• Chen, A. A., , and M. A. Taylor, 2002: Investigating the link between early season Caribbean rainfall and the El Niño+1 year. Int. J. Climatol., 22, 87106.

    • Search Google Scholar
    • Export Citation
  • Comarazamy, D. E., , and J. E. González, 2008: On the validation of the simulation of early season precipitation on the island of Puerto Rico using a mesoscale atmospheric model. J. Hydrometeor., 9, 507520.

    • Search Google Scholar
    • Export Citation
  • Comarazamy, D. E., , and J. E. González, 2011: Regional long-term climate change (1950–2000) in the midtropical Atlantic and its impacts on the hydrological cycle of Puerto Rico. J. Geophys. Res., 116, D00Q05, doi:10.1029/2010JD015414.

    • Search Google Scholar
    • Export Citation
  • Comarazamy, D. E., , J. E. González, , J. C. Luvall, , D. L. Rickman, , and P. J. Mulero, 2010: A land–atmospheric interaction study in the coastal tropical city of San Juan, Puerto Rico. Earth Interact., 14. [Available online at http://EarthInteractions.org.]

    • Search Google Scholar
    • Export Citation
  • Cotton, W. R., and Coauthors, 2003: RAMS 2001: Current status and future directions. Meteor. Atmos. Phys., 82, 529.

  • Dai, A., , K. E. Trenberth, , and T. R. Karl, 1999: Effects of clouds, soil moisture, precipitation, and water vapor on diurnal temperature range. J. Climate, 12, 24512473.

    • Search Google Scholar
    • Export Citation
  • Daly, C., , E. H. Helmer, , and M. Quiñones, 2003: Mapping the climate of Puerto Rico, Vieques and Culebra. Int. J. Climatol., 23, 13591381.

    • Search Google Scholar
    • Export Citation
  • de Almeida, C. M., , A. M. V. Monteiro, , G. Câmara, , B. S. Soares-Filho, , G. C. Cerqueira, , C. L. Pennachin, , and M. Batty, 2005: GIS and remote sensing as tools for the simulation of urban land-use change. Int. J. Remote Sens., 26, 759774.

    • Search Google Scholar
    • Export Citation
  • Dickinson, R. E., , A. H. Sellers, , P. J. Kennedy, , and M. F. Wilson, 1986: Biosphere-atmosphere transfer scheme (BATS) for the NCAR Climate Community Model. NCAR Tech. Note NCAR/TN-275+STR, 69 pp.

  • Duchon, C. E., 1986: Temperature trends at San Juan, Puerto Rico. Bull. Amer. Meteor. Soc., 67, 13701377.

  • Elmore, K. L., , M. E. Baldwin, , and D. M. Schultz, 2006: Field significance revisited: Spatial bias errors in forecasts as applied to the Eta model. Mon. Wea. Rev., 134, 519531.

    • Search Google Scholar
    • Export Citation
  • Environmental and Natural Resources Ministry of the Dominican Republic, 2007: Políticas para la gestión efectiva del Sistema Nacional de Areas Protegidas de la República Dominicana. Environmental and Natural Resources Ministry of the Dominican Republic Rep., 59 pp.

  • Feddema, J., , K. Oleson, , G. Bonan, , L. Mearns, , W. Washington, , G. Meehl, , and D. Nychka, 2005: A comparison of a GCM response to historical anthropogenic land cover change and model sensitivity to uncertainty in present-day land cover representations. Climate Dyn., 25, 581609, doi:10.1007/s00382-005-0038-z.

    • Search Google Scholar
    • Export Citation
  • González, J. E., , J. C. Luvall, , D. Rickman, , D. E. Comarazamy, , and A. J. Picón, 2006: Urban heat island identification and climatologic analysis in a coastal, tropical city: San Juan, Puerto Rico. Urban Remote Sensing, Q. Weng and D. Quattrochi, Eds., CRC Press, 223–252.

  • Helmer, E. H., 2004: Forest conservation and land development in Puerto Rico. Landscape Ecol., 19, 2940.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471.

  • Kennaway, T., , and E. H. Helmer, 2007: The forest types and ages cleared for land development in Puerto Rico. GISci. Remote Sens., 44, 356382.

    • Search Google Scholar
    • Export Citation
  • Landsberg, H. E., 1981: The Urban Climate. Academic Press, 275 pp.

  • Land Use Planning Office of Puerto Rico, 2006: Plan de uso de terrenos de Puerto Rico. Land Use Planning Office of Puerto Rico Rep., 89 pp.

  • Lawton, R. O., , U. S. Nair, , R. A. Pielke Sr., , and R. M. Welch, 2001: Climatic impact of tropical lowland deforestation on nearby montane cloud forests. Science, 294, 584587.

    • Search Google Scholar
    • Export Citation
  • Lemonsu, A., , V. Masson, , and E. Berthier, 2007: Improvement of the hydrological component of an urban soil–vegetation–atmosphere–transfer model. Hydrol. Processes, 21, 21002111.

    • Search Google Scholar
    • Export Citation
  • Livezey, R. E., , and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111, 4659.

    • Search Google Scholar
    • Export Citation
  • López, E., , G. Bocco, , M. Mendoza, , and E. Dahau, 2001: Predicting land-cover and land-use change in the urban fringe: A case in Morelia city, Mexico. Landscape Urban Plann., 55, 271285.

    • Search Google Scholar
    • Export Citation
  • Lugo, A. E., , and E. Helmer, 2004: Emerging forests on abandoned land: Puerto Rico’s new forests. For. Ecol. Manage., 190, 145161.

  • Magaña, V., , J. A. Amador, , and S. Medina, 1999: The mid-summer drought over Mexico and Central America. J. Climate, 12, 15771588.

  • Malmgren, B. A., , A. Winter, , and D. Chen, 1998: El Niño–Southern Oscillation and North Atlantic Oscillation control of climate in Puerto Rico. J. Climate, 11, 27132717.

    • Search Google Scholar
    • Export Citation
  • Mansell, M., , and F. Rollet, 2006: Water balance and the behaviour of different paving surfaces. Water Environ., 20, 7–10.

  • Martilli, A., , A. Clappier, , and M. W. Rotach, 2002: An urban surface exchange parameterisation for mesoscale models. Bound.-Layer Meteor., 104, 261304.

    • Search Google Scholar
    • Export Citation
  • Martinuzzi, S., , W. A. Gould, , and O. M. Ramos González, 2007: Land development, land use, and urban sprawl in Puerto Rico integrating remote sensing and population census data. Landscape Urban Plann., 79, 288297.

    • Search Google Scholar
    • Export Citation
  • Masson, V., 2000: A physically-based scheme for the urban energy budget in atmospheric models. Bound.-Layer Meteor., 94, 357397.

  • Maurer, E. P., , and D. P. Lettenmaier, 2003: Predictability of seasonal runoff in the Mississippi River basin. J. Geophys. Res., 108, 8607, doi:10.1029/2002JD002555.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–846.

  • Myers, N., , R. A. Mittermeier, , C. G. Mittermeier, , G. A. B. da Fonseca, , and J. Kent, 2000: Biodiversity hotspots for conservation priorities. Nature, 403, 853858.

    • Search Google Scholar
    • Export Citation
  • Nair, U. S., , R. O. Lawton, , R. M. Welch, , and R. A. Pielke Sr., 2003: Impact of land use on Costa Rican tropical montane cloud forests: Sensitivity of cumulus cloud field characteristics to lowland deforestation. J. Geophys. Res., 108, 4206, doi:10.1029/2001JD001135.

    • Search Google Scholar
    • Export Citation
  • Oke, T. R., 1987: Boundary Layer Climates. Routledge, 435 pp.

  • Pielke, R. A., and Coauthors, 1992: A comprehensive meteorological modeling system—RAMS. Meteor. Atmos. Phys., 49, 6991.

