Abstract

The sensitivity of near-surface urban meteorological conditions to three different soil moisture initialization experiments under heat-wave conditions is investigated for the city of Melbourne, Australia. The Weather Research and Forecasting Model is used to simulate a domain over Melbourne and its surrounding rural areas. The experiments employ three suites of simulations. Two suites initialize the model with soil moisture from the top layer of the ERA-Interim soil moisture data with a 3-month and 24-h coupled spinup period, respectively. The third suite initializes the model with the arguably more realistic soil moistures from the Australian Water Availability Project (AWAP), which are an order of magnitude drier than the ERA-Interim data, again using a 24-h spinup period. The simulations employing the AWAP data are found to have smaller errors when compared with observations, with biases in urban maximum temperature reduced by 4.1°C and biases in the skin temperature reduced by 3.0°C relative to the biases of the 3-month-spinup experiment. Despite urban areas only having a small proportion of soil-covered surfaces, the results show that urban soils have a greater influence on urban near-surface temperatures at night, whereas rural soils have a greater influence on urban near-surface temperatures during the daytime.

1. Introduction

Interactions between soil moisture and surface temperature are important on a range of spatial and temporal scales, from cities to continents and from daily to decadal and longer time scales (Seneviratne et al. 2010). A negative soil moisture anomaly results in reduced evapotranspiration from the surface and a consequent increase in the sensible heat flux that leads to an increase in air temperature. Therefore, soil moisture can significantly affect the near-surface climate and can affect the occurrence of extreme heat and heat waves. Moreover, inaccurate initialization of soil moisture in models can affect the vertical profiles of meteorological variables in the lower boundary layer and their thermal-stability characteristics (Husain et al. 2014).

The role of antecedent soil moisture conditions on the dynamics and predictability of heat waves varies across the globe. The interaction has been thoroughly studied in the context of the devastating 2003 European heat wave (Lorenz et al. 2010; Miralles et al. 2012), for which the heat anomalies would have been considerably smaller if the soil moisture had not been so dry (Fischer et al. 2007). This was in part due to reduced cooling through latent heat and the positive feedback mechanism between soil moisture, temperature, and continental-scale circulation. Furthermore, during the 2003 European heat wave, drier soils promoted upper-air anticyclonic circulations and higher emissions of sensible heat, which led to higher temperatures (Zampieri et al. 2009). For the 2006 heat wave in California, it was found that ocean advection (Gershunov et al. 2009) rather than soil moisture anomalies (Miralles et al. 2012) caused the deadly heat.

The regions with the highest modeled soil moisture–temperature coupling strength were identified for the boreal summer (Koster et al. 2006). In particular, transitional climate zones and the boundary between wet and dry climates were “hot spots” for soil moisture and atmospheric temperature coupling activity. Moreover, the response of the atmosphere to soil moisture varies with geographic features (Stéfanon et al. 2012). Near to coastlines drier soil moisture can enhance the sea-breeze effect, resulting in cooler onshore temperatures, yet in mountainous areas drier soils increase the sensible heat flux, triggering convection.

Heat waves in southeastern Australia are typically driven by synoptic-scale processes comprising an anticyclone to the east of the continent that advects hot dry air from the desert interior of Australia toward the southeast of the country (Pezza et al. 2012). Perkins et al. (2015) found that antecedent soil moisture conditions had a large impact on Australian heat waves, with drier soils meaning longer and hotter heat waves and more heat-wave days per season. Indeed, antecedent soil moisture and its negative correlation with maximum temperature were shown to contribute greatly to the intensity and spatial extent of the January–February 2009 heat wave in southeastern Australia when simulated with a short lead time (Kala et al. 2015).

Kala et al. (2015) modeled the influence of antecedent soil moisture on the meteorological conditions of the Australian heat-wave and bushfire events that occurred in the summer of 2009. They found that, when modeling with short lead times, perturbations in the soil moisture were linearly related to changes in the maximum temperature and an intensification of an upper-level cyclone. The relationship became more nonlinear as the modeling lead time increased, with some parts of Australia showing an increase in maximum temperature with an increase in soil moisture. As such, once the lead time reaches 15 days, the impact of soil moisture perturbations on the maximum temperature became negligible. Notwithstanding, for southeastern Australia, accurate soil moisture profiles for model initialization are essential for simulating heat waves (Kala et al. 2015).

When considering the lack of soil or vegetation-covered surfaces in urban areas, the impacts of applying more accurate soil moisture to urban model simulations have yet to be established (Husain et al. 2014). Urban soil moisture can affect the magnitude of the urban heat island (UHI) (Hafner and Kidder 1999), the strength and stagnation of sea breezes (Chen et al. 2011b), and the soil temperature and air pollution concentration (Jacobson 1999). For Oklahoma City, Oklahoma, drier soil moisture is associated with warmer near-surface temperatures and enhanced low-level wind (Husain et al. 2014). Of interest is that altering the rural soil moistures surrounding a city has a far greater impact on urban and rural near-surface temperatures than adjusting only the urban soil moisture, which has a minimal influence on rural near-surface temperatures (Husain et al. 2014). Moreover, in Oklahoma City, there was a stronger impact of rural soil moisture on the nighttime UHI effect, whereas the impacts of urban soil moisture changes were confined to the magnitude of the daytime urban cool island. This result was due to the large impervious surface fraction in the city limiting the influence of urban soil moisture.

On 28–30 January 2009, the city of Melbourne in southeastern Australia experienced a record-breaking heat wave, with three days above 43°C and one night above 30°C. Following the event, over 374 excess deaths and 714 hospitalizations were attributed to the heat wave (Victorian Government Department of Human Services 2009). The January 2009 event came at the end of a 13-yr drought for southeastern Australia (Ummenhofer et al. 2009), with no rain recorded at Melbourne Airport weather station between 9 January and 7 February 2009 (Engel et al. 2013). The daily maximum temperature in Melbourne is known to be 1°–3°C higher after drought conditions (Nicholls and Larsen 2011), which may have contributed to the intensity of the 2009 heat wave.

There is considerable evidence that the initial conditions of soil moisture affect the modeling of urban environments by influencing the surface layer meteorological behavior (Husain et al. 2014; Jacobson 1999). Our research simulates the 28–30 January 2009 heat wave using three different methods of initializing soil moisture: one long coupled spinup simulation, one short spinup with wetter initial conditions, and one short spinup with drier initial conditions. We determine whether improving the accuracy of soil moisture influences the modeling of extreme heat in urban areas using Melbourne as a case study.

Heat waves over Australia are predicted to become hotter, longer, and more frequent with climate change (Cowan et al. 2014), and in some cities they have been shown to exacerbate the UHI (Li and Bou-Zeid 2013), which creates a higher risk of heat stress in urban areas (Fischer et al. 2012). There is also evidence that human resilience to heat is very sensitive to particular temperature thresholds (Nicholls et al. 2008), making more accurate simulations crucial—in particular, over cities. Therefore, it remains pertinent to simulate heat waves accurately and to understand the full impact of soil moisture initialization for the modeling of Australian heat waves, particularly in urban areas.

2. Data and methods

This study compares a suite of modeling experiments with instrumental near-surface and remotely sensed observations to determine how different soil-initialization techniques affect the accuracy of a modeled Australian heat wave.

a. Model setup

We used the Weather Research and Forecasting (WRF) Model, version 3.6.0 (Skamarock et al. 2008), driven by 0.7° × 0.7° gridded reanalysis data from ERA-Interim (Dee et al. 2011) to simulate the 28–30 January 2009 heat wave at a 2-km spatial resolution over Melbourne. WRF has previously been used to simulate urban meteorological conditions (Argüeso et al. 2014, 2015; Salamanca et al. 2012) and heat waves (Kala et al. 2015; Stegehuis et al. 2015) in Australia and overseas. WRF was coupled to a single-layer urban canopy model (Kusaka et al. 2001; Chen et al. 2011a) to better represent the urban features of the city.

The physics schemes employed were the Noah land surface scheme, the Mellor–Yamada–Janjić boundary layer scheme, the Kain–Fritsch cumulus scheme (only for the domains with a resolution of ≥10 km), the Dudhia shortwave radiation scheme, the Monin–Obukhov surface similarity scheme, the Rapid Radiative Transfer Model longwave radiation scheme, and the WRF single-moment 5-class microphysics scheme. This group of physics schemes has been deemed one of the best combinations for southeastern Australia from diurnal to weekly time scales (Evans et al. 2012).

