1. Introduction
Over the last century, the peninsula of Florida has significant landscape transformation due to drainage and channel development to accommodate agriculture and population growth (Donders et al. 2008; Florida Department of Environmental Protection 2007; Marshall et al. 2004). Cody (2006) illustrated how population centroids migrated and the modern relocation of human settlements across west-central Florida. The human-induced transformation was confirmed in a recent analysis by Kautz et al. (2007), who showed a significant land use change (LUC) of several nonurban categories into urban or developed. The authors report urban conversions of 14.02% for agriculture and pasture areas, 36.27% for shrubland, 25.44% for dry prairie, and 11.32% for upland forests. Recent studies in Florida include not only the geographical description of land use and land cover change, but the impact of such transformations on climate conditions. Modeling the impact of landscape conversion in Florida, by comparing pre-1900 and 1993 conditions, resulted in modifications of latent and sensible heat fluxes, and sea-breeze thermally induced circulation, resulting in the altering of spatial distributions of convective rainfall (Marshall et al. 2004; Pielke et al. 1999). Additionally, Pielke et al. (2007), using observational and modeling experiments, elucidated the impacts of such transformations on precipitation patterns at different scales and recommended further research to better understand their interconnection. These investigations acknowledged the importance of considering anthropogenic land cover change when studying climate trends.
This study consisted of two complementary parts addressing key questions about the factors driving landscape conversion in central Florida during recent decades and its hypothetical effect on climatic conditions. The first part (section 2) investigates observed LUCs in and around the Southwest Florida Water Management District (SWFWMD) during 1995–2006, when this region underwent a substantial population increase that led to an increased demand for urban developments, and subsequent reduced agriculture and rangeland land uses as well as loss of upland forests. This part also explored the main socioeconomic and biophysical factors related to this landscape conversion. The second part (section 3) uses a case study approach to explore climate simulations for the coast to coast central Florida region using actual 1992–93 spatial land use data (referred to as ALU simulations). These results are compared with outputs from an ideal extreme land use scenario (ELU simulation) in which the dryland crop pasture classes were converted to an urban class. Model results were evaluated using surface and vertical atmospheric observations. The ELU simulation resulted in a land use distribution resembling current patterns over the SWFWMD and leads to discussions regarding the possible impacts of the extreme urbanization. We pay particular attention to the effects of ELU simulation conversion in highly urbanized coastal and inland areas, as well as rural and mixed natural–urban environments. Finally, we conclude with a discussion on the use of this approach to analyze the effects of LUC and urbanization on climate and of the implications for key water management–related parameters and other ecosystem services as defined by Escobedo et al. (2011).
2. Land use and land cover change
a. Data and methods
To accomplish the analysis of land use and land cover we focused on the SWFWMD region, depicted in Fig. 1b, where a substantial ongoing LUC has been observed. We employed imagery used by the SWFWMD state agency to support its regional water management, regulatory, and planning activities. These land use data are categorized according to the Florida Department of Transportation’s (1999) classification system. To better discuss and integrate our landscape transformations and climate modeling sections, we standardized the land use classifications used by the SWFWMD and those employed in regional climate models. The original SWFWMD information was reclassified into 10 categories, indicated in Table 1, to generate the maps of land use as shown in Fig. 1b for 2006. This information consisted of 15.2–30.48-cm-resolution digital photographs taken in 1995, 1999, 2004, and 2006 and photo interpreted at a scale of 1:12 000. Water bodies have significant uncertainties because of the effect of tides. For this classification, water reservoirs were assessed separately and water use distribution was used to help explain changes in this land use category. We did not limit our study to the administrative boundary of the SWFWMD, but chose the wider coast to coast central Florida region in the modeling domain to better analyze comprehensive ocean–atmosphere–terrestrial interactions of major influence on peninsular climate conditions.
(a) The 1995–2006 evolution of LUC showed a dominant increase of 128 000 ha (26.6%) in urban areas, and 23 000 ha (21.8%) in extractive lands (e.g., mines). According to the SWFWMD water use reports, water reservoirs increased 6000 ha (19.9%) to fulfill the public and industrial supply. (b) A 2006 LU map shows more extensive urban areas than there were in 1992–93 (Fig. 2a). The observed increment in development supports the proposed ELU simulation.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
The 53 SWFWMD LU categories reclassified into 10 classes to represent existing 1995–2006 LUC in west-central Florida.