  • Ragab, R., , P. Rosier, , A. Dixon, , J. Bromley, , and J. D. Cooper, 2003: Experimental study of water fluxes in a residential area: 2. Road infiltration, runoff and evaporation. Hydrol. Processes, 17, 24232437.

    • Search Google Scholar
    • Export Citation
  • Ray, D. K., , U. S. Nair, , R. O. Lawton, , R. M. Welch, , and R. A. Pielke Sr., 2006: Impact of land use on Costa Rican tropical montane cloud forests: Sensitivity of orographic cloud formation to deforestation in the plains. J. Geophys. Res., 111, D02108, doi:10.1029/2005JD006096.

    • Search Google Scholar
    • Export Citation
  • Saleeby, M. S., , and W. R. Cotton, 2004: A large-droplet mode and prognostic number concentration of cloud droplets in the Colorado State University Regional Atmospheric Modeling System (RAMS). Part I: Module descriptions and supercell test simulations. J. Appl. Meteor., 43, 182195.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., , and R. W. Reynolds, 2003: Extended reconstruction of global sea surface temperatures based on COADS data (1854–1997). J. Climate, 16, 14951510.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., , R. W. Reynolds, , T. C. Peterson, , and J. Lawrimore, 2008: Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006). J. Climate, 21, 22832296.

    • Search Google Scholar
    • Export Citation
  • Solomon, S., , D. Qin, , M. Manning, , M. Marquis, , K. Averyt, , M. M. B. Tignor, , H. L. Miller Jr., , and Z. Chen, Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.

  • Stein, U., , and P. Alpert, 1993: Factor separation in numerical simulations. J. Atmos. Sci., 50, 21072115.

  • Taylor, M. A., , D. B. Enfield, , and A. A. Chen, 2002: Influence of the tropical Atlantic versus the tropical Pacific on Caribbean rainfall. J. Geophys. Res., 107, 3127, doi:10.1029/2001JC001097.

  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

  • Tso, C. P., 1996: A survey of urban heat island studies in two tropical cities. Atmos. Environ., 30, 507519.

  • Van der Molen, M. K., 2002: Meteorological impacts of land use change in the maritime tropic. Vrije Universititeit Amsterdam Rep., 274 pp.

  • Van der Molen, M. K., , A. J. Dolman, , M. J. Waterloo, , and L. A. Bruijnzeel, 2006: Climate is affected more by maritime than by continental land use change: A multiple scale analysis. Global Planet. Change, 54, 128149.

    • Search Google Scholar
    • Export Citation
  • Velazquez-Lozada, A., , J. E. Gonzalez, , and A. Winter, 2006: Urban heat island effect analysis in San Juan, Puerto Rico. Atmos. Environ., 40, 17311741.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences. 2nd ed. Academic Press, 648 pp.

  • View in gallery

    A 30-yr (1970–2000) monthly precipitation climatology (mm) for 15 COOP stations distributed throughout Puerto Rico (adapted from Comarazamy and González 2011).

  • View in gallery

    Time series and linear regression trend of daily maximum and minimum temperatures (°C) for (top) 1950–2006 averaged over 26 surface COOP stations located throughout Puerto Rico and (bottom) 1956–2006 averaged over those stations located within the San Juan metropolitan area. The slope m has a unit of degrees Celsius per decade. Statistical significance p values are also shown.

  • View in gallery

    (right) Map showing the LCLU specifications in northeastern Puerto Rico for (top) 1951 and (bottom) 2000; 2000 information is complemented with remote sensing data obtained from the ATLAS sensor. The thick solid vertical line represents the location of the north–south vertical cross section in Figs. 8 and 9. (left) (top) Histogram of historical LCLU changes in percent of total area covered from 1951 to 2000 and (bottom) description of the most relevant vegetation and land classes with percent change and conversion rates.

  • View in gallery

    Location of modeling grids used. The largest grid—Grid 1 (G1)—shows the Caribbean Basin and islands; grid 2 (G2) covers the island of Puerto Rico and adjoining Vieques, Culebra, and Virgin Islands to the east of the domain; and grid 3 (G3) covers the northeastern region of Puerto Rico, the main area of interest.

  • View in gallery

    Scatterplots of observed vs simulated daily minimum (blue circles) and maximum (red circles) temperatures (°C) averaged over the corresponding 5-yr period over the SJMA for (a) the present and (b) the past and (c) for the present over El Yunque (LEF).

  • View in gallery

    Differences in minimum temperature (°C) due to (a) LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) LCLU changes while driving the model with past atmospheric conditions, (d) global warming with present LCLU specifications, (e) global warming with past LCLU specifications, (c) total change due to LCLU changes and global warming, and (f) the contribution due to nonlinear interaction among the LCLU and global warming factors. The thick black line delineates urban areas and the thick green line delineates the 350-m-topography contour. The 45° hatch marks identify regions without statistically significant differences (at αLS = 0.05).

  • View in gallery

    As in Fig. 6, but for maximum temperature.

  • View in gallery

    Vertical cross sections, at the time when minimum temperatures occur through the north–south line in Fig. 3 (66.05°W), of differences in temperature (°C; shading), liquid water mixing ratio (g kg−1), and wind vectors due to (a) LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) LCLU changes while driving the model with past atmospheric conditions, (d) global warming with present LCLU specifications, (e) global warming with past LCLU specifications, and (c) total change due to LCLU changes and global warming. The thick blue line delineates sea surface, and the thick green shading indicates underlying topography. The vertical velocity is 10× the reference vector.

  • View in gallery

    As in Fig. 8, but for maximum temperatures.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 84 84 40
PDF Downloads 72 72 35

Climate Impacts of Land-Cover and Land-Use Changes in Tropical Islands under Conditions of Global Climate Change

View More View Less
  • 1 NOAA/CREST Center, City College of New York, New York, New York
  • 2 NOAA/Cooperative Remote Sensing Science and Technology Center (CREST), and Department of Mechanical Engineering, City College of New York, New York, New York
  • 3 Global Hydrology and Climate Center, NASA Marshall Space Flight Center, Huntsville, Alabama
  • 4 Department of Meteorology and Climate, San Jose State University, San Jose, California
© Get Permissions
Full access

Abstract

Land-cover and land-use (LCLU) changes have significant climate impacts in tropical coastal regions with the added complexity of occurring within the context of a warming climate. The individual and combined effects of these two factors in tropical islands are investigated by use of an integrated mesoscale atmospheric modeling approach, taking the northeastern region of Puerto Rico as the test case. To achieve this goal, an ensemble of climate simulations is performed, combining two LCLU and global warming scenarios. Reconstructed agricultural maps and sea surface temperatures form the past (1955–59) scenario, while the present (2000–04) scenario is supported with high-resolution remote sensing LCLU data. Here, the authors show that LCLU changes produced the largest near-surface (2-m AGL) air temperature differences over heavily urbanized regions and that these changes do not penetrate the boundary layer. The influence of the global warming signal induces a positive inland gradient of maximum temperature, possibly because of increased trade winds in the present climatology. These increased winds also generate convergence zones and convection that transport heat and moisture into the boundary layer. In terms of minimum temperatures, the global warming signal induces temperature increases along the coastal plains and inland lowlands.

Corresponding author address: Daniel E. Comarazamy, NOAA/CREST, T-107, Steinman Hall, 140th St. and Convent Ave., New York, NY 10031. E-mail: dcomarazamy@ccny.cuny.edu

Abstract

Land-cover and land-use (LCLU) changes have significant climate impacts in tropical coastal regions with the added complexity of occurring within the context of a warming climate. The individual and combined effects of these two factors in tropical islands are investigated by use of an integrated mesoscale atmospheric modeling approach, taking the northeastern region of Puerto Rico as the test case. To achieve this goal, an ensemble of climate simulations is performed, combining two LCLU and global warming scenarios. Reconstructed agricultural maps and sea surface temperatures form the past (1955–59) scenario, while the present (2000–04) scenario is supported with high-resolution remote sensing LCLU data. Here, the authors show that LCLU changes produced the largest near-surface (2-m AGL) air temperature differences over heavily urbanized regions and that these changes do not penetrate the boundary layer. The influence of the global warming signal induces a positive inland gradient of maximum temperature, possibly because of increased trade winds in the present climatology. These increased winds also generate convergence zones and convection that transport heat and moisture into the boundary layer. In terms of minimum temperatures, the global warming signal induces temperature increases along the coastal plains and inland lowlands.