To define urban areas, the Moderate Resolution Imaging Spectroradiometer (MODIS) land surface categories in WRF were applied. MODIS has only one urban category, but the single-layer urban canopy model in WRF allows for three different urban categories: low, medium, and high density. To define these three urban categories for Melbourne, the land surface dataset from Jackson et al. (2010) was employed. The distribution of MODIS land surface categories for the innermost WRF domain can be seen in Fig. 1, where water, evergreen broadleaf forest, and croplands are the most common land-use types.

Fig. 1.

The distribution of MODIS land-use categories across the inner WRF domain. Regions of high-, medium-, and low-density urban are defined by Jackson et al. (2010).

Fig. 1.

The distribution of MODIS land-use categories across the inner WRF domain. Regions of high-, medium-, and low-density urban are defined by Jackson et al. (2010).

Variations in urban surface characteristics such as building density and height have been shown to alter the surface energy balance, the boundary layer, and the urban climate as a whole in observational (Coutts et al. 2007) and modeled (Barlage et al. 2016) studies. Although the WRF default parameters were validated for Sydney, Australia (Argüeso et al. 2014), for our study the default parameters were unable to simulate a UHI in low-density regions of Melbourne. Therefore, the default WRF urban fraction parameters representing commercial (0.95), high-density (0.90), and low-density (0.50) urban were not used in this study. Instead, we defined high-density (0.80), medium-density (0.77), and low-density (0.71) urban regions as the fraction of urban surface that is not vegetation, on the basis of observed values over Melbourne (Coutts et al. 2007). All other urban parameters were left as the default WRF values for the single-layer urban canopy scheme.

b. Experimental design

Three experiments were performed to determine the sensitivity of temperature and humidity in heat waves to the initial conditions of soil moisture.

  1. Experiment 1 used ERA-Interim as the model driver from 1 November 2008 until 31 January 2009, giving almost 3 months of coupled spinup before the heat wave. The ERA-Interim soil moisture was applied here as the initial condition and then allowed to evolve. This long spinup method has been used to give the land-surface model time to adjust to the very dry conditions in this case study (e.g., Kala et al. 2015; Fig. 2a).

  2. Experiment 2 used ERA-Interim as the model driver for a 72-h simulation, with the first 24 h discarded as model spinup. The process was repeated until the heat wave was fully simulated. This was based on the methods of Grossman-Clarke et al. (2010). These 3-day simulations were repeated to form an 8-member ensemble to reduce the sensitivity to initial conditions (Georgescu et al. 2014). Each ensemble member was initialized 6 h apart (Fig. 2b).

  3. Experiment 3 ran WRF with the same setup as experiment 2 but with the top layer of ERA-Interim soil moisture (0–7 cm) being replaced with gridded 0.05° × 0.05° soil moisture data for January 2009 from the Australian Water Availability Project (AWAP; http://www.csiro.au/awap/) (Fig. 2c). The top layer of AWAP soil moisture 1 day before the heat wave began is an order of magnitude drier than the ERA-Interim soil moisture, showing the large difference between the two products (Fig. 2). We ran each ensemble from 1800 UTC 24 January until 1800 UTC 31 January 2009.

Fig. 2.

The top layer of soil moisture at 0000 UTC 27 Jan (1100 LT), the day before the heat wave began, (a) after 3 months of model spinup, (b) initialized with the reanalysis from ERA-Interim, and (c) initialized with AWAP soil moisture. Note that the scale only varies between 0 and 0.3 but in WRF soil moistures can vary between 0 and 1. Areas shown in white are water-covered surfaces and hence have soil moisture of 1.

Fig. 2.

The top layer of soil moisture at 0000 UTC 27 Jan (1100 LT), the day before the heat wave began, (a) after 3 months of model spinup, (b) initialized with the reanalysis from ERA-Interim, and (c) initialized with AWAP soil moisture. Note that the scale only varies between 0 and 0.3 but in WRF soil moistures can vary between 0 and 1. Areas shown in white are water-covered surfaces and hence have soil moisture of 1.

The AWAP soil moisture is a high-resolution reanalysis product for Australia (Raupach et al. 2009). The AWAP soil moisture uses observational soil information from the Digital Atlas of Australian Soils (McKenzie and Hook 1992; Mckenzie et al. 2000) as well as the gridded AWAP maximum temperature, minimum temperature, precipitation, and solar irradiance (Jones et al. 2009). It combines these with the remotely sensed land surface temperature and brightness temperature from the NOAA Advanced Very High Resolution Radiometer and Advanced Along-Track Scanning Radiometer satellite instruments and monthly climatological means of vegetation greenness derived from the “fraction of absorbed photosynthetically active radiation” from the SeaWiFS satellite instrument. It then uses a model-data fusion approach that combines the observations and models (Raupach et al. 2009). The AWAP soil moisture has been validated against multidecadal streamflow data from over 200 catchments around Australia.

One-dimensional soil moisture models (Reen et al. 2006) and multilevel groundwater-resolving (Maxwell et al. 2007) land-surface models have been shown to have an impact on simulation of the atmospheric boundary layer. Therefore, to test the impact of changing only the top layer of soil moisture in experiment 3, we repeated the simulations but with the second layer of soil moisture reduced by 50%. The difference was minimal, justifying our decision to alter only the top layer of soil moisture in experiment 3.

For experiment 1, we had three domains, with Melbourne and the surrounding area as the innermost domain. The largest domain had a 50-km grid resolution, the middle domain had a 10-km resolution, and the innermost domain had a 2-km resolution (Fig. 3). For experiments 2 and 3, we used the 10-km and 2-km inner domains from experiment 1. All domains used one-way nesting.

Fig. 3.

The three WRF domains used in experiment 1; experiments 2 and 3 use the inner two domains only.

Fig. 3.

The three WRF domains used in experiment 1; experiments 2 and 3 use the inner two domains only.

c. Observational data

To validate our three model experiments, we used four observational datasets. First, we obtained gridded AWAP daily maximum and minimum temperature data and 0900 and 1500 LT vapor pressure data (Jones et al. 2009; hereinafter LT indicates Australian eastern daylight time). These data were interpolated from observational weather stations around Australia to a 0.05° × 0.05° grid. To compare the 2-km-resolution WRF with the ~5-km-resolution AWAP data, we interpolated the WRF data to the AWAP grid using a nearest-neighbor method so that the two datasets were directly comparable. The nearest-neighbor method selects the value from the original grid point that is closest to the new grid point.

To compare the modeled skin temperature with observations, we obtained MODIS “MYD11A1” land surface temperature data on a 1-km grid (Salomonson et al. 1989; Wan 2008). The land surface temperature data were regridded to the WRF 2-km grid using a nearest-neighbor method. We compare the land surface temperature at 1400 LT on the first day of the heat wave, 28 January, because this period is when the satellite data are available and coverage was the highest.

Next, 3-hourly 2-m temperature data from 11 urban and rural Australian Bureau of Meteorology weather stations in Melbourne and the surrounding areas were obtained to compare the results from the closest WRF grid point with the weather stations (Table 1). The weather stations were classified into urban and rural categories following the method of Trewin (2013), which uses the land use of a 2-km radius around the station to determine whether it is urban, urban fringe, or rural. Stations that are classified as urban fringe are counted as urban in this study.

Table 1.

The latitude, longitude, altitude, and classification of the 11 weather stations.

The latitude, longitude, altitude, and classification of the 11 weather stations.
The latitude, longitude, altitude, and classification of the 11 weather stations.

We calculated the root-mean-square error (RMSE) between WRF and the weather station data for the warmest part of the day (1200–1800 LT) for the three days of the heat wave and for the coolest part of the day (0000–0600 LT) for the two middle nights of the heat wave. A cold front swept across Melbourne at 1800 LT on the final day of the heat wave, yet WRF simulated the cold front arriving at 2000 LT. This resulted in an unusually large error between the observed 2-m temperature and the WRF values, which led to an overinflation of the RMSE values. Therefore, for the RMSE calculations for the warmest part of the day, we have excluded the 1800 LT value on 30 January, the final day of the heat wave. The observational data are from 0000 LT 28 January to 0000 LT 31 January 2009; details about the weather stations are in Table 1.