b. Socioeconomic factors behind LUC
According to the Bureau of Economic and Business Research (2008), the total population in major metropolitan statistical areas within the SWFWMD increased by 17.1% and 15.8% during 1990–2000 and 2000–07, respectively. Metropolitan areas in the SWFWMD include Tampa–St. Petersburg–Clear Water (Hillsborough, Pinellas, Pasco, and Hernando Counties), Sarasota–Bradenton–Venice (Sarasota and Manatee Counties), and Lakeland (Polk County), which grew from 2.96 million people in 1990 to 4.02 million people in 2007. In response to rapid population growth, the area of urban and developed land uses increased while agricultural, rangeland, and upland forest areas decreased, as shown in Fig. 1a. Socioeconomic factors such as real estate prices of undeveloped areas (i.e., forests, rangelands, or other undeveloped ecosystems) during analysis years and the “boom” in population and the housing markets resulted in a subsequent need for increased land development and related infrastructures prior to 2008 (Kautz et al. 2007; Montes Rojas et al. 2007). Additionally, the decreased value of the agricultural and forestry sector products relative to real estate values and subsequent conversion of forest to developed areas are often associated with increased urbanization during the analysis period (Long 2005, here after FDF05). Wear and Greis (2002) have identified these socioeconomic factors as drivers of land use change in other areas in the southeastern United States. The effects of the conversion of natural to urban areas on ecosystem services, including effects on climate, are discussed in Escobedo et al. (2011).
The SWFWMD’s estimated water use reports from 1998–2002, 2003–2004, 2006, and 2008 (available online at http://www.swfwmd.state.fl.us/documents/#reports) indicate that freshwater was mainly allocated to the public (domestic and institutional), agricultural, industrial, mining, and recreational sectors. These reports corroborate a higher water demand for public (50% and 47.8%) than for agricultural usage (35.6% and 34.4%) during 2006 and 2008, respectively, when compared to an average of 40% for both sectors during 1998–2002. An expansion of mining activities in central Florida during the early part of the analysis period, around 2002, also increased the area of extractive LUs (i.e., mines) since Florida’s phosphate industry supplies about 75% of the United States’ and 25% of the world’s phosphate supply (Brown 2005). Multiyear restoration projects (e.g., conversion of mining areas to pre-disturbance conditions) have been ongoing since 1975 and resulted in 63% of the land being, or in the process of being, reclaimed (Brown 2005). Nevertheless, mining continues to convert from 2023 to 2428 ha of land annually (Florida Department of Environmental Protection 2003). Figure 1a illustrates the increase in mining lands during 1995–2006 at a rate of approximately 23 340 ha or 2120 ha yr−1. Recreational lands (e.g., parks and golf courses) increased to 3420 ha (13.5% of the area) in 1995 and used 82 million gallons of water per day in 2006, as evidenced in the above SWFWMD water use reports. These reports also confirmed that in 2006, recreation activities consumed 30% more water than that used in 1998 (61 mg day−1).
In 2006, counties in the SWFWMD also supported 61.9% of the U.S. citrus production and accounted for 48.2% of Florida’s citrus acreage (Florida Agricultural Statistics Service 2007). In 1996, the citrus-growing area reached 347 000 ha, but declined by 26.7% through 2008 (Florida Department of Citrus 2009). This decline of citrus-growing land, corresponding to 92 650 ha, is consistent with the 93 000-ha decrease in agricultural areas in this study. Similar trends in LUC and the factors behind them have also been observed elsewhere in the southeastern United States during the analysis period (Wear and Greis 2002). Figure 1a shows the change from agricultural, upland forest, and rangeland classes to urban class, which when combined account for losses of 93 000, 48 000, and 44 000 ha in the land uses, respectively.
c. Conservation policies and other biophysical factors behind LUC
Forest land use/cover declined in recent decades; however, consistent comparisons across years are difficult because of dataset unavailability and different definitions and procedures used by land management and environmental agencies [e.g., land cover definition differences between the SWFWMD and the U.S. Geological Survey (USGS)]. During our analysis period of 1995–2006, the area of upland forest decreased by 48 147 ha. A decline of similar magnitude was also found by the USDA (2009) using county-level data. They reported that the area of land classified as woodlands (e.g., natural and planted woodlots) decreased by 62 887 ha within the SWFWMD counties between 2002 and 2007. Land development regulations, conservation objectives, and smart growth policies have also been cited as factors that drive LUC and increase urban development (Wear and Greis 2002). Forest land ownership in 1995 was dominated by nonindustrial private landowners (49%) and the forest industry (32%), while public forest land (i.e., national forests, state parks, etc.) accounts for 19% (FDF05). Changes in land use did occur during the last decade because of the increase in land values as previously mentioned, but conservation policies helped conserve some forested areas (Hodges et al. 2005).
Finally, natural disturbances such as hurricanes, climate extremes (freezes and droughts), and forest- and agriculture-related pest–disease outbreaks have also detrimentally affected the region’s agricultural sector and led to LUC (FDF05; Spreen et al. 2006). Florida’s forests and agricultural areas have also been impacted by hurricanes that have resulted in measurable damage to the forest lands and various agricultural sectors. For example, Hurricanes Charlie (August 2004), Frances (September 2004), Jeanne (September 2004), and Wilma (October 2005) contributed to the spread of canker and greening diseases, impacting citrus areas in central Florida (Spreen et al. 2006). Along with diseases, FDF05 reported that wildfires associated with droughts have affected forest, rangeland, and agriculture land conversion in west-central Florida. Marion, Levy, and Hernando counties in the SWFWMD were impacted in 1997–2000 by pine beetle infestations that resulted in the loss of 3237 ha of pineland, and from 1995 to 2003, wildfires burned 769 000 ha, affecting forests, wetlands, rangelands, and other undeveloped areas (FDF05).