Corresponding author address: Daniel E. Comarazamy, NOAA/CREST, T-107, Steinman Hall, 140th St. and Convent Ave., New York, NY 10031. E-mail: dcomarazamy@ccny.cuny.edu

1. Introduction

Anthropogenic land-cover and land-use (LCLU) changes have profound climate and environmental impacts. One of the most extreme cases of LCLU change is urbanization, with its clearest indicator as the urban–rural thermal phenomenon known as the urban heat island (UHI). The UHI is defined as a dome of high temperatures observed over urban centers, as compared to the relatively cooler rural surroundings (Landsberg 1981; Oke 1987). One factor that leads to UHI formation is the widespread use of construction materials, such as concrete, asphalt, steel, and glass. UHIs are greater in clear and calm conditions and tend to disperse in cloudy and windy weather by the effects of thermal and mechanical mixing. Because of its proximity to ocean waters and because of the strong influence of trade winds on tropical islands, UHI events in northeastern Puerto Rico tend to be stronger and have greater impacts during the Caribbean dry and early rainfall seasons (Velazquez-Lozada et al. 2006). These two seasons compose the period from December to June (Fig. 1). UHIs in tropical regions have been identified in Kuala Lumpur, Malaysia; the island state of Singapore (Tso 1996); and San Juan, Puerto Rico (Velazquez-Lozada et al. 2006), by comparison of historical temperature differences recorded by urban and rural surface weather stations, followed in some studies by numerical simulations. Broadening the scope of these studies would allow discerning between local and regional impacts due to land-use changes and due to global climate fluctuations (Tso 1996).

Fig. 1.
Fig. 1.

A 30-yr (1970–2000) monthly precipitation climatology (mm) for 15 COOP stations distributed throughout Puerto Rico (adapted from Comarazamy and González 2011).

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Besides urbanization, other cases of LCLU changes include but are not limited to deforestation, reforestation, and use of land for agriculture. It has been reported that lowland deforestation may be a leading cause for increases in cloud-base heights and in thinner clouds in Central American rain forests, resulting in increased regional droughts (Lawton et al. 2001; Nair et al. 2003; Ray et al. 2006). Regional model simulations showed that lowland and premontane deforestation increased air temperatures and sensible heat fluxes and decreased dewpoint temperatures and latent heat fluxes of the air masses that eventually form orographic cloudbanks at higher elevations. However, a similar numerical experiment conducted in Puerto Rico reported an increase in cloud-base height for a forested island (Van der Molen 2002; Van der Molen et al. 2006). Forested runs produced a stronger sea breeze, with a more defined convergence front and consequently with stronger updrafts that transport moisture to higher levels, thus increasing cloud-base heights, compared to pasture simulations. Model simulations performed for northeastern Puerto Rico show that the global warming signal produces higher clouds over mountainous terrain, less total liquid water content on the atmospheric column, and reduced surface accumulated precipitation (Comarazamy and González 2011).

As mentioned, a further consideration is that UHI effects and LCLU change impacts in tropical coastal regions both occur under conditions of global climate change and environmental changes, in the particular case of Puerto Rico in a region highly susceptible to large-scale multidecadal oscillations (Chen and Taylor 2002; Taylor et al. 2002; Malmgren et al. 1998). Since the late 1800s, global average concentrations of atmospheric CO2 and other greenhouse gases have been increasing, mainly a consequence of anthropogenic activities, with these increases being related to increases in air temperatures (0.6° ±0.2°C) and changes in several other climate variables (Solomon et al. 2007; Trenberth et al. 2007). Furthermore, the Caribbean Basin is likely to be among the most seriously impacted regions in the world by global climate change (Myers et al. 2000). Historical temperature trends in Puerto Rico, based on National Oceanic and Atmospheric Administration (NOAA) Cooperative Observer Program (COOP) station data, show a contradictory pattern when comparing long-term daily maximum and minimum temperatures averaged over the entire island to the averages of stations located within the San Juan metropolitan area (SJMA), one of the most densely populated and heavily urbanized cities in the Caribbean. From 1950 to 2006, the island-wide trend shows decreasing maximum temperatures and increasing minimum temperatures, while the SJMA shows a pronounced increase in both daily maximum and minimum temperatures (Fig. 2). This SJMA trend, particularly the increase in maximum temperatures, might be due to an increasing UHI effect (Duchon 1986; Velazquez-Lozada et al. 2006) generated by the rapid growth of the SJMA (~829% growth in 50 yr; Fig. 3). The increase in minimum temperatures is attributed to global warming and to an overall increase in water vapor produced by enhanced ocean evaporation (Dai et al. 1999; Daly et al. 2003).

Fig. 2.
Fig. 2.

Time series and linear regression trend of daily maximum and minimum temperatures (°C) for (top) 1950–2006 averaged over 26 surface COOP stations located throughout Puerto Rico and (bottom) 1956–2006 averaged over those stations located within the San Juan metropolitan area. The slope m has a unit of degrees Celsius per decade. Statistical significance p values are also shown.

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Fig. 3.
Fig. 3.

(right) Map showing the LCLU specifications in northeastern Puerto Rico for (top) 1951 and (bottom) 2000; 2000 information is complemented with remote sensing data obtained from the ATLAS sensor. The thick solid vertical line represents the location of the north–south vertical cross section in Figs. 8 and 9. (left) (top) Histogram of historical LCLU changes in percent of total area covered from 1951 to 2000 and (bottom) description of the most relevant vegetation and land classes with percent change and conversion rates.

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Whereas these past studies focused on the climate effects of one particular factor, in the work presented in this paper we present the impacts of LCLU changes on temperature under global warming conditions. The rationale for this study is that the dome of temperature differences caused by local LCLU changes is located within and affected by a larger dome of temperature differences caused by regional climate changes and global warming. We have developed an atmospheric modeling–based methodology that integrates a mesoscale cloud-resolving atmospheric model with remote sensing information, GIS maps (section 2), and statistical techniques (section 3), which allows quantification of the individual and combined contribution of LCLU changes and global warming to the total climate change observed in tropical coastal regions, taking the northeastern region of Puerto Rico as the test case (sections 4 and 5).

2. Methodology and numerical experiment setup

Given the importance of tropical coastal regions, and the relatively lack of studies that include all contributing factors leading to local climate changes at resolutions fine enough to draw strong conclusions, the following research questions are proposed to study the possible climate impacts of LCLU changes in tropical coastal regions under conditions of global warming:

  1. What has been the effect of LCLU changes on the local and regional Caribbean early rainfall season (ERS; from April to June) climate in tropical coastal regions?
  2. What has been the climatic impact of global warming on tropical coastal regions during the region’s ERS?
  3. Under the conditions of LCLU changes and global warming, what has been their combined effect in tropical coastal regions?
To answer these questions, numerical simulations are configured combining two LCLU scenarios (representing current and pre-urban conditions) with two large-scale atmospheric conditions (representing different periods of global warming and their corresponding levels of greenhouse gases emissions), resulting in the run matrix in Table 1. The time frame for the present and past climatologies were selected to reduce the influence of El Niño–Southern Oscillation (ENSO) and North Atlantic Oscillation (NAO) on the Caribbean ERS climate, as identified by previous studies (e.g., Chen and Taylor 2002; Taylor et al. 2002; Malmgren et al. 1998) and in accordance with periods of major historical LCLU changes (see next subsection). The 5-yr periods from 1955 to 1959 (past) and from 2000 to 2004 (present) are the best available in the long-term record, in terms of ENSO and NAO indices, to perform the simulations (Comarazamy and González 2011). Present and past atmospheric conditions were dynamically downscaled from the National Centers for Environmental Prediction (NCEP) 2.5° reanalysis data (Kalnay et al. 1996), with sea surface temperature (SST) specifications derived from Smith and Reynolds Extended Reconstructed SSTs, version 3b (ERSST v3b), (Smith and Reynolds 2003; Smith et al. 2008).
Table 1.

