantly maritime, typical of the Caribbean islands. Temperatures exhibit small seasonal variation, due to close proximity to the equator. Temperatures at sea level are quite warm, but decrease markedly with increasing elevation. Humidity is generally high, due to the presence of warm ocean waters. Rainfall and cloudiness are influenced strongly by topography, with mountainous regions being much cloudier and wetter than adjacent lowlands and ocean areas. During May to November, precipitation is produced mainly by easterly waves, which are disturbances embedded in the generally east-to-west trade winds across the region. The strength of these waves ranges greatly, from lowintensity systems accompanied by little or no rainfall, to intense systems that produce flooding rains. The greatest rainfall occurs when tropical storms or hurricanes occasionally develop on these waves and move across or near Puerto Rico. During November to April, cold fronts moving off the eastern US seaboard, that penetrate far enough south to affect the islands, can produce significant precipitation.
2.2. Climate Data Monthly average minimum (Tmin) and maximum (Tmax) temperatures for the years 1963 -1995 were calculated for 47 National Weather Service cooperative stations using daily data from the National Climate Data Center (NCDC 1995) (Figure 1). Monthly and annual average Tmin and Tmax over the period were derived from averaging monthly and yearly data over all years, ignoring missing data. Three estimated minimum temperature sites were added in the vicinity of Adjuntas to aid in the spatial definition of a local nighttime temperature minimum in this area. Estimates were made based on an analysis of likely topographic constraints to cold air drainage. Mean monthly and annual precipitation totals were calculated for the period 1963-1995 using monthly data for 108 stations from the NCDC. Additional precipitation data were obtained from a station at El Verde, in the Luquillo Mountains. This station, operated by the University of Puerto Rico, had 19 years of record (Garcia-Martino et al., 1996a).
Figure 1: Terrain map of Puerto Rico, Vieques, and Culebra, showing climate station locations Copyright © 2003 Royal Meteorological Society Int. J. Climatol. 23: 1359-1381 (2003)
2.3. PRISM Model Formulation for Climate Mapping PRISM adopts the assumption that for a localized region, elevation is the most important factor in the distribution of temperature and precipitation (Daly et al., 2002).
PRISM calculates a linear climate-elevation relationship for each DEM grid cell, but the slope of this line changes locally with elevation as dictated by the data points. Beyond the lowest or highest station, the function can be extrapolated linearly as far as needed. A simple, rather than multiple, regression model was chosen because controlling and interpreting the complex relationships between multiple independent variables and climate is difficult. Instead, weighting the data points (discussed later) controls the effects of variables other than elevation.
The climate-elevation regression is developed from x,y pairs of elevation and climate observations supplied by station data. A moving-window procedure is used to calculate a unique climate-elevation regression function for each grid cell. The simple linear regression has the form
Y =β1 X + β0(1)
where Y is the predicted climate element, β1 and β0 are the regression slope and intercept, respectively, and X is the DEM elevation at the target grid cell. The DEM elevation is expressed at a scale appropriate for the climate element being mapped (see Section 2.3.1).