SEARCH

SEARCH BY CITATION

Keywords:

  • land use/land cover change;
  • land-atmosphere interactions;
  • anthropogenic impact;
  • precipitation;
  • regional climate modeling;
  • Phoenix, Arizona

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] This paper is part 1 of a two-part study that evaluates the climatic effects of recent landscape change for one of the nation's most rapidly expanding metropolitan complexes, the Greater Phoenix, Arizona, region. The region's landscape evolution over an approximate 30-year period since the early 1970s is documented on the basis of analyses of Landsat images and land use/land cover (LULC) data sets derived from aerial photography (1973) and Landsat (1992 and 2001). High-resolution, Regional Atmospheric Modeling System (RAMS), simulations (2-km grid spacing) are used in conjunction with consistently defined land cover data sets and associated biophysical parameters for the circa 1973, circa 1992, and circa 2001 time periods to quantify the impacts of intensive land use changes on the July surface temperatures and the surface radiation and energy budgets for the Greater Phoenix region. The main findings are as follows: since the early 1970s the region's landscape has been altered by a significant increase in urban/suburban land area, primarily at the expense of decreasing plots of irrigated agriculture and secondarily by the conversion of seminatural shrubland. Mean regional temperatures for the circa 2001 landscape were 0.12°C warmer than the circa 1973 landscape, with maximum temperature differences, located over regions of greatest urbanization, in excess of 1°C. The significant reduction in irrigated agriculture, for the circa 2001 relative to the circa 1973 landscape, resulted in dew point temperature decreases in excess of 1°C. The effect of distinct land use conversion themes (e.g., conversion from irrigated agriculture to urban land) was also examined to evaluate how the most important conversion themes have each contributed to the region's changing climate. The two urbanization themes studied (from an initial landscape of irrigated agriculture and seminatural shrubland) have the greatest positive effect on near-surface temperature, increasing maximum daily temperatures by 1°C. Overall, sensible heat flux differences between the circa 2001 and circa 1973 landscapes result in a 1 W m−2 increase in domain-wide sensible heating, and a similar order of magnitude decrease in latent heating, highlighting the importance of surface repartitioning in establishing near-surface temperature trends. In part 2 of this study, we address the role of the surface budget changes on the mesoscale dynamics/thermodynamics, in context of the large-scale environment.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] Recent work suggests that comprehensive assessment of anthropogenic influence on climate must extend beyond the traditionally based evaluation of global-scale forcings [National Research Council, 2005]. In particular, it has been argued that the conventional radiative forcing framework is inappropriate when assessing the effect of patchy, heterogeneous, landscape change on near-surface climate, as this approach does not take into account a number of important factors, such as the modification of the surface energy balance and its effect on surface temperatures [Davin et al., 2007], significant regional changes that may alter large-scale circulation [Matsui and Pielke, 2006], and teleconnections to remote regions [Pielke et al., 2002]. Consequently, regional-scale studies quantifying the effect(s) of land use and land cover change (LULCC) through a regional lens have the potential to improve our overall understanding of the full spectrum of anthropogenic climate change, its impacts, and consequences.

[3] The effects of LULCC in urban/suburban areas, or areas that are rapidly being developed and converted in this way, are particularly important because this is where the majority of the world's population resides [Jin and Shepherd, 2005] (and the percentage is growing [Arnfield, 2003]). Inhabitants of large and expanding urban complexes located in arid and semiarid regions of the western United States are especially vulnerable to any changes in climate (either globally or regionally driven), due to reliance on increasingly limited water supplies [Barnett and Pierce, 2008; Barnett et al., 2008]. Consequently, in anticipation of potential western U.S. water crises [Barnett et al., 2004; Christensen et al., 2004; McCabe and Wolock, 2007; Seager et al., 2007] and with the hope of helping to mitigate economic and societal effects, some states located within the Colorado River Basin (CRB), such as Arizona, have recently developed their first drought plan [Jacobs et al., 2005].

[4] The Greater Phoenix region is the largest metropolitan area within the CRB. This urban complex has witnessed tremendous territorial expansion and population growth throughout the 20th century. During the first half of the century, expansion of the urban and suburban area occurred largely at the expense of decreases in the agricultural area immediately surrounding the urban core, while agricultural development and expansion continued at greater rates along the outer fringes of existing plots, serving to replace the preexisting seminatural desert and shrubland cover [Knowles-Yanez et al., 1999]. Throughout the second half of the century changing economic priorities led to a shift in urban and suburban development with increased conversion occurring from formerly desert landscapes as opposed to agricultural areas [Grimm and Redman, 2004].

[5] The rapid growth and associated LULCC of the Greater Phoenix region has altered key physical and biophysical properties that control the surface radiation balance and the ensuing transport of heat and moisture into the atmosphere, with significant implications for near-surface temperature and the overlying atmosphere [Cayan and Douglas, 1984; Hsu, 1984; Brazel et al., 2000; Baker et al., 2002; Grossman-Clarke et al., 2005; Brazel et al., 2007; Georgescu et al., 2008]. In addition, observationally based work has suggested that growth and development of this semiarid metropolitan region has altered summer precipitation downwind of the urban core [Diem and Brown, 2003; Diem, 2006; Shepherd, 2006].

[6] More recently, Georgescu et al. [2008] used present-day and presettlement land cover scenarios in conjunction with high-resolution regional climate simulations to demonstrate a first-order regional effect of LULCC on the summertime climate of the Greater Phoenix area. We briefly revisit these findings and then describe the extension of this research to quantify the effects of major LULCC since the early 1970s on the land surface radiation and energy budgets, and subsequently, on the thermodynamic and dynamic processes in the lower atmosphere.