3. Regional modeling simulations
We explored the impacts of these LUCs on mesoscale climate using modeling experiments and a simplified ideal urbanization scenario. The baseline model land uses were USGS’s 24 land use categories (Brown et al. 1993). Figure 2a displays the 1992–93 USGS model land use map, or ALU simulation, while Fig. 2b shows the extreme land use distribution, or ELU simulation, obtained by applying an ideal conversion where dryland crop/pasture was converted into the urban class. We focused our analysis of urbanization effects on the surface and low-atmosphere conditions through regional numerical experiments during a July 1993 modeling period. This year, concurrent with the model actual land use distribution, is used as a representative baseline to provide a comparison with the proposed extreme LUC scenario, which resulted in a land use distribution similar to contemporary patterns. Interactions between synoptic, mesoscale, and local processes are typical in Florida throughout the year; however, the summer months show higher ocean–land thermodynamic gradients increasing sea-breeze circulation and the occurrence of convection with storms on a daily cycle (Pielke 1974). July was also selected since regional–local processes are more influential in air circulation and we aim to see the effect of local land use alterations on regional climate conditions.
(a) USGS LU map at 1-km2 resolution used in the regional climate simulations. The region contained 9 of the default 24 categories. (b) The ELU simulation where dryland crops and pasture areas were converted into urban class. We use the latitude band (rectangle) to analyze temporal and spatial variability in Figs. 4, 5. (c) The model was configured at these two geographical domains at 3-km (magenta box) and 1-km (green box) resolutions. We use the 1-km output to evaluate the model at the indicated COAPS stations and to perform the sensitivity analysis to changes in the surface and atmospheric conditions after the ELU conversion scenario is applied.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
a. Modeling system, data, and methods
The fifth-generation Pennsylvania State University–National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5) (Grell et al. 1995) was configured for application in central Florida using the following set of physics schemes: the radiation scheme was applied to the rapid radiation transfer model, designed to treat several atmospheric gases and their absorption coefficients for longwave spectral range (Mlawer et al. 1997). The cumulus parameterization was set to the Grell option, where the vertical integration of moisture convergence defines the amount of convection (Grell et al. 1995). The explicit moisture was fixed to a simple ice scheme that is active whenever grid-scale saturation is reached, taking supersaturation as precipitation (Grell et al. 1995). This scheme adds an ice phase in the precipitation processes. The planetary boundary layer was set to the High-Resolution Blackadar choice that uses similarity theory to compute surface heat, moisture fluxes, and vertical mixing velocities and ratios (Grell et al. 1995). Under this planetary boundary layer option, we set the IZOTOPT parameter to 2, so diffusion was allowed for cloudy air and the thermal roughness length was treated different than the roughness for momentum. The land surface scheme consisted of the five-layer (1, 2, 4, 8, and 16 cm) soil model (Grell et al. 1995), where the vertical heat transfer is controlled by the soil thermal inertia of each land use category. This soil scheme calculates temperature tendencies by taking the residual of the surface energy budget (including sensible, latent, and radiative heat fluxes) and keeps the substrate at a constant temperature. Soil moisture was allowed to vary by setting the IMOIAV parameter to 1 if precipitation events occurred during the running time. The model was then run in a two-way nesting option and configured to 23 atmospheric levels ranging from 1001 to 580 mb. Below 880 mb, the atmospheric levels were finer to better explain the dynamics of the lower atmosphere.
Nested modeling domains at 3- and 1-km2 resolutions were configured in MM5 for the region as indicated in Fig. 2c. Boundary conditions necessary to run MM5 were taken from the National Centers for Environmental Prediction (NCEP)–NCAR reanalysis (Kalnay et al. 1996), which originally had a 2.5° × 2.5° spatial resolution with a 6-h interval. The model integration was achieved for the period 2–23 July 1993 with a 3-h interval. We examined the effects on climate conditions by taking the difference between the ALU and ELU simulations using outputs from the 1-km2 modeling domain. Model results were evaluated against meteorological and vertical atmospheric observations. This work uses hourly datasets from five meteorological stations (722110, 722116, 747880, 722040, and 722045), depicted in Fig. 2c, from the Florida State University Center for Ocean-Atmospheric Prediction Studies (COAPS; see online at http://www.coaps.fsu.edu/). These meteorological stations were used to compute the eastern and western mean time series to support the model evaluation and analysis. The model time series were estimated at each station location and then a regional average was obtained among both western stations and eastern stations separately. Vertical atmospheric observations were taken from the Integrated Global Radiosonde Archive (IGRA), version 1.20 (Durre et al. 2006), which consists of an extensive record of radiosonde and balloon measurements since 1970 in a set of globally distributed stations. Morning and evening profiles at the Tampa station (latitude: 27.70°N, longitude: 82.38°W) during July 1993 were used in the model evaluation.