Simulation matrix.

Table 1.

a. LCLU specifications

Digital maps of LCLU available for 1951 and 2000 (Kennaway and Helmer 2007) were analyzed to verify that historical LCLU changes are in accordance with the two time frames selected; and then were configured for the simulations. The first step in this configuration consists of a reclassification of LCLU classes to match the modeling grids (Fig. 4) and the land classification index system in the atmospheric model (Dickinson et al. 1986), thus obtaining the LCLU specifications used as surface characteristics for simulations driven with both past and present atmospheric and oceanic conditions. The methods to derive the 1951 and 2000 LCLU specifications differ only in the use of high-resolution remote sensing data for the SJMA, Caguas, El Yunque, and surrounding areas obtained with the Airborne Thermal and Land Applications Sensor (ATLAS) instrument (González et al. 2006) to complement the present LCLU dataset.

Fig. 4.
Fig. 4.

Location of modeling grids used. The largest grid—Grid 1 (G1)—shows the Caribbean Basin and islands; grid 2 (G2) covers the island of Puerto Rico and adjoining Vieques, Culebra, and Virgin Islands to the east of the domain; and grid 3 (G3) covers the northeastern region of Puerto Rico, the main area of interest.

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Urban areas within the region of interest were identified using the sensor’s visible spectrum, resulting in a configuration that closely resembles current land development, land use, and urban sprawl mapping in Puerto Rico (Martinuzzi et al. 2007). Optical information corresponding to these urban areas was then obtained from its thermal spectrum to create new urban and built-up land classes, producing an improvement in the representation of the urban areas and to update model input surface characteristics (Comarazamy et al. 2010). Analysis of past and present LCLU information indicates a large increase in surface areas covered by urbanization and a shift from an agriculture-based economy (Helmer 2004; Lugo and Helmer 2004; Kennaway and Helmer 2007; Martinuzzi et al. 2007), with an ~34% conversion rate of the surface covered by urbanization and natural vegetation (i.e., shrub land, forest, and woodland) from that of agricultural lands (Fig. 3).

b. Atmospheric model general description and configuration

An ensemble of numerical atmospheric model simulations was performed to separate the signals discussed before (i.e., LCLU change and global climate change). The chosen model is the Regional Atmospheric Modeling System (RAMS), a highly versatile numerical code developed at Colorado State University to simulate and forecast meteorological phenomena (Pielke et al. 1992; Cotton et al. 2003). The version of RAMS used in this investigation, 4.3, contains an explicit cloud microphysics module with eight hydrometeor types and the prognostic number concentration of cloud droplets through activation of cloud condensation nuclei (Saleeby and Cotton 2004).

The simulations are centered on the northeastern region of Puerto Rico, focusing on the SJMA and the Luquillo Experimental Forest (LEF), locally referred to as El Yunque, and with three nested grids. Grid 1 covers the Caribbean Basin with a horizontal resolution of 25 km and 104 horizontal grid points. Grid 2, which is nested within grid 1, covers the island of Puerto Rico with 3100 horizontal grid points at a resolution of 5 km. Grid 3, with 9184 horizontal grid points and a 1-km resolution, is nested within grid 2 and covers the SJMA, LEF, nondeveloped regions west and south of the city, and ocean areas to the north (Fig. 4). The vertical grid has a grid spacing Δz of 30 m near the surface and then is stretched at a constant ratio of 1.15 until Δz = 1000 m. The depth of the model reaches approximately 26 km. A model-derived variable time step Δt was specified for all grids; initially Δt was calculated as 60 s for grid 3.

All runs are performed during the Caribbean ERS period from April to June of each year of the two time frames selected and for each LCLU and atmospheric condition scenario following the number of ensembles specified by the factor separation technique (see next section). This translates to 20 total 3-month-long simulations, where four 5-yr model-produced climatologies, one for each run in Table 1, were used in all validation and analysis exercises. This procedure of calculating a 5-yr average assures that the analysis is done with datasets where possible year-to-year variations are eliminated. The ERS is the most convenient time to study UHI effects (Velazquez-Lozada et al. 2006) and represents the end of the dry season and the onset of the midsummer drought (Magaña et al. 1999), a critical period in the annual hydrological cycle of the island, during which the atmospheric model has previously performed satisfactorily (Comarazamy and González 2008). A spinup time of one week at the start of each 3-month simulation is specified to allow numerical stabilization of the main atmospheric model, submodels, and parameterizations.

After all input information and parameters were incorporated to the atmospheric model and the simulations were performed, the 5-yr average of daily minimum and maximum temperatures from the PRESENT1 and PAST2 simulations were validated and compared with corresponding observed values from COOP stations located within the SJMA and LEF, following procedures developed for the same model applied to the region of interest (Comarazamy and González 2008; Comarazamy et al. 2010). Results for minimum temperatures follow almost entirely a one-to-one correlation (Fig. 5, blue circles). Results for maximum temperature follow a similar pattern of accuracy (Fig. 5, red circles), especially over the forested area. Over the SJMA, on the other hand, although producing satisfactory results, the model showed a tendency for over prediction of maximum temperatures for both the past and present periods.

Fig. 5.
Fig. 5.

Scatterplots of observed vs simulated daily minimum (blue circles) and maximum (red circles) temperatures (°C) averaged over the corresponding 5-yr period over the SJMA for (a) the present and (b) the past and (c) for the present over El Yunque (LEF).

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

The overprediction of maximum temperatures over heavily urbanized areas could be an indication of the model mishandling the full effects of volumetric soil moisture content in vegetated areas within cities and of the effects of wet building materials (e.g., concrete, asphalt) on atmospheric surface heat fluxes. It has been recently reported and it is starting to be widely accepted that such materials as concrete and asphalt absorb and retain water after precipitation events (Ragab et al. 2003; Mansell and Rollet 2006) and that the effects of wet building materials are not included in the surface exchange schemes of current releases of mesoscale atmospheric models (Lemonsu et al. 2007). Given all the results, it is concluded that the atmospheric mesoscale modeling methodology used in this work may not fully account for the effects of soil moisture in urban areas and wet building materials on surface and air temperatures and that this should be addressed in future research. It was also concluded that, since the maximum-temperature overprediction bias was found in the same proportion for the PRESENT1 and PAST2 simulations, it would not affect the comparison between the various scenarios presented in Table 1 and the analysis in section 4.

In general, after the model results were validated and analyzed and acknowledging slight tendencies for over prediction of maximum temperatures (Fig. 5), the model chosen for the study, complemented with digital LCLU maps and airborne remote sensing information and driven with past and present atmospheric and oceanic conditions, is an adequate tool to study the climate impacts due to LCLU changes in coastal tropical regions within the context of global warming.

3. Analysis methods

The application of statistical methods to the modeling results represents an important component of the research in this document. These methods include a factor separation technique to quantify the climate impacts of the two factors of interest for this work (Stein and Alpert 1993). A two-sample hypothesis test was applied to model results to assess the statistical significance of the difference in means obtained after the factor separation calculation. To further test the statistical significance of the model results, a Monte Carlo method was used to test for field significance (Livezey and Chen 1983). All statistical analysis techniques applied to model results are described in detail in the next subsections.

a. Factor separation

A simple method for calculating the interactions among various factors influencing the results of atmospheric models in sensitivity experiments is used: that is, the factor separation technique of Stein and Alpert (1993). The method includes calculation of the individual contribution of each factor, as well as the combined contribution of the factors in question to the predicted total change in any given atmospheric field. The separation of n factors requires 2n simulations to separate the contribution and possible interactions of LCLU changes and global warming (GW); the four simulations in Table 1 are thus adequate.