[7] Georgescu et al. [2008] used a circa 1992 LULC data set based on Landsat and a presettlement land cover scenario, in which the anthropogenic landscape of irrigated agriculture and urban pixels in the 1992 data set was replaced with current seminatural vegetation, as land cover inputs to high-resolution (2-km grid spacing) simulations for three “dry” Julys and three “wet” Julys over the Greater Phoenix region. Differences (i.e., 1992 landscape minus presettlement) between standard meteorological parameters illustrated a dipole pattern for both temperature and dew point, with positive temperature (negative dew point) differences over the urban areas and negative temperature (positive dew point) differences over agricultural areas, revealing the important dual roles of anthropogenic landscape change. This pattern is magnified during the dry simulations as compared to the wet simulations. For example, for the dry experiments, positive temperature (negative dew point) differences between the 1992 landscape and presettlement scenario are enhanced over urban areas (agricultural areas) as compared to differences resulting from the wet simulations. The repartitioning of surface absorbed energy into greater sensible (latent) as opposed to latent (sensible) heating over the urban (agricultural) areas is responsible for warming (cooling) over these locations. At the regional scale, however, warming and cooling effects nearly cancel, leading to negligible net temperature tendency with LULCC. In addition to the thermal impact, we believe that these modeling results are the first to demonstrate enhanced precipitation downwind of the Greater Phoenix metropolitan complex.

[8] This two-part study builds upon the results of Georgescu et al. [2008] by investigating the sensitivity of the near-surface climate, atmospheric dynamics, and thermodynamics over the Greater Phoenix region to intensive LULCC since the early 1970s. Quantifying the response of the regional climate to this landscape alteration (e.g., reduction of irrigated agriculture in comparison to the increasing coverage of urban/suburban land) is critical for improving our understanding of the coupled land-atmosphere system over this semiarid urban complex. In addition, these results may provide important insights toward assessment of future regional climate change, and associated vulnerabilities, in response to projected urban expansion and continued population growth [Arizona State University, 2003], not only in this region, but also in other rapidly urbanizing semiarid regions of the United States (e.g., Las Vegas) or at similar locations elsewhere in the world.

[9] The specific research questions to be addressed in this two-part study are as follows:

[10] 1. How has the landscape for the Greater Phoenix region evolved since the early 1970s on the basis of analysis of consistently defined, multidecadal land cover data sets derived from aerial photography and satellite observations?

[11] 2. Using high-resolution summertime (i.e., when the North American Monsoon System (NAMS) is active) simulations for the Greater Phoenix region, can we quantify the response of the region's climate to the observed evolution of the landscape?

[12] 3. What is the role of land use forcing given interannual variability of the NAMS for wet and dry seasons?

[13] 4. Are the physical pathways responsible for summertime rainfall enhancement downwind of the urban core, as suggested by Georgescu et al. [2008], robust?

[14] This paper, part 1, focuses on documentation of the region's landscape evolution since the 1970s and its climatic impact on fundamental near-surface parameters (e.g., temperature and dew point) with a particular focus on the surface radiation and energy budgets. We also determine the effects of four distinct land use conversion themes (e.g., conversion from irrigated agriculture to urban land) and evaluate how the most important themes have contributed to the region's changing climate. Georgescu et al. [2009], part 2 of this study, address how the changes in the surface radiation and energy budgets in turn affect mesoscale dynamics and thermodynamics within the lower atmosphere and in context of the large-scale environment.

2. Method

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[15] Our assessment of the climatic effect of LULCC over the Greater Phoenix region is based on a number of experiments with a mesoscale numerical model. The evolution of the region's landscape over an approximate 30-year period since the early 1970s is documented based primarily on analyses of consistently defined, multidecadal LULC data sets, associated region-specific biophysical parameters, and selected Landsat images. To distinguish whether a particular landscape-induced signal in the meteorology operates during dissimilar large-scale settings, we repeat the procedure of Georgescu et al. [2008] by focusing on both dry and wet NAMS seasons (i.e., by experimental design, we simulate periods both when the large-scale signal is weaker, and when it is stronger, ensuring that the clear identification of the local/mesoscale landscape signal can be distinguished from that of the large scale). In addition, we focus on the summer season because it is during this time of the year that the surface exerts its greatest control on precipitation, as opposed to the atmospheric, synoptic-scale, control associated with the winter precipitation season. Furthermore, it is during the summer season when the region's climate claims its greatest influence on the stress levels of Greater Phoenix' residents, a distinguishing feature resulting from excessive summertime heat [Balling and Brazel, 1986] and flood-producing rainfall.

2.1. Model Description

[16] We use the Regional Atmospheric Modeling System (RAMS) Version 4.3 [Pielke et al., 1992; Walko and Tremback, 2000] for all experiments. RAMS, which has been used for a variety of applications ranging from large-eddy simulations to large-scale climate integrations, is a nonhydrostatic model that solves the full nonlinear equations of motion and includes a comprehensive soil and vegetation component, the Land Ecosystem-Atmosphere Feedback Model, version two (LEAF-2 [Walko et al., 2000]). RAMS also includes multiple parameterization options for convection, turbulence, radiation and cloud microphysics.

2.2. Landscape Representation

[17] The initial tasks of this two-part study involved the evaluation of the region's extensive landscape alterations from the early 1970s to 2001 and the development of consistent, multidecadal land cover and associated biophysical parameter gridded data sets for use in the RAMS/LEAF-2 simulations. In addition to analysis of selected Landsat images, land cover data sets for the circa 1973, circa 1992, and circa 2001 time periods were developed from available U.S. Geological Survey (USGS) digital land cover data sets for the conterminous United States. The source of our circa 1973 LULC data was the USGS Geographic Information and Analysis System (GIRAS) Land Use Data (LUDA) that was derived from 1972 and 1973 aerial photography for central Arizona: a 1-km dominant land cover subset was extracted from the USGS 1:250,000 scale digital data for Phoenix (L. Steyaert, personal communication, 2007). Our land cover data sets for circa 1992 and circa 2001 were adapted from the national land cover data sets that were originally derived by the USGS and the Multiresolution Land Characteristics Consortium (MRLC). They derived the 1992 National Land Cover Dataset (NLCD) from early 1990s Landsat TM imagery [Vogelmann et al., 2001] and the NLCD 2001 data set from early 2000s Landsat 7 ETM+ imagery [Homer et al., 2004].