b. Model evaluation
Figure 3 presents a comparison of simulated (ALU) and observed time series at the western (lhs) and eastern (rhs) regions. Table 2 provides the mean, standard deviation, correlation coefficient (CC), root-mean-square error (RMSE), and bias for each variable and region. To facilitate the later discussion on the effects of the extreme urbanization scenario on climate, Table 2 also shows descriptive statistics for the ELU simulation. The predicted 2-m temperature presents high correlations to measurements in both regions (0.85 and 0.93). Simulated means (29.54° and 29.14°C) were comparable to observed means (28.05° and 27.94°C), and model standard deviations (2.99° and 3.02°C) agreed with the observed variability (2.82° and 3.29°C).
Comparison of ALU simulation and observed time series for 2-m air temperature (first row), 10-m wind speed (second row), 10-m wind direction (third row), and daily accumulated precipitation (fourth row). The mean time series at the (left) western and (right) eastern region. Model temperature and wind speed display satisfactory correlations with observations, as confirmed in Table 2. Wind direction displays typical southeast and southwest directions as observed at this time of the year. Sporadically, the model precipitation tends to follow the observations, and western amounts were larger than those in the eastern region, in agreement with COAPS data.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
Model evaluation and statistics. The 1-km-resolution model outputs for ALU and ELU simulations are evaluated against COAPS observations. ELU simulation outputs are in parentheses. DT2m, wind speed, wind direction, and precipitation data are averaged in the western and eastern regions. The statistics consists of model mean, model standard deviation, observation mean, observation standard deviation, CC, RMSE, and bias.
In general, model temperature was warmer than observations according to a positive bias in Table 2 and visually in Fig. 3. Wind speed (second row in Fig. 3) shows a daily signal confirming the thermal circulation and sea-breeze effects. The model and observed wind speeds are similar in regard to averages and standard deviations, with acceptable CCs and bias close to zero in both regions. The model predicts larger variability for wind speed on the east coast of Florida, confirmed by the observations. The RMSE over the Gulf of Mexico (1.03 m s−1) shows a lower absolute deviation than the corresponding one on the Atlantic coast (1.41 m s−1). The stations on the Atlantic coast (straight shoreline) are more exposed to an open large oceanic basin, likely enhancing a coherent daily thermal circulation signal shown in Fig. 3. Similarly, a less organized daily variation of wind speed is observed in stations in the Tampa Bay area (irregular shoreline) on the west coast. Simulated wind direction follows the observed pattern on the eastern coast more closely (CC = 0.63), with prevailing winds oscillating between southeast and southwest. A daily oscillating wind speed signal over the region was evident, particularly before 16 July. In the western region, the model and observations showed a synoptic influence after 16 July, when southwest winds prevail. In the Florida peninsula, the typical summer atmospheric circulation includes light southwest and southeast flows depending on synoptic weather circumstances (Boybeyi and Raman 1992), and a similar pattern was found in this study.
Precipitation was a complex parameter to predict both spatially and temporarily, particularly at the finer scales of 1 km as used in this study. Therefore, we computed daily accumulated precipitation in the model and observations to achieve the corresponding analysis. The precipitation daily time series (fourth row of Fig. 3 and last two lines in Table 2) displayed the lowest performance in this study, as low CCs were found for precipitation. The model was drier than observations over both coasts; however, the model was consistent in that the western region was wetter with frequent rains (higher variability) than the eastern region as confirmed by station data.
Figure 4 presents the mean ALU simulation (blue) and observed (red) atmospheric profiles (1000–800 mb) from radiosonde data at the IGRA Tampa station. Figure 4 presents temperature (lhs), wind direction (middle), and wind speed (rhs) vertical distributions and bars represent the standard deviation. The top panels show profiles at 0700 EST and bottom panels at 1900 EST when the radiosondes were released in July 1993. Each panel indicates the CC, mean difference (observed − model), and RMSE. There is a high correlation between model and observed temperatures as evidenced by morning profiles being closer to observations according to their smaller bias and RMSE. Wind direction (middle panels) shows acceptable correlations; however, model variability below 900 mb is significantly lower than observed in the evening. Above 900 mb, the model follows radiosonde wind direction more closely. Observed and model morning profiles skewed to the south-southwest direction likely follow synoptic circulation with a weak sea-breeze influence. Nevertheless, evening profiles tend to be southwest at the surface and southeast above in response to enhanced sea–land temperature gradients influencing wind direction. Wind speed (Fig. 4, rhs) in the evening shows a very good correlation along with lower bias and RMSE. Morning predictions of wind speed deviate from observations; however, above 900 mb the model presents a better agreement. It is evident in Fig. 4 (rhs) that the enhanced land–sea thermal gradient increases both observed and model surface wind speeds during the afternoon.