Let a be any model predicted field that depends on initial and boundary conditions, as well as on model parameterizations and user specified options. If a change is made in two of these input variables, ψ(b) and φ(c), which are a function of changing coefficients, b and c, which in turn assume values of 1 and 0 depending if changes in the factor are taken into account, then a can be expressed as a constant component {a[ψ(0), φ(0)]} plus a component dependent the changing coefficients {â[ψ(b), φ(c)]}. In our case, ψ(b) is the LCLU factor and φ(c) is the atmospheric condition factor. Related to the changing coefficients values they are expressed as
e1
and the simulations in Table 1 are expressed as
e2
In this formulation, a[ψ(0), φ(0)], the PAST2 scenario, is the control simulation, in which none of the factors ψ and φ are present; it is equal to the constant part of a that is independent of the two factors being analyzed, â[(0), φ(0)]. The rest of the left-hand side terms in (2), a[ψ(1), φ(1)], a[ψ(1), φ(0)], and a[ψ(0), φ(1)], are defined as the PRESENT1, PRESENT2, and PAST1 scenarios, respectively. It is also important to be clear about the meaning of â[ψ(1), φ(1)], â[ψ(1), φ(0)], and â[ψ(0), φ(1)]. In that respective sequence, they represent the fraction of a induced by the combination of ψ and φ (both LCLU and GW factors), by ψ only (LCLU), and by φ only (GW). So to complete the formulation of the factor separation method, we solve for â in (2) to obtain
e3
To summarize, putting (3) in terms of the simulation identification in Table 1, the climate change induced by LCLU changes is given by the operation (PRESENT2 − PAST2), the climate change due to the global warming signal by (PAST1 − PAST2), and the combined contribution of LCLU changes and global warming to the total climate change by either [(PRESENT1 − PRESENT2) − (PAST1 − PAST2)] or [(PRESENT1 − PAST1) − (PRESENT2 − PAST2)]. In practice, the calculation of the nonlinear interaction between the two factors involved in the research can be viewed either as the contribution of global warming to the climate impacts due to LCLU changes or as the contribution of LCLU changes to the global warming signal. In either case, the result is the same regardless of viewpoint taken. To better visualize this, Figs. 69 also contain the calculations for (PRESENT1 − PAST1) and (PRESENT1 − PRESENT2), with appropriate text in the next section.
Fig. 6.
Fig. 6.

Differences in minimum temperature (°C) due to (a) LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) LCLU changes while driving the model with past atmospheric conditions, (d) global warming with present LCLU specifications, (e) global warming with past LCLU specifications, (c) total change due to LCLU changes and global warming, and (f) the contribution due to nonlinear interaction among the LCLU and global warming factors. The thick black line delineates urban areas and the thick green line delineates the 350-m-topography contour. The 45° hatch marks identify regions without statistically significant differences (at αLS = 0.05).

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Fig. 7.
Fig. 7.

As in Fig. 6, but for maximum temperature.

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Fig. 8.
Fig. 8.

Vertical cross sections, at the time when minimum temperatures occur through the north–south line in Fig. 3 (66.05°W), of differences in temperature (°C; shading), liquid water mixing ratio (g kg−1), and wind vectors due to (a) LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) LCLU changes while driving the model with past atmospheric conditions, (d) global warming with present LCLU specifications, (e) global warming with past LCLU specifications, and (c) total change due to LCLU changes and global warming. The thick blue line delineates sea surface, and the thick green shading indicates underlying topography. The vertical velocity is 10× the reference vector.

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

Fig. 9.
Fig. 9.

As in Fig. 8, but for maximum temperatures.

Citation: Journal of Climate 26, 5; 10.1175/JCLI-D-12-00087.1

b. Statistical significance tests

The statistical significance test applied for this research considers two separate significance levels: local significance and field or global significance. Local significance is obtained by testing for statistical significance at individual grid points. A two-sample t test is used to test the null hypothesis H0 that the data in the two vectors at a grid point are independent random samples from normal distributions with equal means and equal (but unknown) variances. The test returns the p value of the test statistic
e4
where a is any model predicted field with mean and variance var(ai) over effective samples at each grid point. The effective size sample is calculated to account for serial correlation of the underlying data and is determined by
e5
where ρ1 is the lag-1 serial autocorrelation coefficient.

The p value is the probability, under the null hypothesis, that the calculated value of the test statistic will occur by chance. Therefore, the null hypothesis is rejected if the p value is less than or equal to the test significance level (Wilks 2006). The test significance level used throughout the analysis, unless otherwise specified, is 95% (αLS = 0.05), so the 45° hatch marks in Figs. 6 and 7 represent grid points with 0.05 ≤ p ≤ 1.0: that is, areas where the test statistic fails to reject the null hypothesis.

After the local tests are performed, it still remains to test for the overall, collective significance of the differences in means between fields; this evaluation is called field significance. If we take, for example, the total change in daily averaged maximum temperatures (Fig. 7c), 89.51% of land area rejects H0 at the 95% level, but what is the probability that this could have occurred by chance? What percent of land area with significant changes would be equaled or exceeded by accident? These questions arise given the spatial correlation of the underlying data (Livezey and Chen 1983; Elmore et al. 2006). Spatial correlation describes how variations in one grid point are reflected at or affected by other grid points because of processes larger than the grid’s horizontal resolution. To answer these questions, a null distribution of the percent of land area with significant test statistic results is needed.

A Monte Carlo method is proposed to develop such a distribution to test for field significance (Livezey and Chen 1983). In a Monte Carlo simulation, the experiment tested is duplicated with fictitious data generated in the same manner as the original but in which the null hypothesis is true; it is then repeated many times with random inputs. The percent of grid points with statistically significant test statistic results at the αLS level in each Monte Carlo calculation is recorded and used to estimate a specified percent tail of the distribution and to set a threshold fraction required for field significance at a αFS level. In this research αLS = αFS = 0.05, but different significance levels could be used in the local and field significance tests.

As mentioned above, the random inputs for Monte Carlo experiments should be performed to mimic the original data being tested to maintain the spatial correlation of the underlying data, whose effects are the main reason to test for field significance. Another important consideration is that enough simulations need be performed to estimate the probability distribution accurately. To comply with the first consideration, a day from the PRESENT1 and PAST2 scenarios is chosen randomly with replacement, the resulting latitude–longitude slice is assigned as day 1 in alternative PRESENT1 and PAST2 sets, and the process is repeated ni times until two alternative datasets are completed. The test statistic (4) is applied to the original and alternative sets, and the percent of grid points with statistical significant results at αLS is stored, referred to as pctLS. The process is repeated 10 000 times (with replacement) to comply with the second consideration. After all trials are completed, the percent of points with significant differences pctFS is recorded and (1 − αFS) quantile of percentages determined. If the percent of grid points showing significant differences is larger than this threshold, then the difference in means observed has field significance. This procedure has been applied to situations where local significance analysis of the historical and fictional datasets clearly represented an acceptance of H0 (Livezey and Chen 1983; Maurer and Lettenmaier 2003; Feddema et al. 2005; Elmore et al. 2006). In our case, much consideration was put in the resampling of the original data to obtain fictional sets that would represent a failure to reject H0.

The entire statistical significance determination process employed in this research can be summarized as follows:

  1. Local significance:
    1. The t tests are performed to determine the statistical significance of the difference in means between two scenarios in Table 1.
    2. The percent of grid points over land regions with statistically significant test statistic results below the αLS level is determined pctLS.
  2. Field significance:
    1. A Monte Carlo method is used to estimate the distribution of the percent of points that would show significant differences by chance pctFS.
    2. The percent of points with significant differences pctLS is compared with the upper αFS quantile of pctFS to determine if the differences in means have field significance.
The fraction of land area from Figs. 6 and 7 with significant differences in means at the 95% confidence level (i.e., pctLS), along with the required percent for field significance, is summarized in Tables 2 and 3 for maximum and minimum temperatures, respectively. Clearly, the results for difference in the means in maximum and minimum temperatures are located well out of the upper tail and therefore are considered to have field significance. A more adequate resampling technique for this case should be explored, since the procedure performed seems to generate alternative datasets that do not show the level of randomness observed in the fictitious sets generated for similar field significance tests in previous studies.
Table 2.