[18] In addition to these three land cover data sets, Landsat images from various sources and time periods were also used to evaluate land use changes. For example, we used the 60-m reprocessed North American Landscape Characterization (NALC) image data set that provided Landsat Multispectral Scanner (MSS) time series triplicates over the greater Phoenix region for 1973, 1986, and 1991 (the Landsat MSS sensor was onboard Landsat 1–5). Because of differences in the land cover classification scheme of each USGS land cover data set (i.e., 1973, 1992, and 2001), the next step was to aggregate and adapt the classes to redefine a consistent set of harmonized land cover classes that adequately characterized all three time periods for use in the RAMS/LEAF-2 simulations. This task was somewhat simplified because the NLCD 1992 and the NLCD 2001 land cover classes represented modifications to the Anderson Level II classification scheme [Anderson et al., 1976] that was the basis for the 1973 LUDA data set. Our analysis resulted in the harmonized set of LEAF-2 land cover classes (Table 1). However, for convenience and consistency among the individual LEAF-2 data for each time period, we have adopted a simplified notation: NLCD73, NLCD92, and NLCD01. This notation does not imply that the 1973 data are a component of the MRLC program or methodology.

Table 1. Landscape Classification Used for All Numerical Experimentsa
LEAF-2 ClassαɛLAIvfraczo
  • a

    Also shown are the biophysical parameters that were determined for each LEAF-2 land cover class description: α, albedo; ɛ, emissivity; LAI, leaf area index; vfrac, vegetation fraction; zo, roughness length.

Water0.140.990.00.00.00
Barren0.210.860.50.330.05
Shrubland0.190.950.50.190.18
Urban low intensity0.160.900.40.400.50
Urban high intensity0.160.880.20.260.50
Grassland0.180.960.50.430.13
Irrigated agriculture0.180.956.00.750.06
Evergreen needleleaf0.110.965.00.681.00

[19] When performing modeling studies, it is vital to account for the appropriate, regionally specific biophysical parameters that characterize the area's landscape. We began with the standard LEAF-2 biophysical parameters, then analyzed the literature [e.g., Kuchler, 1964; Gibbens et al., 1996; Grimmond and Oke, 1999; Scurlock et al., 2001; Pielke, 2001], including results from semiarid field studies, and land products derived from satellite data to update and refine the biophysical parameters appropriate for the Greater Phoenix region (Table 1). We also analyzed satellite land products including the USGS 1-km AVHRR Normalized Difference Vegetation Index (NDVI) data (http://edcsns17.cr.usgs.gov/1KM/comp10d.html) and selected Moderate Resolution Imaging Spectroradiometer (MODIS) Land Products available at the Oak Ridge National Laboratory, and Distributed Active Archive Center (ORNL DAAC) with the online North American Subsetting and Visualization Tool (http://www.modis.ornl.gov/modis/modis_subsets5.cfm). For example, we followed the methodology of Zeng et al. [2000] and used the 90th percentile (based on the histogram of the seasonal NDVI maxima for each pixel) to calculate the appropriate vegetation fraction corresponding to the particular class of shrubland that predominates over the Greater Phoenix area during summer. The ORNL MODIS subsetting tool provided time series plots and images of MODIS albedo, LAI, surface reflectance, and vegetation index products for our region of interest.

2.3. Model Configuration and Experiments

[20] We briefly summarize the model setup here and review the numerical experiments performed. The RAMS simulation domain is identical to the one described by Georgescu et al. [2008]; for details about domain extent, vertical grid stretching, initialization and boundary conditions, the interested reader is referred to Georgescu et al. [2008]. A triply nested domain configuration, permitting the downscaling of the large-scale synoptic flow from the coarse outer grid to the fine inner grid, is centered over Phoenix' Sky Harbor International Airport. A graphical representation of the entire model domain is presented in Figure 1. We use the Kain-Fritsch convective parameterization scheme [Kain and Fritsch, 1992] for the outer two grids (i.e., grids 1 and 2), but leave it turned off for the fine grid (i.e., grid 3), where we make use of a 2-km grid spacing. Last, we configured RAMS to use 11 soil layers, from the surface down to a depth of 2m. Standard RAMS two-way interaction is used to communicate information between the grids and their respective parent grid. Model output is written every 3 h, the initial 30 h of all experiments are discarded, and analysis is performed for the month of July: 1 July, 1200 UTC, to 31 July, 1200 UTC.

image

Figure 1. Geographical representation of the RAMS nested grid configuration with topography overlaid and the number of grid cells for each grid shown as numbers in parentheses.

Download figure to PowerPoint

[21] LEAF-2, the soil-vegetation-atmosphere transfer scheme used in RAMS, is prognostic in energy and water exchange between the land surface and overlying atmosphere. Notably, LEAF-2 allows for the division of each grid cell into multiple patches, and thus accounts for subgrid-scale landscape heterogeneity, with each patch interacting with the overlying atmosphere individually, based on the particular vegetation, soil textural class, and soil moisture representation. For the outer two grids, in all experiments, the standard RAMS LULC data set, based on the USGS 1-km Advanced Very High Resolution Radiometer (AVHRR) Olsen Global Ecosystem (OGE) land cover data [Olson, 1994], was used. NLCD73, NLCD92, and NLCD01 landscape data were used only on grid 3 (consequently, all results presented will correspond to the grid 3 domain). As LEAF-2 permits the division of each RAMS grid cell into multiple patches, each RAMS 2-km cell was further subdivided into five patches of decreasing LULC fractional area (four plus water), thus retaining much of the information available in the higher resolution NLCD73, NLCD92 and NLCD01 data.

[22] One important issue is the appropriate representation of the effects of extensive warm season irrigation of the agricultural areas. To replicate these effects, we adopted the simple approach of forcing the soil moisture for each irrigated agricultural land cover patch (within grid 3 only) to saturation at each model time step. Sensitivity to this assumption was discussed by Georgescu et al. [2008].