Calculations of the UHI were made using three urban stations (Laverton, Scoresby, and Viewbank) and three rural stations (Avalon, Coldstream, and Sheoaks). We averaged the 3-hourly temperatures for the urban group and the rural group and then took the difference between urban and rural to calculate the UHI. For calculation of the UHI in WRF, the closest model grid point to the observational weather station was used. The rural stations Avalon, Coldstream, and Sheoaks were classified in WRF as croplands, grasslands, and grasslands, respectively, and the urban stations Laverton, Scoresby, and Viewbank were classified as low-density, medium-density, and medium-density urban, respectively. The RMSE of the three WRF experiments was then calculated for the UHI over the entire 3 days of the heat wave.

To determine how well the three experiments simulate the atmospheric profile, particularly in the lower atmosphere, we sourced 12-hourly atmospheric soundings from the University of Wyoming archive (http://weather.uwyo.edu/upperair/sounding.html) to compare with our WRF simulations. The boundary layer is of importance because it is where the largest difference between urban and rural weather conditions occurs, as a result of the influence of the urban form on the energy and momentum budgets at the surface. The soundings are available at 0000 and 1200 UTC (1100 and 2300 LT).

We calculated the temperature and vapor pressure bias at 2 m between WRF and AWAP and the skin temperature bias between WRF and MODIS. We also calculated the RMSE between WRF, the observational weather stations, and the UHI. Last, we compared the boundary layer temperature, dewpoint temperature, and wind speed between WRF and atmospheric soundings. The experiment with the lowest bias or RMSE will highlight which soil moisture initialization experiment best simulates the January 2009 heat wave over Melbourne.

3. Results

a. Comparison of near-surface variables

When looking exclusively at the rural areas of the innermost domain, all three WRF experiments overestimated the minimum temperature (warm bias) and underestimated the maximum temperature (cool bias) (Table 2). As such, WRF was unable to capture the diurnal temperature range of the heat wave in rural areas, where experiments 1, 2, and 3 underestimated the diurnal temperature range by 9.2°, 8.1°, and 5.6°C, respectively. Experiment 1 has the smallest bias in minimum temperature (2.9°C), and experiment 3 has the smallest bias in maximum temperature (−1.7°C). Experiment 1 was the coolest throughout the heat wave, and this result is highlighted by the average maximum temperature bias of −6.2°C for this simulation. However, this helped it to achieve more realistic minimum temperatures as experiments 2 and 3 were too warm at night. After 3 months of model spinup, the soil moisture in experiment 1 was the wettest at the beginning of the heat wave (Fig. 2), causing the cooler temperatures. This confirms past work showing that antecedent soil moisture conditions are known to have a negative correlation with maximum temperature (Kala et al. 2015).

Table 2.

The heat-wave area average bias (WRF minus observations) over the rural grid points in the innermost domain only. Boldface text indicates the experiment that performed best.

The heat-wave area average bias (WRF minus observations) over the rural grid points in the innermost domain only. Boldface text indicates the experiment that performed best.
The heat-wave area average bias (WRF minus observations) over the rural grid points in the innermost domain only. Boldface text indicates the experiment that performed best.

The bias in rural 1400 LT skin temperature produces results that are similar to those for the maximum temperature; all experiments produce negative biases, and experiment 3 has the smallest skin temperature bias of −3.2°C. This bias is 5.8° and 3.1°C lower than those for experiments 1 and 2, respectively. Drier soils heat faster during the day than wetter soils because they have a lower heat capacity; hence, experiment 1 has the largest negative bias in skin temperature and experiment 3 has the smallest. The negative model bias in skin temperature and maximum temperature for all three experiments suggests that, despite the very dry AWAP soils, the model simulations still need to be hotter and drier.

These results are further highlighted by the spatial differences between the three experiments and the AWAP maximum and minimum temperature (Fig. 4). Experiment 1 has the largest difference between its maximum temperatures and the observations; the simulation is up to 9°C colder than AWAP (Fig. 4a). In contrast, experiment 2 represents an improvement, with maximum temperatures up to 6°C cooler than observed (Fig. 4b). Experiment 3 performs best, with temperatures up to 3°C cooler than observed (Fig. 4c).

Fig. 4.

The difference between the 1500 LT temperature and the AWAP maximum temperature for (a) experiment 1, (b) experiment 2, and (c) experiment 3, and the difference between the 0300 LT temperature and the AWAP minimum temperature for (d) experiment 1, (e) experiment 2, and (f) experiment 3 for the second day of the heat wave.

Fig. 4.

The difference between the 1500 LT temperature and the AWAP maximum temperature for (a) experiment 1, (b) experiment 2, and (c) experiment 3, and the difference between the 0300 LT temperature and the AWAP minimum temperature for (d) experiment 1, (e) experiment 2, and (f) experiment 3 for the second day of the heat wave.

The spatial distribution of the minimum temperature bias shows that experiment 1 has the smallest biases across the inner domain (Fig. 4d). The minimum temperature for all three experiments is hotter than observations in the southwestern corner of the domain, with biases exceeding +10°C. The warm bias in minimum temperature over rural areas is smaller for experiment 3 than for experiment 2 (Table 2) because experiment 3 also has a negative bias in minimum temperature across the northwestern and southwestern regions of the domain that helps to reduce the large positive bias in the southeast (Fig. 4f). Experiment 2 has very few regions of negative bias for the minimum temperature (Fig. 4e), resulting in it having the largest minimum temperature bias of the three experiments (Table 2).

Heat waves in southeastern Australia are often dry (Pezza et al. 2012), suggesting that an accurate simulation of the near-surface humidity is essential for the modeling of heat waves. For rural areas, experiment 3 has the smallest vapor pressure bias at 0900 and 1500 LT (+1.9 and +2.1 hPa, respectively), and experiment 1 had a smaller bias than experiment 2 at 0900 LT (Table 2). All three experiments had a positive bias for both time periods, showing that WRF is too humid near the surface, even in the driest simulation. The vapor pressure bias was smaller at 0900 LT for all three experiments because the environment is more humid at that time as a result of the diurnal variability of moisture input to the atmosphere; hence, it is easier for WRF to achieve those conditions. At 1500 LT, WRF is less able to simulate the drier conditions and thus the moist bias increases. Nevertheless, the 1500 LT bias in experiment 3 was considerably smaller than it was in the other experiments (1.5 hPa lower than experiment 1 and 1.1 hPa lower than experiment 2) because of it being initialized with soil moistures that were an order-of-magnitude drier.

Urban areas are drier, are warmer, and have significantly more impervious surfaces than rural regions (Coutts et al. 2013). Therefore, we investigated the importance of soil moisture initialization on simulating the near-surface meteorological variables in urban areas, where there is less vegetation and fewer regions of exposed soil.

The urban maximum and minimum temperature biases in the innermost domain (Table 3) were analogous to the rural results (Table 2). All three experiments underestimated the urban diurnal temperature range, experiment 1 had the smallest minimum temperature bias, and experiment 3 had the smallest maximum temperature bias. In urban areas, however, the maximum temperature biases for experiments 1 and 2 were 0.2° and 0.3°C smaller, respectively, when compared with rural areas, whereas the bias in experiment 3 was 0.2°C larger. The higher rural soil moisture content in experiment 1 meant that it was too cool to capture the maximum temperature of the heat wave, yet the simulation improved slightly in the drier urban areas. In contrast, the skin temperature biases were much larger in urban areas than in rural areas (Table 3). Experiment 3 had the smallest urban skin temperature bias of −7.5°C, yet that is 4.3°C larger than the rural bias from experiment 3. This is because the MODIS land surface temperature at 1400 LT is 2.3°C warmer in urban areas than in rural areas, yet the simulation in experiment 3 is 2.0°C cooler in urban areas, creating a larger skin temperature bias for urban areas. The modeled 1400 LT skin temperature is cooler in urban areas than in rural areas because WRF simulates an urban cool island during the day of the heat wave, albeit of a smaller magnitude than the nighttime UHI. Furthermore, satellites are known to have larger land surface temperature errors over urban areas because of the emissivity errors from the different kinds of rooftops.

Table 3.

The heat-wave area average bias (WRF minus observations) over urban areas in the innermost domain only. Boldface text indicates the experiment that performed best.

The heat-wave area average bias (WRF minus observations) over urban areas in the innermost domain only. Boldface text indicates the experiment that performed best.
The heat-wave area average bias (WRF minus observations) over urban areas in the innermost domain only. Boldface text indicates the experiment that performed best.