ALU simulation output (blue) and IGRA radiosonde (red) mean atmospheric profiles below 800 mb and bars indicating standard deviation. Vertical distributions at (top) 7 a.m. and (bottom) 7 p.m. EST. (left) Temperatures present a high correlation, with lower bias (mean diff) and RMSE. (middle) Wind direction shows acceptable correlations; however, model variability below 900 mb is significantly lower than observed. (right) Wind speed in the evening shows very good correlation along with lower bias and RMSE; nevertheless, above 900 mb the model agrees better with observations.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
We analyzed central Florida from the Gulf of Mexico to the Atlantic coasts to comprehensively describe the modeling region (Fig. 2). According to the 1992–93 USGS land use classifications, the urban areas (3.12% of the study region) were concentrated mostly in the west and smaller patches occurred on the east coast. The agricultural-related classes (53% of the study region including dryland crop and pasture, cropland and grassland mosaic, and cropland wooded mosaic areas) and the evergreen needle leaf forests (5.87%) occupied significant portions during 1993. The ELU simulation (Fig. 2b) resembled land use in 2006 (Fig. 1b) in a swath delimited by 27.7°–28.0°N latitude in the western sections of the study area. We analyzed this latitudinal swath to better explore the effects of ELU simulation transformation on the six geographical zones depicted in Fig. 2b: west coast (WC; longitude < −82.85°), high urban (HU1; from −82.85° to −82.30° longitude), rural (RU; from −82.30° to −82° longitude), high urban (HU2; from −82° to −81.50° longitude), mixed areas (MA; counting crop wood, large natural extensions, and urban patches, from −81.50° to −80.50° longitude), and east coast (EC; longitude > −80.50°).
c. Analysis: Impact of the extreme LUC scenario
Previous results were used to discuss the effects of the ELU simulation on regional climate conditions. First, we use the amounts in parentheses in Table 2 to compare results from the ALU and ELU simulations. The ELU simulation temperature statistics consistently showed slightly smaller mean and standard deviations along with increased precipitation in both regions. The temperature decrease was likely a result of the cooling effect of clouds and precipitation. It seems likely that extreme urbanization tends to inhibit air circulation across the whole region since ELU simulation consistently presents smaller wind speed averages and standard deviations than the ALU simulation modeling results. No major changes were visible regarding wind direction, since comparable averages and standard deviations were evident in the statistics for both regions. In summary, the ideal extreme urbanization scenario in the coastal areas resulted in an inhibited atmospheric circulation with somewhat enhanced precipitation.
As shown in Fig. 5a, the largest 2-m air temperature (DT2m > 0.6°C) was typically in the inland HU2 zone and neighboring rural areas, followed by the coastal HU1 zone. Both the WC and EC zones showed DT2ms closer to zero, and because of this, we focus the discussion on the terrestrial zones in the next analysis. Positive or negative patches of DT2m were just as likely to occur in inland zones; however, they tended to occur in the afternoon–evening periods when low relative humidity (RH) and higher wind speeds prevailed. The COAPS measurements clearly corroborate that wind speed showed a regional daily variability in response to the sea-breeze thermodynamic circulation (Pielke 1974), which simultaneously influences RH. The two largest negative and positive patches of DT2m happened before 16 July and are located over the western inland section with intensification at the highly urbanized HU2 in the extreme scenario. Although it has a low performance in precipitation, the model follows the main temporal and spatial observed patterns. It was clear when discussing Fig. 3 that over the west region wind alternates between southwest and southeast, wind speed is slower, and rain is more frequent. During 3–4 July, over the DT2m negative patch, wind from the southwest brought typically warmer air (from the Gulf of Mexico) into the region. Under these circumstances a sea-breeze front was likely to develop with no precipitation, as the model predicted over part of the region. During 8–9 July, DT2m was positive, simulated wind mainly blew from the south and southeast, and the model predicted 6 mm of precipitation. From 16 to 22 July, the wind speed intensified to a mean of 3.16 m s−1 and the DT2m patches became smaller. These results suggested that the effect of our ELU simulation on summer climate was complex and depended on the prevailing interactions among local (e.g., sea breeze), convective and advective precipitation synoptic systems (e.g., the well-known Bermuda high over the Atlantic, typical low pressure over the continent, and the intertropical convergence zone moving northward) and indeed the land use characteristics in each zone.