[23] A total of 18 different simulations were produced for this study: three simulations each, corresponding to the trio of landscape scenarios, for Julys from six separate years (see Table 2 for a summary of all experiments performed). The choice of year for each selected July was made according to their classification as either wet or dry summer seasons as guided by an appropriate NAMS Index. The distinction between wet and dry monsoon seasons was made to better distinguish the effect of landscape-imposed atmospheric forcing. The hypothesis was that, during times of heavy and widespread rainfall, the regional contribution of the landscape will be difficult to distinguish given the large-scale forcing and atmospheric (as opposed to surface) control of precipitation, consistent with expectations based on previous work [e.g., Findell and Eltahir, 2003; Koster et al., 2004; Anyah et al., 2008]. By keying on wet and dry NAMS conditions, the possibility of greater or lesser sensitivity to LULCC given contrasting hydrometeorological regimes was more easily investigated.

Table 2. Summary of All 18 Experiments Performed
Landscape ConditionsYear From Which Initial and Boundary Conditions Were Used to Force RAMS
Wet YearsDry Years
NLCD731990, 1984, 19831994, 1989, 1979
NLCD921990, 1984, 19831994, 1989, 1979
NLCD011990, 1984, 19831994, 1989, 1979

[24] We chose a number of case study years on the basis of daily observed precipitation data encompassing the period 1948–1998 (i.e., Daily U.S. UNIFIED Precipitation data set [Higgins et al., 2000]). Given that North American Regional Reanalysis (NARR) data were not available prior to 1979, we restricted ourselves to selecting seasons after 1978. The high-resolution nature of these experiments and associated computational limits precluded us from simulating the entirety of each summer monsoon season (i.e., July, August, and September), for a greater number of years, and consequently, we were limited to simulation and analysis of the initial monsoon month (i.e., July) and for 6 years only.

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[25] A detailed evaluation of RAMS' ability to appropriately reproduce the regional climate of Greater Phoenix was presented by Georgescu et al. [2008] and will not be reproduced here.

3.1. Greater Phoenix Landscape Change

[26] As discussed in section 1, an initial goal of this research involved quantifying the evolution of the region's observed landscape change, since the early 1970s. The harmonized NLCD73, NLCD92, and NLCD01 LULC data prepared for this research allow us to characterize landscape trends over time. The major landscape conversion themes are addressed below.

[27] Figures 2a, 2b, 2c, and 2d present the fractional land cover difference patterns (NLCD01 fraction of total area minus NLCD73 fraction of total area) for urban land area (Figure 2a), irrigated agricultural land area (Figure 2b), shrubland area (Figure 2c), and forest area (Figure 2d), while Figures 3a, 3b, 3c, and 3d illustrate a similar calculation but for NLCD92 minus NLCD73. We focus on these landscape alterations as they represent the dominant LULC themes undergoing conversion from 1973 to 2001.

image

Figure 2. GRID 3 fractional land cover difference (NLCD01 fraction of total area minus NLCD73 fraction of total area) in (a) urban land area, (b) irrigated agriculture land area, (c) shrubland area, and (d) evergreen forest area.

Download figure to PowerPoint

image

Figure 3. GRID 3 fractional land cover difference (NLCD92 fraction of total area minus NLCD73 fraction of total area) in (a) urban land area, (b) irrigated agriculture land area, (c) shrubland area, and (d) evergreen forest area.

Download figure to PowerPoint

[28] Inspection of Figures 2 and 3 suggests temporal trends within the NLCD73, NLCD92, and NLCD01 land cover data sets. The changes in urban, agricultural, and shrubland cover since 1973 across the Phoenix metropolitan area are central to this study. For example, urbanization occurred to the northwest and southeast of central Phoenix as urban/suburban development proceeded from NLCD73 to NLCD92. Urban sprawl continued unabated during the ensuing decade, as urban land expanded in nearly all directions. These results are consistent with supporting evidence found in the literature [e.g., Knowles-Yanez et al., 1999] and highlight the importance of broad landscape comparison to complement pixel-by-pixel change assessments. The majority of the increase in urban land was at the expense of losses in plots of irrigated agriculture. However, gains in urban land also occurred at the expense of seminatural shrubland. This is especially evident in a region just east and north of the central business district of Phoenix (centered at about 33.6° N/112° W; see Figure 2c for the difference between NLCD01 and NLCD73).

[29] Figures 2c, 2d, 3c, and 3d show temporal trends in the areas of shrubland and evergreen forest across the northern part of our fine grid (i.e., north of about 34° N latitude). In general, the data suggest that shrubland area has increased over time as a result of the reduction in the evergreen forest area. The NLCD 2001 area of evergreen forest across the north is generally consistent with the area of evergreen forest that was mapped as part of the detailed ecoregional Gap Analysis Program (GAP), in part, based on 1999–2001 Landsat ETM+ images [Prior-Magee et al., 2007]. Potential physical causes for the decrease in evergreen forest area include increased burned area due to wildfire activity [Westerling et al., 2003; Collins et al., 2006], extensive tree mortality resulting from insect infestation and disease [U.S. Department of Agriculture, 1998; Williams et al., 2008], commercial logging [Raish et al., 1997], and urbanization [Raish et al., 1997]. Each of these land cover changes would lead to regenerating vegetation primarily consisting of shrub and scrub species. The evidence does suggest that these trends are real, although our model results for this northern zone are tempered given LULC data set ambiguity and a dearth of detailed ground validation observations.

[30] Overall, the landscape change analysis presented in this section shows the consistent nature of LULCC that has occurred from the early 1970s, through the 1990s and finally, through 2001. Of the trio of landscape snapshots in time, NLCD73 and NLCD01 illustrate the most extreme combinations of LULC: NLCD73 has the greatest coverage of irrigated agriculture and the least coverage of urban land, while NLCD01 has the least coverage of irrigated agriculture and the greatest coverage of urban land. Therefore, results presented from this point forward will reflect differences among parameters resulting from the most extreme landscape cases, i.e., differences between NLCD01 and NLCD73.