When comparing urban biases with rural biases, the minimum temperature bias was 0.5° and 0.1°C larger for experiments 1 and 3, respectively, in urban areas yet was 0.2°C smaller for experiment 2 (cf. Table 3 with Table 2). In a similar way, the diurnal temperature range bias was 0.2° and 0.3°C larger for urban areas in experiments 1 and 3, respectively, yet was 0.3°C smaller for urban areas in experiment 2. The reason for experiment 2 having smaller biases in urban areas while experiments 1 and 3 had smaller biases in rural areas is the considerably drier urban soils relative to the rural soils in experiment 2 (Fig. 2). In contrast, the urban and rural soil moistures were much more uniform across the inner domain in experiments 1 and 3.

The three WRF experiments had smaller vapor pressure biases at 0900 and 1500 LT in urban areas than in rural areas (cf. Table 3 with Table 2). The driest simulation, experiment 3, had the smallest urban vapor pressure bias yet was still too humid at the surface when compared with observations. The urban vapor pressure bias for experiment 3 was 0.9 and 0.5 hPa smaller than in rural areas for 0900 and 1500 LT, respectively.

Of interest is that, for experiments 1 and 2, the considerably smaller urban vapor pressure biases when compared with the rural biases suggest that it is the urban features such as impervious surfaces and high runoff rates, rather than the different soil moisture initializations, that have the largest impact on simulating the near-surface urban humidity. Hence, the simulations in urban areas are closer to the unusually dry reality of the January 2009 heat wave. Note, however, that the soil moisture over the whole domain of the simulation still determines the magnitude of the bias (see section 4), as demonstrated by experiment 2 being drier than experiment 1 and as a result having smaller vapor pressure biases.

The rural energy balances for experiments 2 and 3 are as expected in that the drier experiment 3 has a larger sensible heat flux and smaller ground and latent heat fluxes when compared with experiment 2 (Figs. 5b,c). The largest difference between experiments 2 and 3 for the rural sensible heat flux and ground heat flux occurs at 1300 LT, with differences of 67.7 and 53.3 W m−2, respectively, whereas the largest difference in the latent heat flux of 30.4 W m−2 occurs at 1500 LT. The greater amounts of soil moisture in experiment 2 enable more energy to be stored in the surface, resulting in a larger ground heat flux. The extra soil moisture is also available for evapotranspiration and latent heat exchange. In the drier experiment 3, more energy is partitioned into the sensible heat flux because it cannot be stored as efficiently in the ground.

Fig. 5.

The 24-h sensible heat flux (red lines), latent heat flux (blue), and ground heat flux (gray) for (a),(d) experiment 1, (b),(e) experiment 2, and (c),(f) experiment 3 over (top) rural and (bottom) urban areas of the innermost domain averaged over the three days of the heat wave. Note that a positive heat flux means that energy is moving away from the ground.

Fig. 5.

The 24-h sensible heat flux (red lines), latent heat flux (blue), and ground heat flux (gray) for (a),(d) experiment 1, (b),(e) experiment 2, and (c),(f) experiment 3 over (top) rural and (bottom) urban areas of the innermost domain averaged over the three days of the heat wave. Note that a positive heat flux means that energy is moving away from the ground.

The urban results for experiments 2 and 3 are analogous to the rural results, albeit with much smaller differences between the sensible, latent, and ground heat fluxes because of the larger portion of impervious surfaces reducing the effect of soil moisture (Figs. 5e,f). For example, the maximum urban sensible heat flux for experiment 3 is 13.6 W m−2 larger than that for experiment 2. The ground heat flux is larger for urban areas than for rural areas because of the urban geometry and available heat storage from buildings and roads. Despite the smaller vegetated fraction in urban areas, the results still show that the sensible heat flux is larger and the ground heat flux and latent heat flux are smaller in the drier simulation.

In contrast, the wetter experiment 1 has a rural sensible heat flux that is 30.9 W m−2 larger than that of experiment 2 at 1100 LT (Fig. 5a) and has the largest urban maximum sensible heat flux in urban areas of any of the three experiments at 339.2 W m−2 (Fig. 5d). Furthermore, the urban ground heat flux is the smallest for experiment 1 despite it containing larger amounts of soil moisture. These results were not expected. The difference in net radiative flux among the three experiments is minimal (not shown), suggesting that the results for experiment 1 could be influenced by changes to the advective weather systems in the 3-month coupled spinup.

When comparing the 2-m temperature RMSE from the warmest (1200–1800 LT) and coolest (0000–0600 LT) parts of the day, there is no clear evidence that WRF is able to simulate urban or rural conditions better at the 11 observational weather stations (Table 4). Indeed, two rural stations, Sheoaks and Coldstream, have the smallest and largest RMSE values, respectively. In contrast, it is clear that experiment 3 has lower RMSEs than experiments 1 and 2 when comparing the closest WRF grid point with the weather stations (Table 4). This finding confirms the gridded-data results in which the drier simulation enables WRF to better capture the magnitude of the heat wave. Note that WRF is on a 2-km grid, and so choosing one representative point to compare with a weather station is a strict test of the model. Nevertheless, the weather-station results broadly agree with the gridded data.

Table 4.

The RMSE between the observational weather stations and the closest WRF grid point over the warmest part of the day (1200–1800 LT) and the coolest part of the day (0000–0600 LT) for the three experiments. Boldface text for each time period indicates the experiment that performed best.

The RMSE between the observational weather stations and the closest WRF grid point over the warmest part of the day (1200–1800 LT) and the coolest part of the day (0000–0600 LT) for the three experiments. Boldface text for each time period indicates the experiment that performed best.
The RMSE between the observational weather stations and the closest WRF grid point over the warmest part of the day (1200–1800 LT) and the coolest part of the day (0000–0600 LT) for the three experiments. Boldface text for each time period indicates the experiment that performed best.

With Moorabbin used as a typical urban example, Fig. 6 shows that experiment 3 has a greater diurnal variability than experiments 1 and 2 and that, when compared with observations, it is the most accurate simulation during both the warmest and coolest parts of the heat wave. The 2-m temperatures in experiment 1 do not exceed 40°C at any stage during the simulation; experiment 2 exceeds 40°C for each day of the heat wave but also had a large warm bias at night.

Fig. 6.

(a) The 3-hourly 2-m temperature for Moorabbin weather station observational data (gray line), experiment 1 (blue), experiment 2 (green), and experiment 3 (orange). (b) Experiment 1 minus observations (blue line), experiment 2 minus observations (green), and experiment 3 minus observations (orange). Times are local time (UTC + 11 h).

Fig. 6.

(a) The 3-hourly 2-m temperature for Moorabbin weather station observational data (gray line), experiment 1 (blue), experiment 2 (green), and experiment 3 (orange). (b) Experiment 1 minus observations (blue line), experiment 2 minus observations (green), and experiment 3 minus observations (orange). Times are local time (UTC + 11 h).

In general, experiment 3 has the smallest difference between observations and model simulations throughout the warmest and coolest part of the day (Table 4). However, there were times when experiment 1 or 2 performed better, especially on the first night of the heatwave. The 11 weather stations are spread throughout Melbourne and the surrounding region at a range of altitudes and geographic locations, from coastal areas to a mountain valley. Therefore, variability is expected between the weather conditions and how well each model experiment can simulate the idiosyncrasies for each station. This result suggests that the errors are not uniform across the heat wave and that each experiment performed better during different days and nights of the heat wave. Nevertheless, the RMSEs show that, overall, experiment 3 was the closest to reality.

b. Soundings

While we have established differences at the surface among the three soil moisture experiments, we also look at soundings to investigate whether the differences extend to the lower atmosphere. Atmospheric observations from weather balloons are only available from Melbourne Airport weather station at 0000 and 1200 UTC, which corresponds to 1100 and 2300 LT. At these times, the atmospheric soundings from the WRF grid point that is closest to Melbourne Airport were almost identical between experiments 2 and 3, except in the boundary layer, which is defined here as being below 850 hPa. Although no sounding data were available, the three experiments showed the largest differences during the warmest and coolest part of the day (approximately 1500 and 0300 LT) rather than at 1100 or 2300 LT. Because the largest differences among experiments were in the lower atmosphere and boundary layer, we now compare the temperature, dewpoint temperature, and wind speed between the three experiments and the atmospheric soundings between the 1000- and 850-hPa pressure levels only.