Local and field statistical significance results for maximum temperature.

Table 2.
Table 3.

As in Table 2, but for for minimum temperature.

Table 3.

4. Climate impacts of LCLU changes and global warming in coastal tropical regions

Climate impacts on air temperature are analyzed by the calculation of the difference between the combinations of simulations in Table 1 that have a physical meaning. These combinations explain the climate impacts due to LCLU changes under present and past atmospheric conditions, the climate impacts due to global warming with present and past LCLU specifications, what is determined to be total change due to LCLU changes and global warming, and the nonlinear interaction between the two factors (Stein and Alpert 1993). Table 4 summarizes these different combinations, what contribution to the total change they explain, and which panels in Figs. 69 show each effect. The results presented herein refer to the 5-yr averages of each simulation scenario described in Table 1.

Table 4.

Climate impact scenario description and identification.

Table 4.

a. Impacts on minimum and maximum temperatures: Horizontal distributions

Since the most noticeable indicator of climate change is changes in near-surface temperature, 2-m AGL differences are analyzed for mean daily minimum and maximum temperatures over land areas in northeast Puerto Rico at a 1-km horizontal resolution grid (Figs. 6 and 7, respectively). Results for minimum-temperature total change (Fig. 6c) show positive (i.e., warming) differences that range from 1.0° to 2.5°C throughout the domain; the interpretation of this temperature difference field lies in the results for the signal separation exercise. Mean daily minimum temperature differences due to LCLU changes between the two study periods, while driving the model with present atmospheric conditions and SSTs and with 1955–59 atmospheric and SST conditions, are shown in Figs. 6a and 6b, respectively. Statistical significance tests reveal that minimum-temperature differences over land areas for the two LCLU change cases are significant at αLS = 0.05 only over major urban regions, though the full distribution does not clearly indicate an urban influence but a relatively homogeneous pattern of temperature differences over land regions. Differences are also significant in pockets of increased temperature differences over portions of El Yunque and the Central Mountains. The global warming signal, with present LCLU surface characteristics (Fig. 6d), shows results significant over most of the land region in the domain, with the exception of elevated terrain areas, specifically along the Central Mountain ridge and El Yunque. Figure 6e also shows the global warming signal in the mean daily minimum temperature differences but with past LCLU surface characteristics. Both global warming scenarios show the same horizontal near-surface heating trend of 0.8°–1.6°C, with increased heating of ~1.6°C along the northeastern and southeastern coastlines and interior lowlands.

The fact that Figs. 6a and 6b are strikingly similar indicates that the impact of LCLU changes on minimum temperatures during the period analyzed is relatively independent of large-scale climate conditions and vice versa when looking at Figs. 6d and 6e. The difference between the temperature changes in Figs. 6a and 6b (in Fig. 6f), which yields the same results as the calculation of the differences between Figs. 6d and 6e, shows small temperature differences with no statistical significance throughout the domain. This indicates that the sum of the individual effects is not significantly different from the combined linear effect of both (Stein and Alpert 1993). Thus, for late-twentieth-century changes in minimum temperature, the differences induced by LCLU changes and large-scale climate change are independent and able to be combined linearly to obtain the total change distribution (Fig. 6c).

Maximum-temperature differences due to LCLU changes, while driving the model with present atmospheric conditions and SSTs (Fig. 7a), clearly show the influence of the urban areas (including the SJMA, Caguas, and the ring of ongoing development around El Yunque) on early afternoon daytime high temperature changes. These urban areas show statistically significant (at αLS = 0.05) increases that range between 2° and 5°C. The rest of the domain shows temperature increases over land regions between 0.8° and 2.4°C in isolated locations where LCLU conversions less dramatic than urbanization have occurred from 1951 to 2000 (see Fig. 3).

Results for temperature differences due to LCLU changes, while driving the model with the 1955–59 atmospheric and oceanic conditions, show the same pattern of maximum-temperature change as the results with present climate conditions in Fig. 7a (Fig. 7b). These results show that the strong SJMA UHI has a significant influence in near-surface air temperatures and that this influence has been increasing over time, which is in agreement with previous studies on the effects of the SJMA UHI (Velazquez-Lozada et al. 2006; González et al. 2006).

Results for maximum-temperature differences due to global warming while keeping the present LCLU specification constant (Fig. 7d), present a more homogeneous pattern of significant temperature changes over land regions than the two LCLU change cases in Figs. 7a and 7b. The pattern shows a temperature-difference gradient increasing inland from the coast to near the Central Mountain ridge. Of interest to climate impacts in coastal tropical regions is the slight cooling observed along the easternmost coastline of Puerto Rico (0°–0.8°C), due to the global warming signal, which could be attributed to increased surface trade winds orthogonal to the coastline in the eastern to southeastern region of the island.

Maximum-temperature differences due to global warming, using the past LCLU specifications as surface characteristics (Fig. 7e), show the same temperature increase pattern as Fig. 7d. The coastal cooling in the global warming signal (Figs. 7d,e) might be mitigating the heating occurring along the same coastline because of ongoing urbanization, as observed in the LCLU change signal (Figs. 7a,b). This is of extreme importance in the northeastern part of the domain, where residential and tourism projects account for much of the development around El Yunque, a naturally sensitive protected rain forest reserve. Maximum-temperature changes along the northeastern coast of Puerto Rico, with the exception of the section of coast corresponding to the SJMA, were not significant at αLS = 0.05 and αLS = 0.1, possibly because of the low values found in the differences in mean daily maximum temperatures for those areas.

The maximum-temperature changes in Figs. 7a and 7b show nearly similar temperature change patterns, indicating that the impact due to LCLU changes during the approximate 50-yr period between 1951 and 2000 is relatively independent of the large-scale climate conditions of the 5-yr periods of 1955–59 and 2000–04 and vice versa (Figs. 7d,e). This is evident in the small temperature change values (with a low significance) of the difference between the maximum temperatures in Figs. 7a and 7b or between Figs. 7d and 7e (i.e., Fig. 7f). This indicates that changes in maximum temperatures due to LCLU changes and due to global warming are again independent from each other and therefore can again be added linearly to obtain the total change distribution (Fig. 7c). Regarding the small coastal cooling discussed above in the global warming signal as a mitigating factor to temperature increases in coastal urban regions observed in the LCLU change signal, the total change shows that the impact due to LCLU changes still dominates the overall maximum-temperature change pattern in the region of interest (Fig. 7c), especially in areas with more urbanization and only a small coastal-cooling effect. In general, the horizontal heating pattern produced by the global warming signal, especially in the interior of the domain, has an enhancing effect to the maximum-temperature changes produced by the LCLU change signal.

b. Impacts on minimum and maximum temperatures: Vertical profiles

To observe the relative and total impacts that LCLU changes and global warming have on the boundary layer (BL), vertical cross section plots were constructed in the north–south plane identified in Fig. 3 at 66.05°W longitude, which runs through the SJMA, Caguas, and the Central Mountains. The vertical cross sections in Figs. 8 and 9 show temperature differences and wind component (vectors constructed with east–west υ and vertical w components) differences at the time of minimum and maximum temperature, respectively. For minimum temperature, differences due to LCLU changes with present atmospheric conditions and SSTs (Fig. 8a) are limited to the near-surface region occupied by the SJMA (centered around 18.4°N latitude) and are almost nonexistent at higher vertical levels. The LCLU change signal when driving the model with past atmospheric and SST conditions (Fig. 8b) shows changes in temperature only near the surface over the city with no changes in temperature or wind at upper levels. The global warming signal with present LCLU (Fig. 8d) produces positive temperature changes from the surface up to 900 m, a band of no changes with a nucleus of small negative values (cooling) from 900 to 2100 m, and then increased heating at higher elevations. This pattern of minimum-temperature changes in the vertical might be explained as the influence of the input NCEP data, which shows similar behavior in the vertical (Comarazamy and González 2011). Given the similarity between Figs. 8a and 8b and between Figs. 8d and 8e, the LCLU change signal and the global warming signal combine to produce the vertical total change pattern (Fig. 8c), which in turn is similar to the global warming signal with the added near-surface air heating over the city produced by the LCLU change signal.