3.2. Simulated Differences in Temperature and Dew Point

[31] Figures 4a, 4b, 4c, and 4d show the RAMS simulated ensemble differences (NLCD01 minus NLCD73) for both air temperature (Figures 4a and 4c) and dew point temperature (Figures 4b and 4d), for all three wet and dry years. For both the wet and dry years, maximum warming, of the order of 1°C, occurs over the areas of greatest urbanization. As reported by Georgescu et al. [2008] concerning their analysis of hypothetical landscape change, peak warming is enhanced during the dry years. In contrast to the aforementioned analysis, however, the extent of cooling has diminished in size considerably for NLCD01, a trend reflective of the reduction in irrigated agriculture during this period. The wet simulations reveal a combination of warming and cooling to the north of the Greater Phoenix area, while the dry simulations show a widespread area of enhanced cooling over this same region. Both sets of experiments illustrate a regional warming of 0.12°C during the roughly three decade period of landscape change, in qualitative agreement with recent estimates of urban heat island tendency over the Phoenix area [Stone, 2007].

image

Figure 4. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in (a) first atmospheric level [24.1 m] air temperature (°C) and (b) dew point (°C), for all three wet years; (c) first atmospheric level [24.1 m] air temperature (°C) and (d) dew point (°C), for all three dry years.

Download figure to PowerPoint

[32] The decrease in dew point temperature (Figures 4b and 4d) resulting from the rapid decline in irrigated agriculture is centered over areas that have experienced the greatest urbanization (compare Figure 4 to Figure 2). Although both sets of simulations indicate a drying tendency, the dry experiments illustrate a regional enhancement of roughly twice the magnitude seen in the wet experiments. Balling and Brazel [1986] analyzed 88 years (1896–1984) of dew point temperature and relative humidity (RH) over the Greater Phoenix area, and find that RH values fell at a statistically significant rate over this time period, a decrease they attribute largely to increases in temperature. Despite contentions that dew point was on the rise, Balling and Brazel found no such long-term trend but did find evidence of a recent downward trend which they attribute to recent reduction of agricultural lands. Though changes in dew point are nearly nonexistent to the north of Greater Phoenix for the wet simulations, there is a pronounced moistening (of about 0.2°C) for the dry years, coinciding with the area of cooling noted previously.

[33] Landscape modifications have also impacted the daily simulated monthly average of maximum and minimum low-level air temperature (Figures 5a, 5b, 5c, and 5d). Daily maximum temperature changes (Figures 5a and 5c) show similar patterns for both the wet and the dry years, although as already noted with the mean daily temperature, maximum temperature differences for the dry simulations are enhanced relative to the wet year experiments. In contrast to simulated maximum low-level air temperatures, the greatest effect on minimum temperature changes (Figures 5b and 5d) does not occur directly over those areas where the landscape was altered, but is shifted to the north of the Greater Phoenix area. In addition, the peak increase in minimum (i.e., nighttime) temperatures is less than half the magnitude of the peak increase in maximum (i.e., daytime) temperatures. The regional nighttime warming, however, is slightly greater than the daytime warming (0.02°C), for both the wet and dry year experiments, indicative of a decrease in the regional diurnal temperature range (DTR) over the entire domain.

image

Figure 5. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in (a) first atmospheric level [24.1 m] daily maximum air temperature (°C) and (b) daily minimum air temperature (°C), for all three wet years; (c) same as Figure 5a, but for all three dry years, and (d) same as Figure 5b but for all three dry years.

Download figure to PowerPoint

[34] Notwithstanding the regional decrease in the DTR, a result consistent with the phenomenon of urbanization [e.g., Kalnay and Cai, 2003], the precise locales where urbanization occurred actually produced an increase in the DTR, as maximum temperatures increased more than minimum temperatures. Daytime heating of the surface in this region coincides with weak wind flow, and the additional absorption of radiation by the landscape drive the local temperature response. As evening and nighttime draw near, wind flow, in a climatological sense, increases. Therefore, advection plays an important role in translating and rearranging low-level air over the domain.

[35] While this helps explain why maximum nighttime warming does not coincide with the precise areas of greatest urbanization in our simulations, there is still the question of whether the simulated magnitude of warming and impact on the DTR over the city is reasonable. The thermal field of the urban environment is in large part determined by the choice of building and road construction materials. In addition to anthropogenic heat release, a characteristic not taken into account by RAMS, thermal properties such as thermal diffusivity, heat capacity, and surface emissivity of the three-dimensional urban structure may affect both the daytime and nighttime urban environment [e.g., Hamdi and Schayes, 2008]. In addition, geometric effects of the city structure also play an important role, for example, through surface shading from and trapping of insolation during daytime. The geometric configuration, commonly referred to as the aspect ratio effect, also influences nighttime emission of longwave radiation to space. The impact on longwave loss is directly related to the sky view factor [e.g., Oke, 1987]. The lack of such a comprehensive urban canopy parameterization in RAMS, able to account for the presence of and interaction between building walls, rooftops, road surfaces, and the overlying atmosphere, likely results in an underestimate of the nocturnal UHI intensity. Therefore, the effects of urbanization we simulate (on temperatures and atmospheric processes) likely represent underestimates as well.

3.3. Simulated Differences in Albedo and Surface Energy Balance

[36] To better understand the pathways by which landscape evolution has altered the near-surface temperature, we analyze the regional surface radiation budget.

[37] RAMS simulated ensemble differences in net surface broadband albedo are presented in Figure 6 for both the wet (Figure 6a) and dry years (Figure 6b). The regional difference in albedo pattern, for both hydrometeorological regimes, is nearly identical, though peak values are enhanced for the dry years. The greatest decreases in surface albedo are apparent near the center of the domain. This region underwent significant urbanization during the three-decade period we are considering (see Figure 2 and section 3.1). Landscape conversion occurred at the expense of seminatural shrubland, whose albedo values are greater than that of urban areas, resulting in a strong positive radiative forcing on the shortwave component of the surface radiation budget (Figures 7a and 7b). Over this local region, the dry years show peak shortwave absorption differences in excess of twice the magnitude of the wet years. Differences in incident radiation for the dry years are generally consistent with changes in land cover. Though the primary effect on the radiation budget is through changes in albedo, incident radiation is also influenced, to an extent, by changes in cloud cover, particularly during the wet years.

image

Figure 6. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in surface albedo for the (a) wet years and (b) dry years.