Between 1000 and 950 hPa, experiments 2 and 3 are hotter, drier, and windier than experiment 1 since they have significantly lower soil moistures (Fig. 7). Evidence from past studies shows that drier urban soil moistures increase low-level wind speeds (Jacobson 1999; Husain et al. 2014), and this is further demonstrated by experiment 3 having the fastest wind speeds from 1000 to 850 hPa (Fig. 7c). The wind speeds from experiment 3 are the most realistic (i.e., closest to observations) between 1000 and 925 hPa, with the slightly slower wind speeds of experiment 2 being more accurate from 925 to 850 hPa. The increase in low-level winds is due to the higher ground temperature raising the sensible heat flux; hence, there is a greater turbulent transport of momentum to the surface (Husain et al. 2014).

Fig. 7.

The (a) air temperature, (b) dewpoint temperature, and (c) wind speed vs pressure (i.e., height) for the three experiments and observations from Melbourne Airport for 0000 UTC (1100 LT) on the second day of the heat wave.

Fig. 7.

The (a) air temperature, (b) dewpoint temperature, and (c) wind speed vs pressure (i.e., height) for the three experiments and observations from Melbourne Airport for 0000 UTC (1100 LT) on the second day of the heat wave.

The temperatures in experiments 2 and 3 are the closest to observations between 1000 and 950 hPa, and experiment 3 performs better above 950 hPa (Fig. 7a). The difference between the observational temperatures and experiment 3 is approximately 1°–2°C above 950 hPa. In contrast, the temperatures in experiment 1 are approximately 4°C colder than observations throughout the boundary layer.

The minimal difference between the temperatures of experiments 2 and 3 in the lower atmosphere are not replicated in the dewpoint temperature. All three experiments are much too moist throughout the boundary layer when compared with observations, although experiment 3 performs better below 950 hPa and experiment 1 performs better above that level (Fig. 7b). This result demonstrates that, despite improvements to the soil moisture, WRF is still unable to capture the exceedingly dry conditions in the boundary layer, including the dewpoint temperature inversion during the heat wave. Of interest is that at the surface experiment 1 has a dewpoint temperature that is approximately 8°C higher than observations but above 950 hPa it becomes the driest and, hence, the closest to reality. This only occurs for the second day of the heat wave (shown in Fig. 7b). On the first and final days, experiment 1 has the highest and most inaccurate dewpoint temperature throughout the boundary layer (not shown).

Because the greatest variation among the three experiments occurred during the warmest and coolest parts of the day, we produced soundings from the output of each experiment for Melbourne Airport at 0300 and 1500 LT on the second day of the heat wave (Fig. 8). At 0300 and 1500 LT, experiment 3 is the driest and warmest in the boundary layer, experiment 1 is the coolest at 1500 LT, and experiment 2 is the wettest at 1500 LT. The difference in the boundary layer dewpoint temperature for experiments 1 and 3 is negligible at 1500 LT. (Fig. 8b). Above 700 hPa, the temperature and wind profiles in the atmosphere are almost identical for experiments 2 and 3, suggesting that the significantly drier soils only affect the simulation in the lower levels of the atmosphere.

Fig. 8.

Soundings of experiment 1 (blue lines and wind barbs), experiment 2 (green), and experiment 3 (orange) on 29 Jan 2009, the second day of the heat wave, for (a) 0300 and (b) 1500 LT.

Fig. 8.

Soundings of experiment 1 (blue lines and wind barbs), experiment 2 (green), and experiment 3 (orange) on 29 Jan 2009, the second day of the heat wave, for (a) 0300 and (b) 1500 LT.

Experiment 1 produces an unrealistically dry atmosphere at 0300 LT between 700 and 450 hPa (Fig. 8a), and this occurs throughout each day of the heat wave in this experiment. This is a result of WRF oversimulating atmospheric subsidence. While subsidence above the boundary layer was found to contribute greatly to mid- and upper-tropospheric warming during this heat-wave event (Parker et al. 2014), dewpoint temperatures below −30°C are not credible, providing an indication that the long coupled spinup in experiment 1 had caused WRF to drift too far from reality.

c. Implications for UHI simulation

We have investigated how the three soil moisture initialization experiments affected the simulation of the UHI. The observational UHI was calculated as the difference between the average of the 3-hourly temperature from three urban and three rural weather stations (see section 2c). The simulated UHI was calculated using the WRF grid point that is closest to the observational weather stations selected for the UHI analysis.

The RMSEs of the UHI between experiments 1, 2, and 3 and observations were 1.9°, 1.6°, and 1.6°C, respectively, indicating that experiments 2 and 3 best simulated the UHI. Indeed, the largest difference between the model simulations and observations occurred at 0300 LT on the second day of the heat wave, at which time experiment 1 underestimated the UHI by more than 4°C (Fig. 9). The RMSE of the model experiments was increased by WRF mistiming the peak of the UHI on each day of the heat wave. Experiment 3 greatly overestimated the urban cool island at 0900 LT on the first and second days of the heat wave yet was the closest to observations at 0900 LT on the third day of the heat wave.

Fig. 9.

(a) The UHI from 28 to 30 Jan 2009 on the basis of three urban stations (Laverton, Scoresby, and Viewbank) and three rural stations (Avalon, Coldstream, and Sheoaks) for the observations (gray line), experiment 1 (blue), experiment 2 (green), and experiment 3 (orange). (b) Difference between the simulated UHI and observations for experiment 1 (blue), experiment 2 (green), and experiment 3 (orange). Times are local time (UTC + 11 h).

Fig. 9.

(a) The UHI from 28 to 30 Jan 2009 on the basis of three urban stations (Laverton, Scoresby, and Viewbank) and three rural stations (Avalon, Coldstream, and Sheoaks) for the observations (gray line), experiment 1 (blue), experiment 2 (green), and experiment 3 (orange). (b) Difference between the simulated UHI and observations for experiment 1 (blue), experiment 2 (green), and experiment 3 (orange). Times are local time (UTC + 11 h).

We have diagnosed that the improvement of the UHI simulation in experiments 2 and 3 relative to experiment 1 was due to improvements in the urban areas as opposed to in the rural areas. This was because the improvement in RMSE for each station used in UHI calculations over the heat wave was greater for the urban areas than for the rural areas. For example, the average difference in RMSE between experiment 1 and experiment 3 for the urban stations was 1.2°C whereas for the rural stations it was 0.6°C. This result suggests that, despite little soil coverage in urban areas, more accurate soil moistures in urban tiles in WRF can greatly improve simulation of the UHI.

The results described in this section have demonstrated that having more accurate urban soil moisture initial conditions improves the WRF simulation of the near-surface temperature and vapor pressure, skin temperature, lower atmospheric temperature, dewpoint temperature, and wind speeds and, therefore, the UHI during this heatwave event.

4. Discussion

Despite 3 months of coupled-model spinup during extreme drought conditions where there had been no rainfall since 9 January 2009 (Engel et al. 2013), experiment 1 had an urban top layer of soil moisture that was 6 times as wet as that of experiment 3. The discrepancy was similar in rural areas. With these errors in soil moisture, the ability to simulate the heat wave was limited to such an extent that the event did not register as a severe heat wave in this experiment, with maximum temperatures of approximately 37°C each day, as opposed to the maximum temperatures of 43°–45°C that occurred in reality. Past studies have used long coupled model spinup times in WRF (e.g., 3 months) to allow the boundary conditions, such as soil moisture, to stabilize into a more realistic state (Kala et al. 2015; Georgescu et al. 2014; Cortés-Hernández et al. 2016), yet this means that the boundary conditions are also susceptible to inevitable model biases. Our results provide evidence that such a long coupled spinup can enable the model to “drift” too far from reality in terms of soil moisture, as it has in our series of experiments. We have shown that this effect can make a considerable difference to the simulation of extreme heat in rural areas as well as urban areas that have a lower soil content and, hence, receive less influence from soil moisture initialization. These results should not be confused with long offline land-surface model spinups (often greater than 1 yr) that have been proven to provide more realistic soil moisture and soil temperature initial conditions in urban areas (Sharma et al. 2017).

The difference between the observed maximum and minimum temperature during the heat wave exceeded 20°C on the second day of the heat wave (Fig. 6a). This is a considerable diurnal temperature range for WRF to simulate, and in experiment 1 it underestimated the diurnal temperature range of the heat wave by almost 10°C. Argüeso et al. (2014) found similar results when using WRF to model summer temperatures in Sydney, whereby WRF also underestimated the diurnal temperature range when compared with observations. By including the drier and more accurate AWAP soil moisture in experiment 3, WRF was able to almost halve the size of the diurnal temperature range bias for the heat wave without the need for bias-correction techniques (Argüeso et al. 2014; Dosio and Paruolo 2011).