Evolution of the difference between the ALU and ELU simulations for DT2m. (a) A latitude band (Fig. 2b) mean was applied to obtain the DT2m longitude–time (Hovmöller) diagram. Days in the vertical axis are placed at 00 h EST. (b)–(d) The regional mean of wind speed, wind direction, and RH, respectively, from COAPS observations.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
Figure 6 shows the difference between the ELU and ALU simulations in surface temperature, wind speed, wind direction, and precipitation modeling time series across the representative land use zones indicated in Fig. 2b. Table 3 summarizes the statistics, median, and RMSE of the difference for each variable and zone. We use median because of the skewness of statistical distributions and RMSE as a measure of variability. Figure 6 and Table 3 support our previous discussion on the effect of the extreme urbanization scenario. In general, the magnitude of the impact depends on land use at each zone, the zone’s location with respect to the coasts, and the prevailing atmospheric circulation. A daily oscillatory pattern in the temperature and wind speed differences is clear in Fig. 6 in response to the associated evolution of the ocean–land thermal gradient. According to Table 3, in general, medians are close to zero for all variables; however, Fig. 6 clearly confirms a daily cycle for temperature and wind speed. The difference presented higher variability at the inland HU2 and RU zones. Because of the proximity to oceanic basins, the differences are less affected in HU1 and MA as confirmed by the corresponding smaller medians and lower variability in Fig. 6 and Table 3. At the beginning of the study period (before 16 July) when local processes control surface wind direction, such differences are more pronounced for temperature, wind speed, and precipitation. After 16 July, synoptic processes are more influential, observed wind direction prevails from the southwest, and wind speed intensifies as found in previous analyses. During this part of the study period, the variability of ELU–ALU simulation differences (temperature, wind direction, and precipitation) decreases, however HU2 and RU, temperature and wind speed show larger RMSEs.
To analyze the impact of the extreme urbanization scenario on surface climatological conditions, we consider the time series of temperature, wind speed, wind direction, and precipitation in the characteristic zones indicated in Fig. 2b, and here show the differences in those time series between the ALU and ELU simulations. Table 3 summarizes the statistics of the differences. The magnitude of the impact depends on the underlying LU at the zones, the location of the zones with respect to the coast, and the dominance between local and synoptic atmospheric circulations. The RMSE of the differences for all climate variables is higher in the inland HU2 and RU zones.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
Statistics of the differences in time series for surface temperature, wind speed, wind direction, and precipitation between ALU and ELU simulations.
The next part of the analysis is based on a simulated atmospheric cross section during two contrasting conditions of the atmospheric boundary layer. One is before sunrise (0400 EST) when a stable atmosphere is expected and the second before sunset (1600 EST) when a convective mixed layer is established. During nighttime, the modeled lower atmosphere was stable (humidity and temperature decreased with height) and presented inward patches of moisture from the Gulf of Mexico (Fig. 7). This was a characteristic summer nocturnal pattern of higher surface RH and low air circulation, which seemed to dominate in both the ALU and ELU simulations. Because the inland portion of the peninsula (longitudes between −82.85° and −80.45°) had larger sensible heat flux than the lateral coastal portions, the temperature distribution at 4 p.m. exhibited a discernable dome shape at lower levels with an updraft moisture flow confirmed by the RH contours. The dome was somewhat extended at the base and skewed toward the eastern side of the peninsula compared to the afternoon cross section of the ALU experiment. The RHs of Fig. 7 shows lower surface RH in the afternoon, in agreement with the COAPS observations. The ALU and ELU simulations presented some differences, most noticeably in the evening when wind and urban heat island (UHI) interactions affected both cross sections. The 29°C contour of temperature in the ALU simulation was elongated toward the west in the ELU simulation in response to more urbanized areas in the west. In the ELU simulation, above the 29°C line, the atmosphere was drier because of a vertical moisture flow that lifted the 54% RH contour. Size and location with respect to coastal areas and their effect on atmosphere thermodynamics of the UHI have been previously analyzed (Ohashi and Kida 2004).
Mean atmospheric vertical cross sections from the surface up to 920 mb (about 840-m height) of the latitude band shown in Fig. 2b. (left) The temperature (°C; shaded colors) and RH (red contour) sections at 4 a.m. EST during the period from 1 to 22 Jul 1993 and (right) analogous estimates at 4 p.m. for the (top) ALU and (bottom) ELU simulation results.
Citation: Weather, Climate, and Society 4, 3; 10.1175/WCAS-D-11-00019.1
Complex land use modeling and detailed descriptions of the physics and interactions in our modeling efforts are beyond the scope of this study. However, this study elucidates future areas of research and has some implications for the effects of urbanization on mesoscale climate. Peninsular Florida’s unique geographical settings with two distinct ocean basins adjacent to flat inland areas located in tropical and subtropical climates will likely influence UHI effects regionally and locally. Although the HU1 and HU2 urbanized areas were similar in size (about 54-km width), they were characterized by dissimilar land use classes, and their locations—with respect to ocean basins—were different. Land use categories in HU1 include a large portion of Tampa Bay, while those in HU2 are mostly urbanized land uses after the extreme transformation. In the ELU modeling scenario, HU1 and HU2 presented a reduction of surface RH with enhanced vertical flow of moisture at the center of the modeling region. The 54% contour of relative humidity was taller and wider in HU2, meaning a drier atmosphere above. Over the coastal HU1, the moisture transport inland is enhanced by the sea-breeze circulation. These features were more pronounced in the ELU simulation than in the ALU simulation. The extended urban areas in the ELU simulation made the 54% RH contours closer over the RU and a reduction of surface wind speed was significant on these urban and rural zones (as confirmed in Table 3). Within MA in the region close to HU2, a depression in the 54% RH contour is evident under ELU-simulation circumstances. The WC and EC environments displayed modest daily changes in surface temperature due to a larger heat capacity of the ocean compared to its terrestrial counterpart.