Download figure to PowerPoint

image

Figure 7. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in incident radiation (W m−2) for the (a) wet years and (b) dry years. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in net longwave flux (W m−2) for the (c) wet years and (d) dry years.

Download figure to PowerPoint

[38] Surrounding this centralized region, and marking the extent of Greater Phoenix' metro limits, are smaller pockets of relatively minor increases and decreases in broadband albedo (±0.02). Though the landscape was significantly modified in these areas as well, urbanization occurred at the expense of irrigated agriculture (in contrast to the aforementioned changes at the expense of shrubland), a land use class possessing similarly low values of albedo. Therefore, our results suggest that the net effect on the observed change in the shortwave component of the surface radiative budget over the Greater Phoenix metropolitan area is, in large part, due to the conversion from natural shrubland to urban landscape.

[39] To the north of the Greater Phoenix area, a considerable increase in simulated broadband albedo is noted for the dry year experiments (see Figure 6b). Flanked by these large pockets of albedo increase are relatively smaller patches of albedo reduction. This broad northern region experienced increases in natural shrubland at the expense of evergreen needleleaf forest from the 1970s to the present decade. While the wet years illustrate a patchwork of increases and decreases in incident shortwave radiation over this northern area, in part, due to greater changes in cloud cover, the increased shrubland albedo results in a coherent decrease in incident shortwave radiation for the dry years. Consequently, we attribute the decrease in low-level temperatures over the northern one third of our grid 3 domain (i.e., the fine grid) to changes resulting from decreased surface absorption of shortwave radiation due to the conversion from evergreen needleleaf forest to shrubland.

[40] RAMS-simulated ensemble differences in net longwave flux are presented in Figure 7 for both the wet (Figure 7c) and dry years (Figure 7d). The regional pattern for both hydrometeorological regimes is particularly similar although, as before, the dry year experiments show an enhancement of this radiation budget component. Peak longwave radiation loss occurs over those areas that witnessed the greatest urbanization. In particular, the conversion from irrigated agriculture to an urban landscape facilitated the removal of a significant amount of (low level) atmospheric water vapor that otherwise, through increased downward longwave emission, maintained greater levels of surface longwave absorption. In addition, it is not just removal of water vapor that is important in affecting a change in surface longwave emission. Because of increased surface temperatures, surface emission of longwave radiation must also increase, in accordance with the Stefan-Boltzmann law, thereby lowering the surface longwave even further for NLCD01 relative to NLCD73.

[41] The mean monthly differences in net (shortwave plus longwave) radiative flux (not shown) indicate patchy increases and decreases over the Greater Phoenix area, illustrating the twin contributions of urbanization resulting from shrubland conversion (causing an increase in net surface radiation due to increased incident radiation) and urbanization resulting from extensive loss of irrigated agriculture (causing a decrease in net surface radiation due to decreased longwave radiation). To the north of the Greater Phoenix area, the effect of higher shrubland albedo and its impact on net shortwave radiation results in an overall decrease in net surface radiation.

3.4. Simulated Differences in Regional Surface Fluxes

[42] RAMS simulated ensemble differences in surface sensible heat flux are presented in Figure 8 for the wet (Figure 8a) and dry (Figure 8b) year experiments. The chief difference between the wet and dry year experiments is the enhancement in the peak flux values. As a result of compensating increases and decreases in flux magnitude, regional averages are similar for both sets of experiments, though both indicate a small domain-wide increase in sensible heating of the order of 1 W m−2. Areas where mean monthly flux differences are greatest (60–100 W m−2) coincide with those areas where maximum warming occurred for NLCD01 as compared to NLCD73.

image

Figure 8. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in surface sensible heat flux (W m−2) for the (a) wet years and (b) dry years. RAMS simulated ensemble differences (NLCD01 minus NLCD73) in surface latent heat flux (W m−2) for the (c) wet years and (d) dry years.

Download figure to PowerPoint

[43] Differences in latent heating (Figures 8c and 8d) show a similar, but reversed pattern, with latent heat flux reduction (60–100 W m−2) occurring primarily over regions where irrigated agriculture was converted to an urban landscape. Latent heat flux values also decreased over those areas where shrubland was converted to urban land, although the reduction in magnitude was generally less than half of that seen in the irrigated agriculture to urbanization conversion.

[44] Alternating pockets of higher and lower fluxes in net surface heat flux difference (sensible plus latent) are evident over the Greater Phoenix area (not shown). Because of the increase in surface sensible heating, areas that underwent a conversion from natural shrubland to urban land experienced an increase in surface heat flux. In contrast, regions to the northwest and southeast of Phoenix (recall that these areas underwent significant urbanization at the expense of irrigated agriculture) experienced a considerable reduction in net surface heat flux, with peak decreases for the dry year experiments in excess of 20 W m−2 for the dry year experiments.

3.5. Simulated Differences in Diurnal Cycle

[45] To better understand the impact of landscape evolution on the diurnal cycle, we now explore the individual effects of four distinct themes of landscape transformation that occurred over the Greater Phoenix region during this time period. In particular, we investigate the near-surface atmospheric effect resulting from two different types of urbanization: (1) conversion of shrubland to urban land and (2) conversion of irrigated agriculture to urban land. In addition, we also assess the effect of two different types of increase in seminatural shrubland: (1) conversion of irrigated agriculture to shrubland and (2) conversion of evergreen forest to shrubland.