The RMSEs for the WRF grid point that is closest to the observational weather stations show that there is no systematic bias (e.g., warmer bias at urban locations; cooler bias in rural areas) across the 11 stations. This result is partly due to unusual observations at several weather stations, including the very high nighttime temperature at Melbourne Airport on the first night that do not feature at all of the stations. It is also due to the geographic range of the rural stations. Avalon is located on the coast, Mangalore is approximately 108 km inland, and Sheoaks is over 200 m above sea level. Indeed, the RMSE at Coldstream for experiments 2 and 3 is over 7°C during the coolest part of the day yet is less than 2.5°C during the warmest part (Table 4). The likely reason for the large difference at night between WRF and the observations is the geographical surroundings of Coldstream weather station. Coldstream is located in a valley that is surrounded by mountains and can experience cool-air drainage (Trewin 2005) during the night (B. C. Trewin 2016, personal communication). Therefore, the observed minimum temperatures are significantly lower than the WRF simulations. Moreover, the inner domain of the WRF simulations is on a 2-km grid, which means the topography of the Coldstream region may not be fully recognized.

During the heat wave, the three experiments were too moist at the surface (Tables 2 and3), and these biases extended into the boundary layer (Fig. 7b). Therefore, WRF was unable to completely capture the unusually dry conditions experienced during this heat wave, even when the soil moisture has been initialized with a local and extremely dry product, such as in experiment 3. Nevertheless, initializing experiment 3 with the AWAP soil moisture produced a considerable improvement in WRF’s ability to simulate the diurnal variability of the heat-wave temperature and near-surface humidity. Moreover, it meant that a long coupled spinup to produce accurate initial conditions was unnecessary, even in urban areas with little exposed soil and vegetation.

It is worth considering how much of the improved urban heat-wave simulation of experiment 3 is due to having accurate soil moistures in urban areas or is due to accurate soil moistures across the entire model domain. Given that the maximum temperature bias was 0.2°C larger in urban areas than in rural areas, it would appear that the urban soil moisture exerts a minimal influence on the daytime maximum temperature and that accuracy stems from having correct soil moistures across the whole inner model domain. Indeed, Fig. 8b shows that at 1500 LT there are strong northerly winds near the surface, suggesting a large amount of advection. Therefore, isolating the influence of urban soils would be difficult because the air originates from the rural areas to the north of the city. In contrast, at night, there is a strong inversion and almost no wind near the surface, suggesting a highly stable lower atmosphere, indicating that the urban soil moisture would exert an influence at night (Fig. 8a). These results are in contrast to Husain et al. (2014), who found that urban soil had a greater influence in the daytime and a minimal influence at night in Oklahoma City. Consequently, when using WRF to simulate a dry-climate heat wave in urban areas, accurate soil moistures across the whole domain are vital for achieving realistic maximum temperatures, whereas minimum temperatures also rely on urban soils despite their small coverage. This result highlights the importance of using an accurate product, even for regions where the influence of soil moisture might be thought of as reduced (e.g., in urban areas).

Improving the soil moisture is just one aspect of modeling near-surface urban weather. Urban parameters such as building density and height and the particular WRF urban canopy model used have been shown to greatly influence the accuracy of simulations (Barlage et al. 2016; Lee et al. 2011), and biases can also form from the large-scale forcing conditions (Hawkins and Sutton 2009; Dee et al. 2011)—in this case, inherent biases in the ERA-Interim data. Although we have addressed one large issue of soil moisture accuracy in this research, the simulations could still be improved, as is seen with the overestimation of simulated dewpoint temperatures in Fig. 7b.

One limitation of using the AWAP soil moisture product is that it is only available for Australia and so would not assist regional and urban model simulations elsewhere. For dry-climate modeling, caution should be advised when using the ERA-Interim dataset because it is known to overestimate soil moisture for dry land (Albergel et al. 2012). For example, ERA-Interim has an average bias of +0.1 m3 m−3 for Australia and Spain when compared with in situ soil moisture measurements (Albergel et al. 2013). This bias is large, considering that the field capacity of southeastern Australian soils is approximately 0.4 m3 m−3.

An alternative is to use global datasets of “leaky bucket” soil moisture models (Fan and van den Dool 2004) and satellite-derived datasets such as from the Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) and the Soil Moisture Ocean Salinity (SMOS) projects (e.g., De Jeu et al. 2014; Parinussa et al. 2014). Their coarse resolution is detrimental to regional modeling, however, because the heterogeneity of the soil moistures is not well represented (Piles et al. 2011). Hence, downscaled satellite-derived soil moisture datasets have been developed for the United States (Anderson et al. 2007), the Iberian Peninsula (Piles et al. 2011), southern France (Piles et al. 2016), and the Tibetan Plateau (Zhao and Li 2015). Although not every country has the resources to produce a high-resolution gridded soil moisture dataset, for regional modeling in a dry climate it is recommended that other soil moisture products be explored to increase the accuracy of the simulations.

5. Conclusions

This study examined three different soil moisture initialization experiments in WRF and how they affected the simulation accuracy of a southeastern Australian heat wave. Our results indicate that using the AWAP soil moisture with WRF produced the most accurate heat-wave simulation and, by proxy, also provided a form of validation for the AWAP product for use with Australian heat-wave modeling and analysis. The AWAP soil moisture product is preferable to using the default values that are available from the ERA-Interim global reanalysis since it is known to overestimate soil moistures over dry land. Furthermore, it is preferable to tuning model soil moistures until the desired results are produced, because tuning does not guarantee that the initial conditions are realistic.

The AWAP soil moisture product has never been used in Australian heat-wave simulations before or in regional downscaling studies. The evidence provided here shows that, at least for this case study, it is a viable and preferable alternative to traditional products (e.g., ERA-Interim). Furthermore, we have demonstrated that including a more accurate representation of soil moisture in WRF can improve the simulation of a heat wave even in areas where soil moisture is arguably less important (i.e., included in a small fraction of the land surface tile), such as urban areas.

Validating simulations over an urban area provides information on the veracity of risk assessments of current and future heat waves for urban populations. Whereas the January 2009 heat wave occurred at the end of a 13-yr drought for southeastern Australia, producing unusually dry conditions, climate change is predicted to produce hotter, longer, and more frequent heat waves (Cowan et al. 2014) as well as more droughts in the future for Australia (Mpelasoka et al. 2008). Soil moisture is heavily related to droughts and has a very large impact on the intensity of Australian heat waves (Perkins et al. 2015). Therefore, using the available resources to produce accurate representations of the current climate will help in understanding how best to model heat waves in the future climate.

Acknowledgments

This research was supported by the CRC for Water Sensitive Cities, ARC DECRA Project DE150101297, the ARC Centre of Excellence for Climate System Science (Grant CE 110001028), and the NCI National Facility. The authors thank Daniel Argüeso from the University of Hawai‘i at Mānoa and Carlo Jamandre and Melissa Hart from the University of New South Wales for helping with WRF Model setup, Peter Briggs from the Commonwealth Scientific and Industrial Research Organisation for supplying AWAP soil moisture data, the Australian Bureau of Meteorology for providing the station data, and the ECMWF for supplying the ERA-Interim data.