4. Summary, conclusions, and challenges
This study analyzed two aspects related to LUC in central Florida. First, we used available LU spatial data and the literature to discuss several socioeconomic and biophysical factors affecting LUC in the region and then based on a simplified agriculture-to-urban LUC scenario, we used a modeling approach and sensitivity analyses to explore the potential changes to the surface and low-atmosphere climate. We found that land use/cover changes involve a combination of both biophysical and socioeconomic factors across different spatial and temporal scales. Additionally, the process of urbanization is heterogeneous and not only affects land use patterns (i.e., ecosystem structure), but also specific climate parameters such as temperature and relative humidity (e.g., ecosystem function) that affect resource use and even human well-being (i.e., ecosystem services). Specifically, in Florida, UHI effects are different between coastal areas (WC versus EC) in the study and also in inland locations and will be most pronounced in and adjacent to land uses that become urbanized (HU2 and MA; Figs. 2b, 7).
This modeling analysis and our results have implications for land use planning and water management in urbanizing subtropical areas. While natural resources management, planning, and conservation activities are occurring in these areas, using modeling approaches such as ours to predict the effects associated with urbanization is important for better decision making, as these effects can concurrently act across both temporal and spatial domains. Future research is warranted in identifying specific drivers behind these land use changes and to model the effects of different land use policy scenarios (e.g., land conservation or smart growth development versus urban sprawl policies) on climate. Other research can use field data and remote sensing data to adjust default land use parameters used by MM5. Although it is challenging to analyze the complex relationships among these biophysical and socioeconomic factors and climate effects, as we have shown, a sensitivity analysis of simplified LUC scenarios using regional climate models can provide a practical method to predict the potential effects of LUC on climate, water needs, and UHI effects.
As such, this study’s modeling approach could be used to quantify the effects of urbanization and forest loss on key mesoscale climate-related parameters necessary to model evapotranspiration, urban temperatures, relative humidity, and other ecosystem services. Nonetheless, these modeling experiments need significant improvements and require further research to address the realities of modeling regions and to better understand interactions between natural ecosystems and the built environment. Indeed, the interaction between UHI, thermodynamic circulation, LUC, and the impacts on diverse natural ecosystems and human well-being deserves more detailed research. Beyond the geographical settings of subtropical Florida, this is a topic of fundamental interest pertaining to climate change and LUC dynamics.
Acknowledgments
This research was supported by the Southeast Climate Consortium the University of Florida Water Institute and the University of Florida Department of Agricultural and Biological Engineering. The authors acknowledge the University of Florida High-Performance Computing Center for providing computational resources and support, which have contributed to the research results reported within this paper.
REFERENCES
Boybeyi, Z., and Raman S. , 1992: A three-dimensional numerical sensitivity study of convection over the Florida Peninsula. Bound.-Layer Meteor., 60, 325–359.
Brown, J. F., Loveland T. R. , Merchant J. W. , Reed B. C. , and Ohlen D. O. , 1993: Using multisource data in global land characterization: Concepts, requirements, and methods. Photogramm. Eng. Remote Sens., 59, 977–987.
Brown, M. T., 2005: Landscape restoration following phosphate mining: 30 years of coevolution of science, industry and regulation. Ecol. Eng., 24, 309–329.
Bureau of Economic and Business Research, 2008: Florida Statistical Abstract 2008. University of Florida, 917 pp.
Cody, S. K., 2006: Florida’s population center migrates through history. Florida Focus, Vol. 2, University of Florida, Bureau of Economic and Business Research, 5 pp. [Available online at http://www.bebr.ufl.edu/sites/default/files/FloridaFocus2_1_2006_0_0.pdf.]
Donders, T. H., Gorissen P. M. , Sangiorgi F. , Cremer H. , Wagner-Cremer F. , and McGee V. , 2008: Three-hundred-year hydrological changes in a subtropical estuary, Rookery Bay (Florida): Human impact versus natural variability. Geochem., Geophys., Geosyst., 9, Q07V06, doi:10.1029/2008GC001980.
Durre, I., Vose R. S. , and Wuertz D. B. , 2006: Overview of the Integrated Global Radiosonde Archive. J. Climate, 19, 53–68.