[46] Here, we only show averages over those grid cells that exhibited a shift in dominant land cover. For example, Figures 9a and 9b present the RAMS-simulated time series of diurnally averaged first atmospheric level temperature differences for the wet years (Figure 9a) and the dry years (Figure 9b), for those grid cells that had at least 50% seminatural shrubland fractional coverage in NLCD73 and at least 50% urban fractional coverage for NLCD01 (the dash-dot-dot-dashed line), i.e., designating only those grid cells that underwent a conversion from predominantly shrubland to predominantly urban land. Similar calculations are performed for conversion of irrigated agriculture to shrubland, evergreen forest to shrubland, and agriculture to urban land.

image

Figure 9. RAMS simulated time series of diurnally averaged first atmospheric level [24.1 m] temperature (°C) differences (NLCD01 minus NLCD73) for those grid cells that experienced the following shift in dominant land cover type (from NLCD73 to NLCD01): from shrub to urban (dash-dot-dot-dashed), from irrigated agriculture to shrub (dash-dotted), from evergreen needleleaf to shrub (dot-dotted), from irrigated agriculture to urban (dash-dashed), and all land points on grid 3 (solid), (a) for all three wet years and (b) for all three dry years; (c) same as Figure 9a but for dew point temperature (°C) differences (NLCD01 minus NLCD73) for the wet years and (d) for all three dry years.

Download figure to PowerPoint

[47] The net effect on near-surface temperature for all pixels depicts an overall warming that peaks in the late afternoon/early evening (solid, black line for both plots). The greatest contributions to warming result from the shift in irrigated agriculture to urban land and from the conversion of shrubland to urban land. The previously noted albedo reduction due to forest-to-shrubland conversion results in a temperature reduction that coincides with the time of simulated maximum solar heating.

[48] The simulated effect on dew point temperature differences for the same land cover themes is presented in Figure 9 for both the wet years (Figure 9c) and dry years (Figure 9d). For all pixels, drying is roughly twice the magnitude during the dry years as compared to the wet years. The greatest contribution to drying results from the shift in irrigated agriculture to urban land and from the conversion of shrubland to urban land. The significance of the diurnal cycle is manifested as a rapid increase in the rate of drying near the time of simulated sunrise (1500 UTC, or 0800 LT). The importance of surface absorbed solar radiation and its effect on evapotranspiration becomes most apparent for the irrigated agriculture-to-urban land conversion, as a dew point temperature decrease of the order of 1°C (1.5°C) is noted from sunrise to midafternoon for the wet years (dry years). The lone theme that contributes to low-level moistening, as opposed to drying, occurs because of the conversion of evergreen forest to shrubland. As noted previously, the difference in albedo between the two surfaces lowers surface absorbed radiation for shrubland, serving to decrease low-level temperatures. Consequently, the amount of energy required to sustain shrubland evaporation at the same level as over needleleaf forest is increased. In absence of this energy enhancement the shrubland surface is simulated to remain comparatively cooler and moister (also see section 3.2). Because of the albedo effect already noted, peak moistening (which is, once again, enhanced for the dry years) for this theme coincides with the time when the sun is at its highest point in the sky, and the corresponding effect on evapotranspiration is potentially greatest.

[49] The diurnally averaged net surface shortwave radiation flux difference for all four land cover conversion themes is presented in Figure 10, for the wet (Figure 10a) and dry (Figure 10b) years. Differences in incident shortwave radiation, which occur only during daytime hours, are either positive or negative, depending on the particular land surface conversion of interest. Both modes of urbanization (conversion from either irrigated agriculture or shrubland to urban land) have a positive impact on this budget component, though the shrub-to-urban conversion results in a radiative forcing more than twice the magnitude of the agriculture-to-urban conversion. As expected and discussed previously, the conversion of evergreen forest to shrubland occurring in the northern tier of the domain results in a negative effect on the diurnally averaged surface absorbed shortwave radiation. However, the loss of irrigated agriculture to shrubland also results in a decrease of absorbed shortwave radiation at the surface. For these reasons, the effect on diurnally averaged surface shortwave radiation flux follows the change in albedo for the particular land surface conversions involved, though the relation shows an inverse proportionality (e.g., the shrub to urban landscape conversion results in a lowering of albedo and hence, produces an increase in this radiation budget component). During dry years, though the effect is more pronounced for all conversion themes, the magnitude of increase is larger for landscape conversions that increase fractional coverage of shrubland, resulting in an overall decrease in incident shortwave radiation across the domain (see also Figure 7b).

image

Figure 10. Same as Figure 9 but for net surface shortwave radiative flux (W m-2) (NLCD01 minus NLCD73) (a) for all three wet years and (b) for all three dry years; same as Figure 10a but for net longwave radiative flux (W m-2) (NLCD01 minus NLCD73) (c) for all three wet years, and (d) for all three dry years.

Download figure to PowerPoint

[50] The diurnally averaged net surface longwave radiation flux difference curves are presented in Figure 10, for the wet (Figure 10c) and dry (Figure 10d) years. The effect of low-level water vapor loss due to decreasing fractional coverage of irrigated agriculture is evident, as the two conversion themes contributing most to the surface longwave loss are the irrigated agriculture-to-shrubland and irrigated agriculture-to-urban land shifts.

[51] The total radiative forcing resulting from the net contributions of shortwave and longwave absorption display a slight decrease for the wet years (not shown) and a more pronounced decline during the dry (not shown) years. The lone landscape theme contributing in a positive manner to the total radiative forcing is the shrub-to-urban conversion with maximum increases of nearly 60 W m−2 evident during the early afternoon.

[52] Finally, shifts in land use alter the diurnal cycle of the partitioning between sensible and latent heating. Figures 11a and 11b show that the overall change in sensible heat flux is generally consistent with the warming and cooling noted previously. All surface conversion themes contribute to an increase in sensible heating, except for the evergreen-forest-to-shrub conversion. The removal of vegetation (i.e., irrigated agriculture) at the expense of increased coverage of urban land, shifts surface energy partitioning to increased sensible and decreased latent heat (see also Figures 11c and 11d), a result that is in agreement with previous numerical modeling studies [e.g., Arnfield, 2003; Georgescu et al., 2008]. The agriculture-to-urban land conversion shows the greatest effect on the surface flux energy partitioning with increases (decreases) in turbulent sensible (latent) heating of about 100 W m−2 between the wet and dry years. The forest-to-shrub conversion is the lone conversion theme that increases turbulent latent heating, though it only partially offsets the negative contributions from the remaining three themes.

image

Figure 11. As Figure 9, but for surface sensible heat flux (W m−2) (NLCD01 minus NLCD73) (a) for all three wet years and (b) for all three dry years; same as Figure 11a but for surface latent heat flux (W m−2) (NLCD01 minus NLCD73) (c) for all three wet years and (d) for all three dry years.