As with the horizontal distribution of temperature changes, more complexity exists in the vertical cross section plots at the time of maximum temperature (Fig. 9) than in the plots at the time of minimum temperature. Even though noticeable differences exist in the wind field because of LCLU changes (Figs. 9a,b), changes in temperature are still limited to the near surface, penetrating farther into the atmosphere than in the minimum-temperature LCLU cases, which is particularly noticeable while driving the model with past atmospheric and oceanic conditions (Fig. 9b).

The total change of maximum-temperature differences (Fig. 9c) resembles closely the results of differences due to the global warming signal, whether with present or past LCLU (Figs. 9d and 9e, respectively). This result is a consequence of the small differences found in the LCLU change signal results and of the similarity between Figs. 9a and 9b and between Figs. 9d and 9e. Maximum-temperature total differences and differences due to the two global warming scenarios (Figs. 9c–e, respectively), penetrate up to 1500 m, driven by vertical motions generated by a convergence zone located north of the Central Mountains. This circulation generates a cell with subsidence in the mountain range ridge, the location of the convergence zone in the simulations driven by past atmospheric conditions. A band of no temperature changes from 1500 to about 3000 m is where the main liquid water differences were found and where a climatological cloud layer is believed located (Comarazamy and González 2011).

5. Summary and concluding remarks

The individual and combined effects of LCLU changes and global warming, as evidenced in long-term regional climate change, in tropical coastal regions were investigated with the use of an integrated mesoscale atmospheric modeling approach, taking the northeastern region of Puerto Rico as the test case. To achieve this goal, an ensemble of climate simulations is performed, combining two LCLU and global warming scenarios. Reconstructed agricultural maps and sea surface temperatures form the past (1955–59) scenario, while the present (2000–04) scenario is supported with high-resolution remote sensing LCLU data. Results show that LCLU changes produced the largest air temperature differences over heavily urbanized regions and that these changes occur near the surface. The influence of the global warming signal induces a positive inland gradient of maximum temperature, possibly due to increased trade winds in the present climatology. In terms of minimum temperatures, the global warming signal induces temperature increases along the coastal plains and inland lowlands.

The reported changes in temperature due to LCLU changes and global warming in a coastal tropical region could be representative of various continental or island scale cases. The case of small island developing states (SIDSs) is particularly important, because of their high vulnerability to global climate change and low mitigation capacity. In terms of local environmental changes, SIDSs and other continental coastal tropical regions are responsible for much of the global LCLU changes and therefore for its underlying climate impacts. Recently, local governments in Puerto Rico and Dominican Republic have made great efforts to protect their natural resources and to assign tropical montane rain and cloud forests as natural reserves (Environmental and Natural Resources Ministry of the Dominican Republic 2007; Land Use Planning Office of Puerto Rico 2006), but uncontrolled urban sprawl remains an important contributor to climate change due to LCLU changes. A number of secondary and tertiary social, economic, and cultural implications of climate change also exist because of the combined effects of LCLU changes and global warming in coastal tropical regions. To make adaptation and mitigation strategies with a significant impact in addressing the problem, the most reliable information, with relevant factors included in the analysis, is needed. This is why comprehensive research, like the one presented here, is of extreme importance.

Proper climate simulations that are representative of the regions of interest are key to these efforts. In this work, an update of the atmospheric model surface characteristics was used that considered parameters extracted from high-resolution remote sensing information.

The findings discussed here apply to late-twentieth-century urbanization and climate change, a period that saw dramatic urbanization and other cases of LCLU changes in Puerto Rico while only modest large-scale warming was observed. The implication is that if global warming continues as projected (Meehl et al. 2007), presuming that urbanization has reached a relative stable extent (and may have some reduction of its impact because of expanded use of reflective and green roofs, greater urban vegetation, reforestation programs, etc.), the future may not show the same relative effects experienced during the study period in this document. Projections for future climate changes in tropical coastal regions could be performed using different greenhouse gas (GHG) emission scenarios (Solomon et al. 2007; Meehl et al. 2007) combined with statistical and dynamical modeling of LCLU change and urban growth (López et al. 2001; de Almeida et al. 2005; Velazquez-Lozada et al. 2006).

Where LCLU change includes urban areas with deep urban canyons and where surface energy balances could be affected by tall buildings, a more sophisticated urban canopy energy model should be coupled to the atmospheric model (Masson 2000; Martilli et al. 2002). In addition to improved input parameters and modeling parameterizations, a more adequate resampling method to perform Monte Carlo experiments, in tests of significance, should be explored.

Acknowledgments

This research was funded by NOAA/Cooperative Remote Sensing Science and Technology Center (CREST) Grant NA06OAR4810162. The atmospheric model simulations were performed at the High Performance Computing Facilities of the University of Puerto Rico at Río Piedras.

REFERENCES

  • Chen, A. A., , and M. A. Taylor, 2002: Investigating the link between early season Caribbean rainfall and the El Niño+1 year. Int. J. Climatol., 22, 87106.

    • Search Google Scholar
    • Export Citation
  • Comarazamy, D. E., , and J. E. González, 2008: On the validation of the simulation of early season precipitation on the island of Puerto Rico using a mesoscale atmospheric model. J. Hydrometeor., 9, 507520.

    • Search Google Scholar
    • Export Citation
  • Comarazamy, D. E., , and J. E. González, 2011: Regional long-term climate change (1950–2000) in the midtropical Atlantic and its impacts on the hydrological cycle of Puerto Rico. J. Geophys. Res., 116, D00Q05, doi:10.1029/2010JD015414.

    • Search Google Scholar
    • Export Citation
  • Comarazamy, D. E., , J. E. González, , J. C. Luvall, , D. L. Rickman, , and P. J. Mulero, 2010: A land–atmospheric interaction study in the coastal tropical city of San Juan, Puerto Rico. Earth Interact., 14. [Available online at http://EarthInteractions.org.]

    • Search Google Scholar
    • Export Citation
  • Cotton, W. R., and Coauthors, 2003: RAMS 2001: Current status and future directions. Meteor. Atmos. Phys., 82, 529.

  • Dai, A., , K. E. Trenberth, , and T. R. Karl, 1999: Effects of clouds, soil moisture, precipitation, and water vapor on diurnal temperature range. J. Climate, 12, 24512473.

    • Search Google Scholar
    • Export Citation
  • Daly, C., , E. H. Helmer, , and M. Quiñones, 2003: Mapping the climate of Puerto Rico, Vieques and Culebra. Int. J. Climatol., 23, 13591381.

    • Search Google Scholar
    • Export Citation
  • de Almeida, C. M., , A. M. V. Monteiro, , G. Câmara, , B. S. Soares-Filho, , G. C. Cerqueira, , C. L. Pennachin, , and M. Batty, 2005: GIS and remote sensing as tools for the simulation of urban land-use change. Int. J. Remote Sens., 26, 759774.

    • Search Google Scholar
    • Export Citation
  • Dickinson, R. E., , A. H. Sellers, , P. J. Kennedy, , and M. F. Wilson, 1986: Biosphere-atmosphere transfer scheme (BATS) for the NCAR Climate Community Model. NCAR Tech. Note NCAR/TN-275+STR, 69 pp.

  • Duchon, C. E., 1986: Temperature trends at San Juan, Puerto Rico. Bull. Amer. Meteor. Soc., 67, 13701377.