Escobedo, F., Kroeger T. , and Wagner J. , 2011: Urban forests and pollution mitigation: Analyzing ecosystem services and disservices. Environ. Pollut., 159, 2078–2087.
Florida Agricultural Statistics Service, cited 2007: Citrus summary 2005-06. [Available online at http://www.nass.usda.gov/Statistics_by_State/Florida/Publications/Citrus/cs/2005–06/CS0506all.pdf.]
Florida Department of Citrus, cited 2009: Citrus reference book. [Available online at https://fdocgrower.box.com/shared/dt42f2c7kk#/s/dt42f2c7kk/1/76187836/676024980/1.]
Florida Department of Environmental Protection, cited 2003: Rate of reclamation report: July 1, 1975 through December 31, 2003. 9 pp. [Available online at http://publicfiles.dep.state.fl.us/dwrm/MineReclamation/reports/rate-of-reclamation/2003ror.pdf.]
Florida Department of Environmental Protection, cited 2007: Peace River Cumulative Impact Study (CIS) final report. [Available online at http://www.dep.state.fl.us/water/mines/pr_cis-final.htm.]
Florida Department of Transportation, 1999: Florida land use, cover and forms classification system. Surveying & Mapping Office Handbook, 3rd ed., 93 pp. [Available online at http://www.dot.state.fl.us/surveyingandmapping/Manuals/fluccmanual.pdf.]
Grell, G. A., Dudhia J. , and Stauffer D. R. , 1995: A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 122 pp.
Hodges, A. W., Mulkley W. D. , Alavalapati J. R. , Carter D. R. , and Kiker C. F. , 2005: Economic impacts of the forest industry in Florida, 2003. Final Rep. to the Florida Forestry Association, EDIS Doc. FE538, University of Florida, Institute of Food and Agriculture Sciences, 47 pp. [Available online at http://edis.ifas.ufl.edu/pdffiles/FE/FE53800.pdf.]
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kautz, R., Stys B. , and Kawula R. , 2007: Florida vegetation 2003 and land use change between 1985-89 and 2003. Fla. Sci., 70 (1), 12–23.
Long, M. C., 2005: Florida’s Forest Resources Plan: Setting the course for 2030. Florida Division of Forestry, 44 pp.
Marshall, C. H., Pielke R. A. Sr., Steyaert L. T. , and Willard D. A. , 2004: The impact of anthropogenic land-cover change on the Florida peninsula sea breezes and warm season sensible weather. Mon. Wea. Rev., 132, 28–52.
Mlawer, E. J., Taubman S. J. , Brown P. D. , Iacono M. J. , and Clough S. A. , 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 (D14), 16 663–16 682.
Montes Rojas, G., McGuire S. T. , Ivey S. , and Durrenberger T. , 2007: The Florida housing boom. Florida Focus, Vol. 3, University of Florida, Bureau of Economic and Business Research, 16 pp. [Available online at http://www.bebr.ufl.edu/sites/default/files/FloridaFocus3_2_2007_0.pdf.]
Ohashi, Y., and Kida H. , 2004: Local circulations developed in the vicinity of both coastal and inland urban areas. Part II: Effects of urban and mountain areas on moisture transport. J. Appl. Meteor., 43, 119–133.
Pielke, R. A., Sr., 1974: A three-dimensional numerical model of sea breezes over south Florida. Mon. Wea. Rev., 102, 115–139.
Pielke, R. A., Sr., Walko R. L. , Steyaert L. T. , Vidale P. L. , Liston G. E. , Lyons W. A. , and Chase T. N. , 1999: The influence of anthropogenic landscape changes on weather in south Florida. Mon. Wea. Rev., 127, 1663–1673.
Pielke, R. A., Sr., Adegoke J. , Beltrán-Przekurat A. , Hiemstra C. A. , Lin J. , Nair U. S. , Niyogi D. , and Nobis T. E. , 2007: An overview of regional land-use and land-cover impacts on rainfall. Tellus, 59B, 587–601.
Spreen, T. H., and Coauthors, 2006: An economic assessment of the future prospects for the Florida citrus industry. University of Florida, Institute of Food and Agricultural Sciences Rep., 168 pp. [Available online at http://www.fred.ifas.ufl.edu/floridajuice/pubs/EconAssessment.pdf.]
USDA, 2009: 2007 Census of Agriculture: United States Summary and State Data. Geographic Area Series, Vol. 1, U.S. Department of Agriculture, National Agricultural Statistics Service, 739 pp. [Available online at http://www.agcensus.usda.gov/Publications/2007/Full_Report/usv1.pdf.]
Wear, D. N., and Greis J. G. , Eds., 2002: Southern forest resource assessment - Technical report. Gen. Tech. Rep. SRS-53, U.S. Department of Agriculture, Forest Service, Southern Research Station, 635 pp.