REFERENCES

REFERENCES
Albergel
,
C.
,
P.
de Rosnay
,
G.
Balsamo
,
L.
Isaksen
, and
J.
Muñoz-Sabater
,
2012
:
Soil moisture analyses at ECMWF: Evaluation using global ground-based in situ observations
.
J. Hydrometeor.
,
13
,
1442
1460
, doi:.
Albergel
,
C.
, and Coauthors
,
2013
:
Skill and global trend analysis of soil moisture from reanalyses and microwave remote sensing
.
J. Hydrometeor.
,
14
,
1259
1277
, doi:.
Anderson
,
M. C.
,
J. M.
Norman
,
J. R.
Mecikalski
,
J. A.
Otkin
, and
W. P.
Kustas
,
2007
:
A climatological study of evapotranspiration and moisture stress across the continental United States based on thermal remote sensing: 1. Model formulation
.
J. Geophys. Res.
,
112
,
D10117
, doi:.
Argüeso
,
D.
,
J. P.
Evans
,
L.
Fita
, and
K. J.
Bormann
,
2014
:
Temperature response to future urbanization and climate change
.
Climate Dyn.
,
42
,
2183
2199
, doi:.
Argüeso
,
D.
,
J. P.
Evans
,
A. J.
Pitman
, and
A.
Di Luca
,
2015
:
Effects of city expansion on heat stress under climate change conditions
.
PLoS One
,
10
, e0117066, doi:.
Barlage
,
M.
,
S.
Miao
, and
F.
Chen
,
2016
:
Impact of physics parameterizations on high-resolution weather prediction over two Chinese megacities
.
J. Geophys. Res. Atmos.
,
121
,
4487
4498
, doi:.
Chen
,
F.
, and Coauthors
,
2011a
:
The integrated WRF/urban modelling system: Development, evaluation, and applications to urban environmental problems
.
Int. J. Climatol.
,
31
,
273
288
, doi:.
Chen
,
F.
,
S.
Miao
,
M.
Tewari
,
J. W.
Bao
, and
H.
Kusaka
,
2011b
:
A numerical study of interactions between surface forcing and sea breeze circulations and their effects on stagnation in the greater Houston area
.
J. Geophys. Res.
,
116
,
D12105
, doi:.
Cortés-Hernández
,
V. E.
,
F.
Zheng
,
J.
Evans
,
M.
Lambert
,
A.
Sharma
, and
S.
Westra
,
2016
:
Evaluating regional climate models for simulating sub-daily rainfall extremes
.
Climate Dyn.
,
47
,
1613
1628
, doi:.
Coutts
,
A. M.
,
J.
Beringer
, and
N. J.
Tapper
,
2007
:
Impact of increasing urban density on local climate: Spatial and temporal variations in the surface energy balance in Melbourne, Australia
.
J. Appl. Meteor. Climatol.
,
46
,
477
493
, doi:.
Coutts
,
A. M.
,
N. J.
Tapper
,
J.
Beringer
,
M.
Loughnan
, and
M.
Demuzere
,
2013
:
Watering our cities: The capacity for water sensitive urban design to support urban cooling and improve human thermal comfort in the Australian context
.
Prog. Phys. Geogr.
,
37
,
2
28
, doi:.
Cowan
,
T.
,
A.
Purich
,
S.
Perkins
,
A.
Pezza
,
G.
Boschat
, and
K.
Sadler
,
2014
:
More frequent, longer, and hotter heat waves for Australia in the twenty-first century
.
J. Climate
,
27
,
5851
5871
, doi:.
Dee
,
D. P.
, and Coauthors
,
2011
:
The ERA-Interim reanalysis: Configuration and performance of the data assimilation system
.
Quart. J. Roy. Meteor. Soc.
,
137
,
553
597
, doi:.
De Jeu
,
R. A. M.
,
T. R. H.
Holmes
,
R. M.
Parinussa
, and
M.
Owe
,
2014
:
A spatially coherent global soil moisture product with improved temporal resolution
.
J. Hydrol.
,
516
,
284
296
, doi:.
Dosio
,
A.
, and
P.
Paruolo
,
2011
:
Bias correction of the ENSEMBLES high-resolution climate change projections for use by impact models: Evaluation on the present climate
.
J. Geophys. Res.
,
116
,
D16106
, doi:.
Engel
,
C. B.
,
T. P.
Lane
,
M. J.
Reeder
, and
M.
Rezny
,
2013
:
The meteorology of Black Saturday
.
Quart. J. Roy. Meteor. Soc.
,
139
,
585
599
, doi:.
Evans
,
J. P.
,
M.
Ekström
, and
F.
Ji
,
2012
:
Evaluating the performance of a WRF physics ensemble over south-east Australia
.
Climate Dyn.
,
39
,
1241
1258
, doi:.
Fan
,
Y.
, and
H.
van den Dool
,
2004
:
Climate Prediction Center global monthly soil moisture data set at 0.5° resolution for 1948 to present
.
J. Geophys. Res.
,
109
,
D10102
, doi:.
Fischer
,
E. M.
,
S. I.
Seneviratne
,
P. L.
Vidale
,
D.
Lüthi
, and
C.
Schär
,
2007
:
Soil moisture–atmosphere interactions during the 2003 European summer heat wave
.
J. Climate
,
20
,
5081
5099
, doi:.
Fischer
,
E. M.
,
K. W.
Oleson
, and
D. M.
Lawrence
,
2012
:
Contrasting urban and rural heat stress responses to climate change
.
Geophys. Res. Lett.
,
39
,
L03705
, doi:.
Georgescu
,
M.
,
P. E.
Morefield
,
B. G.
Bierwagen
, and
C. P.
Weaver
,
2014
:
Urban adaptation can roll back warming of emerging megapolitan regions
.
Proc. Natl. Acad. Sci. USA
,
111
,
2909
2914
, doi:.
Gershunov
,
A.
,
D. R.
Cayan
, and
S. F.
Iacobellis
,
2009
:
The great 2006 heat wave over California and Nevada: Signal of an increasing trend
.
J. Climate
,
22
,
6181
6203
, doi:.
Grossman-Clarke
,
S.
,
J. A.
Zehnder
,
T.
Loridan
, and
C. S. B.
Grimmond
,
2010
:
Contribution of land use changes to near-surface air temperatures during recent summer extreme heat events in the Phoenix metropolitan area
.
J. Appl. Meteor. Climatol.
,
49
,
1649
1664
, doi:.
Hafner
,
J.
, and
S. Q.
Kidder
,
1999
:
Urban heat island modeling in conjunction with satellite-derived surface/soil parameters
.
J. Appl. Meteor.
,
38
,
448
465
, doi:.
Hawkins
,
E.
, and
R.
Sutton
,
2009
:
The potential to narrow uncertainty in regional climate predictions
.
Bull. Amer. Meteor. Soc.
,
90
,
1095
1107
, doi:.
Husain
,
S. Z.
,
S.
Bélair
, and
S.
Leroyer
,
2014
:
Influence of soil moisture on urban microclimate and surface-layer meteorology in Oklahoma City
.
J. Appl. Meteor. Climatol.
,
53
,
83
98
, doi:.
Jackson
,
T. L.
,
J. J.
Feddema
,
K. W.
Oleson
,
G. B.
Bonan
, and
J. T.
Bauer
,
2010
:
Parameterization of urban characteristics for global climate modeling
.
Ann. Assoc. Amer. Geogr.
,
100
,
848
865
, doi:.
Jacobson
,
M. Z.
,
1999
:
Effects of soil moisture on temperatures, winds, and pollutant concentrations in Los Angeles
.
J. Appl. Meteor.
,
38
,
607
616
, doi:.
Jones
,
D.
,
W.
Wang
, and
R.
Fawcett
,
2009
:
High-quality spatial climate data-sets for Australia
.
Aust. Meteor. Oceanogr. J.
,
58
,
233
248
, doi:.
Kala
,
J.
,
J. P.
Evans
, and
A. J.
Pitman
,
2015
:
Influence of antecedent soil moisture conditions on the synoptic meteorology of the Black Saturday bushfire event in southeast Australia
.
Quart. J. Roy. Meteor. Soc.
,
141
,
3118
3129
, doi:.
Koster
,
R. D.
, and Coauthors
,
2006
:
GLACE: The Global Land–Atmosphere Coupling Experiment. Part I: Overview
.
J. Hydrometeor.
,
7
,
590
610
, doi:.
Kusaka
,
H.
,
H.
Kondo
,
Y.
Kikegawa
, and
F.
Kimura
,
2001
:
A simple single-layer urban canopy model for atmospheric models: Comparison with multi-layer and slab models
.
Bound.-Layer Meteor.
,
101
,
329
358
, doi:.
Lee
,
S. H.
, and Coauthors
,
2011
:
Evaluation of urban surface parameterizations in the WRF Model using measurements during the Texas Air Quality Study 2006 field campaign