Download figure to PowerPoint

4. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[53] This paper, an extension of the work presented by Georgescu et al. [2008], is the first of a two-part series investigating the summer climate impact of actual landscape evolution, since the early 1970s, over one of the fastest urbanizing metropolitan complexes in the United States, the Greater Phoenix region. We use high-resolution RAMS simulations in conjunction with consistently defined, harmonized circa 1973, circa 1992, and circa 2001 LULC data sets, derived with USGS land cover data for 1973 (source is aerial photography), 1992 (Landsat TM), and 2001 (Landsat ETM+). We also updated the LEAF-2 biophysical parameters for the associated land cover classes for use with LEAF-2. Furthermore, by experimental design, we purposely simulate both dry and wet NAMS seasons, to distinguish whether a particular landscape-induced signal in the meteorology operates during dissimilar large-scale settings (i.e., we simulate periods both when the large-scale signal is weaker, and when it is stronger). The intention is to ensure that the clear identification of the local/mesoscale landscape signal is distinguishable from the forcing of the large scale.

[54] In this paper we first quantify the region's extensive landscape modification, beginning in the early 1970s through 2001. Inspection of the NLCD73, NLCD92, and NLCD01 land cover data sets highlights the broad landscape change that occurred over the region during the last three decades. Urban sprawl occurred nearly uninterrupted to the northwest and southeast of central Phoenix, replacing plots of irrigated agriculture, as territorial expansion proceeded from the 1970s through the 1990s and finally to the early 2000s. Gains in urban land, however, also occurred at the expense of seminatural shrubland, just north and east of the central business district of Phoenix. Last, north of about 34° N latitude (i.e., the north of Greater Phoenix) large increases in seminatural shrubland occurred over time, due to reduction in evergreen forest. While a quantitative inventory detailing forest decrease at the expense of increased territorial coverage of shrubland (and to a lesser extent, urban land) during the period of interest is not readily available, ancillary evidence does help to provide additional physical, qualitative, support for the landscape change provided in our data sets (see also section 3.1).

[55] The broad changes in the region's landscape were then investigated through modeled changes in local/regional albedo and its effect on the modulation of the surface radiation balance. In addition, the effect of distinct land use conversion themes (e.g., conversion from irrigated agriculture to urban land) were examined to evaluate how distinct themes have each contributed to the region's changing climate. Diurnally averaged land use conversion themes illustrate the important impact of landscape change on the land surface radiation and energy budget components and the subsequent impact on near-surface climate. For example, the two urbanization themes studied in this paper (i.e., the conversion from shrubland to urban land and the conversion of irrigated agriculture to urban land) both act to increase surface shortwave absorption, while the pair of themes leading to greater coverage of shrubland (i.e., the conversion of irrigated agriculture to shrubland and the conversion of evergreen forest to shrubland) both lead to a decrease in surface absorbed shortwave radiation. The effect of landscape change on the longwave radiation budget component displays a nearly across-the-board decrease, though there is variability among the different conversion themes.

[56] A key point is that while the net radiative flux difference between the 1973 and 2001 landscapes is small (i.e., for the wet years) or negative (i.e., for the dry years), it is the repartitioning of surface-absorbed energy that is critical for any modification of the near-surface climate. Both urbanization themes and the decrease in irrigated agriculture at the expense of increased shrubland, contribute positively to the diurnally averaged sensible heat flux. Overall, sensible heat flux differences between NLCD01 and NLCD73 result in a 1 W m−2 (this result is similar for both the wet and dry experiments, due to compensating increases and decreases in the enhanced values seen for the dry as opposed to the wet years) increase in domain-wide sensible heating. Coupled with a 2 W m−2 (for the wet years) to 3.6 W m−2 (for the dry years) decrease in latent heating (primarily driven by the loss of irrigated agriculture) between NLCD01 and NLCD73, these results highlight the importance of surface repartitioning in establishing near-surface temperature trends. For example, both sets of experiments illustrate a regional warming of 0.12°C during the roughly three decade period of landscape evolution, in qualitative agreement with recent estimates of urban heat island tendency over the Phoenix area [Stone, 2007].

[57] Finally, although regional absolute temperature has increased, equivalent temperature, which takes into account the variability in atmospheric moisture in addition to the contribution of dry static energy [Pielke et al., 2004] decreased (not shown) by roughly three times the magnitude of the temperature increase, owing to decreasing coverage of irrigated agriculture and its effect on near surface moisture.

[58] As discussed in section 3.3, the effects of urbanization presented in this study (on temperatures and atmospheric processes) likely represent underestimates. A follow-up study analyzing the effect of a more comprehensive representation of the urban landscape would be a beneficial additional contribution [e.g., Freitas et al., 2006].

[59] The partitioning of surface absorbed energy into sensible and latent heating has been shown to be a significant driver of atmospheric circulations and convective activity in many regions of the world and on a variety of scales [Pielke, 2001]. Results presented in part 1 of this overall study document the significant effects of landscape change on key components of the surface radiation and energy budgets and their effect on near-surface climate. In part 2, Georgescu et al. [2009] address the role of the previously discussed surface budget changes, and ensuing thermal gradient, on the mesoscale dynamics and thermodynamics, convective rainfall and the association with the large-scale NAMS storm systems.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[60] This work was funded by NASA through Earth System Science Fellowship Grant NNG04GQ47H. C.P.W. states that the views expressed in this paper are his own and those of the other authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. The authors also wish to thank two anonymous reviewers for their extensive comments.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
jgrd15088-sup-0001-t01.txtplain text document1KTab-delimited Table 1.
jgrd15088-sup-0002-t02.txtplain text document0KTab-delimited Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.