  • Elmore, K. L., , M. E. Baldwin, , and D. M. Schultz, 2006: Field significance revisited: Spatial bias errors in forecasts as applied to the Eta model. Mon. Wea. Rev., 134, 519531.

    • Search Google Scholar
    • Export Citation
  • Environmental and Natural Resources Ministry of the Dominican Republic, 2007: Políticas para la gestión efectiva del Sistema Nacional de Areas Protegidas de la República Dominicana. Environmental and Natural Resources Ministry of the Dominican Republic Rep., 59 pp.

  • Feddema, J., , K. Oleson, , G. Bonan, , L. Mearns, , W. Washington, , G. Meehl, , and D. Nychka, 2005: A comparison of a GCM response to historical anthropogenic land cover change and model sensitivity to uncertainty in present-day land cover representations. Climate Dyn., 25, 581609, doi:10.1007/s00382-005-0038-z.

    • Search Google Scholar
    • Export Citation
  • González, J. E., , J. C. Luvall, , D. Rickman, , D. E. Comarazamy, , and A. J. Picón, 2006: Urban heat island identification and climatologic analysis in a coastal, tropical city: San Juan, Puerto Rico. Urban Remote Sensing, Q. Weng and D. Quattrochi, Eds., CRC Press, 223–252.

  • Helmer, E. H., 2004: Forest conservation and land development in Puerto Rico. Landscape Ecol., 19, 2940.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471.

  • Kennaway, T., , and E. H. Helmer, 2007: The forest types and ages cleared for land development in Puerto Rico. GISci. Remote Sens., 44, 356382.

    • Search Google Scholar
    • Export Citation
  • Landsberg, H. E., 1981: The Urban Climate. Academic Press, 275 pp.

  • Land Use Planning Office of Puerto Rico, 2006: Plan de uso de terrenos de Puerto Rico. Land Use Planning Office of Puerto Rico Rep., 89 pp.

  • Lawton, R. O., , U. S. Nair, , R. A. Pielke Sr., , and R. M. Welch, 2001: Climatic impact of tropical lowland deforestation on nearby montane cloud forests. Science, 294, 584587.

    • Search Google Scholar
    • Export Citation
  • Lemonsu, A., , V. Masson, , and E. Berthier, 2007: Improvement of the hydrological component of an urban soil–vegetation–atmosphere–transfer model. Hydrol. Processes, 21, 21002111.

    • Search Google Scholar
    • Export Citation
  • Livezey, R. E., , and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111, 4659.

    • Search Google Scholar
    • Export Citation
  • López, E., , G. Bocco, , M. Mendoza, , and E. Dahau, 2001: Predicting land-cover and land-use change in the urban fringe: A case in Morelia city, Mexico. Landscape Urban Plann., 55, 271285.

    • Search Google Scholar
    • Export Citation
  • Lugo, A. E., , and E. Helmer, 2004: Emerging forests on abandoned land: Puerto Rico’s new forests. For. Ecol. Manage., 190, 145161.

  • Magaña, V., , J. A. Amador, , and S. Medina, 1999: The mid-summer drought over Mexico and Central America. J. Climate, 12, 15771588.

  • Malmgren, B. A., , A. Winter, , and D. Chen, 1998: El Niño–Southern Oscillation and North Atlantic Oscillation control of climate in Puerto Rico. J. Climate, 11, 27132717.

    • Search Google Scholar
    • Export Citation
  • Mansell, M., , and F. Rollet, 2006: Water balance and the behaviour of different paving surfaces. Water Environ., 20, 7–10.

  • Martilli, A., , A. Clappier, , and M. W. Rotach, 2002: An urban surface exchange parameterisation for mesoscale models. Bound.-Layer Meteor., 104, 261304.

    • Search Google Scholar
    • Export Citation
  • Martinuzzi, S., , W. A. Gould, , and O. M. Ramos González, 2007: Land development, land use, and urban sprawl in Puerto Rico integrating remote sensing and population census data. Landscape Urban Plann., 79, 288297.

    • Search Google Scholar
    • Export Citation
  • Masson, V., 2000: A physically-based scheme for the urban energy budget in atmospheric models. Bound.-Layer Meteor., 94, 357397.

  • Maurer, E. P., , and D. P. Lettenmaier, 2003: Predictability of seasonal runoff in the Mississippi River basin. J. Geophys. Res., 108, 8607, doi:10.1029/2002JD002555.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–846.

  • Myers, N., , R. A. Mittermeier, , C. G. Mittermeier, , G. A. B. da Fonseca, , and J. Kent, 2000: Biodiversity hotspots for conservation priorities. Nature, 403, 853858.

    • Search Google Scholar
    • Export Citation
  • Nair, U. S., , R. O. Lawton, , R. M. Welch, , and R. A. Pielke Sr., 2003: Impact of land use on Costa Rican tropical montane cloud forests: Sensitivity of cumulus cloud field characteristics to lowland deforestation. J. Geophys. Res., 108, 4206, doi:10.1029/2001JD001135.

    • Search Google Scholar
    • Export Citation
  • Oke, T. R., 1987: Boundary Layer Climates. Routledge, 435 pp.

  • Pielke, R. A., and Coauthors, 1992: A comprehensive meteorological modeling system—RAMS. Meteor. Atmos. Phys., 49, 6991.

  • Ragab, R., , P. Rosier, , A. Dixon, , J. Bromley, , and J. D. Cooper, 2003: Experimental study of water fluxes in a residential area: 2. Road infiltration, runoff and evaporation. Hydrol. Processes, 17, 24232437.

    • Search Google Scholar
    • Export Citation
  • Ray, D. K., , U. S. Nair, , R. O. Lawton, , R. M. Welch, , and R. A. Pielke Sr., 2006: Impact of land use on Costa Rican tropical montane cloud forests: Sensitivity of orographic cloud formation to deforestation in the plains. J. Geophys. Res., 111, D02108, doi:10.1029/2005JD006096.

    • Search Google Scholar
    • Export Citation
  • Saleeby, M. S., , and W. R. Cotton, 2004: A large-droplet mode and prognostic number concentration of cloud droplets in the Colorado State University Regional Atmospheric Modeling System (RAMS). Part I: Module descriptions and supercell test simulations. J. Appl. Meteor., 43, 182195.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., , and R. W. Reynolds, 2003: Extended reconstruction of global sea surface temperatures based on COADS data (1854–1997). J. Climate, 16, 14951510.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., , R. W. Reynolds, , T. C. Peterson, , and J. Lawrimore, 2008: Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006). J. Climate, 21, 22832296.

    • Search Google Scholar
    • Export Citation
  • Solomon, S., , D. Qin, , M. Manning, , M. Marquis, , K. Averyt, , M. M. B. Tignor, , H. L. Miller Jr., , and Z. Chen, Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.

  • Stein, U., , and P. Alpert, 1993: Factor separation in numerical simulations. J. Atmos. Sci., 50, 21072115.

  • Taylor, M. A., , D. B. Enfield, , and A. A. Chen, 2002: Influence of the tropical Atlantic versus the tropical Pacific on Caribbean rainfall. J. Geophys. Res., 107, 3127, doi:10.1029/2001JC001097.

  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

  • Tso, C. P., 1996: A survey of urban heat island studies in two tropical cities. Atmos. Environ., 30, 507519.

  • Van der Molen, M. K., 2002: Meteorological impacts of land use change in the maritime tropic. Vrije Universititeit Amsterdam Rep., 274 pp.

  • Van der Molen, M. K., , A. J. Dolman, , M. J. Waterloo, , and L. A. Bruijnzeel, 2006: Climate is affected more by maritime than by continental land use change: A multiple scale analysis. Global Planet. Change, 54, 128149.

    • Search Google Scholar
    • Export Citation
  • Velazquez-Lozada, A., , J. E. Gonzalez, , and A. Winter, 2006: Urban heat island effect analysis in San Juan, Puerto Rico. Atmos. Environ., 40, 17311741.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences. 2nd ed. Academic Press, 648 pp.

Save