.
Atmos. Chem. Phys.
,
11
,
2127
2143
, doi:.
Li
,
D.
, and
E.
Bou-Zeid
,
2013
:
Synergistic interactions between urban heat islands and heat waves: The impact in cities is larger than the sum of its parts
.
J. Appl. Meteor. Climatol.
,
52
,
2051
2064
, doi:.
Lorenz
,
R.
,
E. B.
Jaeger
, and
S. I.
Seneviratne
,
2010
:
Persistence of heat waves and its link to soil moisture memory
.
Geophys. Res. Lett.
,
37
,
L09703
, doi:.
Maxwell
,
R. M.
,
F. K.
Chow
, and
S. J.
Kollet
,
2007
:
The groundwater–land-surface–atmosphere connection: Soil moisture effects on the atmospheric boundary layer in fully-coupled simulations
.
Adv. Water Resour.
,
30
,
2447
2466
, doi:.
McKenzie
,
N. J.
, and
J.
Hook
,
1992
: Interpretation of the Atlas of Australian Soils. CSIRO Division of Soils Tech. Rep. 94/1992, 23 pp.
McKenzie
,
N. J.
,
D. W.
Jacquier
,
L. J.
Ashton
, and
H. P.
Cresswell
,
2000
: Estimation of soil properties using the Atlas of Australian Soils. CSIRO Land and Water Tech. Rep. 11/00, 24 pp.
Miralles
,
D. G.
,
M. J.
van den Berg
,
A. J.
Teuling
, and
R. A. M.
de Jeu
,
2012
:
Soil moisture–temperature coupling: A multiscale observational analysis
.
Geophys. Res. Lett.
,
39
,
L21707
, doi:.
Mpelasoka
,
F.
,
K.
Hennessy
,
R.
Jones
, and
B.
Bates
,
2008
:
Comparison of suitable drought indices for climate change impacts assessment over Australia towards resource management
.
Int. J. Climatol.
,
28
,
1283
1292
, doi:.
Nicholls
,
N.
, and
S.
Larsen
,
2011
:
Impact of drought on temperature extremes in Melbourne, Australia
.
Aust. Meteor. Oceanogr. J.
,
61
,
113
116
, doi:.
Nicholls
,
N.
,
C.
Skinner
,
M.
Loughnan
, and
N.
Tapper
,
2008
:
A simple heat alert system for Melbourne, Australia
.
Int. J. Biometeor.
,
52
,
375
384
, doi:.
Parinussa
,
R. M.
,
M. T.
Yilmaz
,
M. C.
Anderson
,
C. R.
Hain
, and
R. A. M.
de Jeu
,
2014
:
An intercomparison of remotely sensed soil moisture products at various spatial scales over the Iberian Peninsula
.
Hydrol. Processes
,
28
,
4865
4876
, doi:.
Parker
,
T. J.
,
G. J.
Berry
, and
M. J.
Reeder
,
2014
:
The structure and evolution of heat waves in southeastern Australia
.
J. Climate
,
27
,
5768
5785
, doi:.
Perkins
,
S. E.
,
D.
Argüeso
, and
C. J.
White
,
2015
:
Relationships between climate variability, soil moisture, and Australian heatwaves
.
J. Geophys. Res. Atmos.
,
120
,
8144
8164
, doi:.
Pezza
,
A. B.
,
P.
van Rensch
, and
W.
Cai
,
2012
:
Severe heat waves in southern Australia: Synoptic climatology and large scale connections
.
Climate Dyn.
,
38
,
209
224
, doi:.
Piles
,
M.
,
A.
Camps
,
M.
Vall-Llossera
,
I.
Corbella
,
R.
Panciera
,
C.
Rudiger
,
Y. H.
Kerr
, and
J.
Walker
,
2011
:
Downscaling SMOS-derived soil moisture using MODIS visible/infrared data
.
IEEE Trans. Geosci. Remote Sens.
,
49
,
3156
3166
, doi:.
Piles
,
M.
,
G. P.
Petropoulos
,
N.
Sánchez
,
Á.
González-Zamora
, and
G.
Ireland
,
2016
:
Towards improved spatio-temporal resolution soil moisture retrievals from the synergy of SMOS and MSG SEVIRI spaceborne observations
.
Remote Sens. Environ.
,
180
,
403
417
, doi:.
Raupach
,
M. R.
,
P. R.
Briggs
,
V.
Haverd
,
E. A.
King
,
M.
Paget
, and
C. M.
Trudinger
,
2009
: Australian Water Availability Project (AWAP): CSIRO marine and atmospheric research component: Final report for phase 3. CAWCR Tech. Rep. 13, 67 pp. [Available online at http://www.csiro.au/awap/doc/CTR_013_online_FINAL.pdf.]
Reen
,
B. P.
,
D. R.
Stauffer
,
K. J.
Davis
, and
A. R.
Desai
,
2006
:
A case study on the effects of heterogeneous soil moisture on mesoscale boundary-layer structure in the Southern Great Plains, U.S.A. Part II: Mesoscale modelling
.
Bound.-Layer Meteor.
,
120
,
275
314
, doi:.
Salamanca
,
F.
,
A.
Martilli
, and
C.
Yagüe
,
2012
:
A numerical study of the urban heat island over Madrid during the DESIREX (2008) campaign with WRF and an evaluation of simple mitigation strategies
.
Int. J. Climatol.
,
32
,
2372
2386
, doi:.
Salomonson
,
V. V.
,
W. L.
Barnes
,
P. W.
Maymon
,
H. E.
Montgomery
, and
H.
Ostrow
,
1989
:
MODIS: Advanced facility instrument for studies of the Earth as a system
.
IEEE Trans. Geosci. Remote Sens.
,
27
,
145
153
, doi:.
Seneviratne
,
S. I.
,
T.
Corti
,
E. L.
Davin
,
M.
Hirschi
,
E. B.
Jaeger
,
I.
Lehner
,
B.
Orlowsky
, and
A. J.
Teuling
,
2010
:
Investigating soil moisture–climate interactions in a changing climate: A review
.
Earth Sci. Rev.
,
99
,
125
161
, doi:.
Sharma
,
A.
,
H. J. S.
Fernando
,
A. F.
Hamlet
,
J. J.
Hellmann
,
M.
Barlage
, and
F.
Chen
,
2017
:
Urban meteorological modeling using WRF: A sensitivity study
.
Int. J. Climatol.
,
37
,
1885
1900
, doi:.
Skamarock
,
W. C.
, and Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., doi:.
Stéfanon
,
M.
,
P.
Drobinski
,
F.
D’Andrea
, and
N.
De Noblet-Ducoudré
,
2012
:
Effects of interactive vegetation phenology on the 2003 summer heat waves
.
J. Geophys. Res.
,
117
,
D24103
, doi:.
Stegehuis
,
A. I.
,
R.
Vautard
,
P.
Ciais
,
A. J.
Teuling
,
D. G.
Miralles
, and
M.
Wild
,
2015
:
An observation-constrained multi-physics WRF ensemble for simulating European mega heat waves
.
Geosci. Model Dev.
,
8
,
2285
2298
, doi:.
Trewin
,
B. C.
,
2005
:
A notable frost hollow at Coonabarabran, New South Wales
.
Aust. Meteor. Mag.
,
54
,
15
21
.
Trewin
,
B. C.
,
2013
:
A daily homogenized temperature data set for Australia
.
Int. J. Climatol.
,
33
,
1510
1529
, doi:.
Ummenhofer
,
C. C.
,
M. H.
England
,
P. C.
Mclntosh
,
G. A.
Meyers
,
M. J.
Pook
,
J. S.
Risbey
,
A.
Sen Gupta
, and
A. S.
Taschetto
,
2009
:
What causes southeast Australia’s worst droughts?
Geophys. Res. Lett.
,
36
,
L04706
, doi:.
Victorian Government Department of Human Services
,
2009
: January 2009 heatwave in Victoria: An assessment of health impacts. Victorian Government Department of Human Services Rep., 21 pp. [Available online at https://www2.health.vic.gov.au/getfile/?sc_itemid=%7B78C32CE8-A619-47A6-8ED1-1C1D34566326%7D&title=January.]
Wan
,
Z.
,
2008
:
New refinements and validation of the MODIS land-surface temperature/emissivity products
.
Remote Sens. Environ.
,
112
,
59
74
, doi:.
Zampieri
,
M.
,
F.
D’Andrea
,
R.
Vautard
,
P.
Ciais
,
N.
De Noblet-Ducoudré
, and
P.
Yiou
,
2009
:
Hot European summers and the role of soil moisture in the propagation of Mediterranean drought
.
J. Climate
,
22
,
4747
4758
, doi:.
Zhao
,
W.
, and
A.
Li
,
2015
:
A comparison study on empirical microwave soil moisture downscaling methods based on the integration of microwave-optical/IR data on the Tibetan Plateau
.
Int. J. Remote Sens.
,
36
,
4986
5002
, doi:.

Footnotes

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).