Local and Regional Scale Evaluation of the Integrated Urban Land Model by Comparing with the Common Land Model

Land surface evaporation is not only an important parameter in natural land surface modeling, but a crucial important parameter in urban hydrology modeling. A whole-layer soil evaporation scheme was developed in the integrated urban land model (IUM) to improve the soil evaporation simulation. +e impervious surface evaporation (ISE) was used as a component of urban water balance equation. In this paper, the integrated urban land model was validated at one desert site and six urban road sites to emphasize the improvement in the evaporation simulations for arid and urban areas. A sensitivity analysis was implemented in seven basins to expand the utility of the whole layer soil evaporation scheme. For the urban road sites, the validation results indicate that imperious surface evaporation (ISE) plays a crucial role in road surface temperature (RST) simulations on rainy days. For the desert site, the validation results show that the inner layer evaporation is very important in arid regions. For the basins, the analysis results indicate that the relative monthly mean differences in the evapotranspiration (ET) between the simulations with (IUM) and without (Common Land Model (CoLM)) considering the inner layer evaporation range from −8% to 8%, which is proportional to the degree of dryness. In arid areas, especially deserts, the inner layer soil evaporation could not be neglected.


Introduction
Land surface model (LSM) is an important part in the earth system model (ESM), which is indispensable in regional and global climate change prediction.Land surface evaporation is not only an important parameter in natural land surface modeling [1], but also a crucial parameter in urban hydrology modeling [2].For arid and semiarid areas, soil evaporation is a key factor in climate change and water cycle [3], while for urban areas, the imperious surface evaporation (ISE) [4] is an important part in urban energy and water balance.
For natural land surfaces, many parameterization schemes were developed to parameterize soil evaporation [5][6][7][8][9][10][11], but most of these schemes only considered the evaporation processes in soil surface.In fact, water and energy transfer are also happened in inner soil layers especially in the semiarid and arid regions.In order to solve this issue, a whole layer soil evaporation scheme was needed.
As the development of urbanization, human activities changed the land surface characteristics drastically.ese altered the land surface radiation balance, energy balance, and water balance.Moreover, many parameters used in land surface models, such as the roughness length, albedo, permeability, etc., were changed.Traditional LSMs could not simulate urban land surface parameters accurately.To solve this problem, urban land surface models were developed.
Two main categories of models were developed for urban land surface modeling.One is the coupled method and the other is the integrated method.For the coupled method, the urban canopy model (UCM) e.g., [12,13], town energy balance (TEB) model [14], and building energy model (BEM) [15] were developed and coupled with the land surface model for the natural land surfaces.For the integrated method, e.g., [16], the urban land surface model was built directly based on the traditional land surface model.In UCM [17][18][19], the ISE depended on the precipitation, and the drying process of the impervious surface was not considered.While in TEB [14], the drainage is not considered in the water balance equation.In BEM [15], the ISE was not included at all.As an integrated model, the IUM integrates the land surface model for urban and natural land surfaces.It was dedicated to simulate land surface process in all kinds of land use and land cover (LULC) categories and in all kinds of spatial scales.Compared with other coupling land surface models, one of the important improvements of IUM is the mechanism study of evaporation, not only in natural land surface, but in urban impervious surface.In the integrated urban land model (IUM), the impervious water depth was used as the prognostic of urban water balance equation.e ISE depended on the impervious water depth since the inner soil evaporation is more important in arid areas.In this paper, the IUM was tested at one desert site and six urban sites to validate its improvements in the simulation of evaporation, which is a key parameter in land surface modeling and an important part of the improvements offered in the IUM.Besides, a sensitivity analysis was implemented in seven basins in order to expand the application of the whole layer soil evaporation.

Model Description.
e IUM [16] was developed based on the common land model (CoLM) [20].IUM integrates the land surface models for urban and natural land surfaces.For natural land surfaces, a whole layer evaporation scheme was developed to improve the simulation of evaporation especially in arid areas.For urban land surfaces, both the energy and water balance model were developed.

Whole Layer Soil Evaporation Scheme.
Soil evaporation in the CoLM is calculated as follows: where E is the soil evaporation (mm•s −1 ) and h r is the soil relative humidity.Considering the water vapor transfer within the soil, the whole layer soil evaporation in the IUM could be written as follows: where n the number of soil layers, which in the CoLM, n is 10; q is the soil specific humidity (kg•kg −1 ); z h,j is the depth of the soil layer interfaces (m); and r soil [z h,j ] is the aerodynamic resistance for evaporation at the soil layer interface (s•m −1 ).e specific humidity of each soil layers can be estimated as follows: where q satj is the saturated specific humidity (kg•kg −1 ), which is associated with the near-surface air pressure and soil temperature; Ψ j is the soil matrix potential (m), which is associated with the soil texture and the soil moisture; g is the gravity constant (m•s −2 ); R w is the gas constant for water vapor (J•Kg −1 •K −1 ); and T j is the soil temperature (K).e method to calculate the aerodynamic resistance for evaporation at the soil layer interface could be seen in Meng [16].Water vapor transfer cannot be neglected, especially when the soil surface is very dry, so it is more important in arid and desert regions.

Urban Energy Balance Model.
For urban energy balance, in contrast to UCM [12], TEB [14], and BEM [15], only the anthropogenic heat (AH) was added as a radiation source term in IUM.Based on the first law of thermodynamics and the Fourier thermal conductivity law, the energy balance equation for the impervious urban surface can be described as where c is the specific heat capacity of urban surface (J•m −3 •K −1 ); Δz is the surface layer depth (m); T r is the urban surface temperature (K); t k is the thermal conductivity of urban surface (W•m −1 •K −1 ); T 2 is the second layer soil temperature (K); ; and L v is the latent heat of evaporation for water (J•kg −1 ).e diurnal cycle of the AH is from Miao et al. [21]; this methodology is based on Sailor and Lu [22].e nighttime light data was used to expand the anthropogenic heat to regional and global scale [23].

Urban Water Balance Model.
e ISE is an important component in urban water balance.
e water depth in impervious surface is an important parameter in urban waterlogging study.In IUM, the water depth in impervious surface is considered as the prognostic variable of urban balance equation.For road surfaces, the urban water balance equation in IUM could be described as follows: where W is the water depth on impervious surface (mm); t is the time (s); E imp is the impervious surface evaporation (mm•s −1 ); P rcp is the precipitation (mm•s −1 ); and D rain is the drainage of the impervious surface (mm•s −1 ).e ISE in the CoLM is set as 0, while in the IUM it is parameterized as follows: where E p is the potential evaporation (m•s −1 ).Because surface runoff depends on topography and drainage, the drainage was evaluated based on the sensitivity analysis result in 18 road sites in Beijing during a whole summer season, which is approximately 10 mm•d −1 [24].E p can be parameterized as follows: 2 Advances in Meteorology where ρ a is the air density (kg•m −3 ); ρ w is the water density (kg•m −3 ), which is approximately equal to 1000; r d is the aerodynamic resistance for evaporation (s•m −1 ), which could be calculated as follows: where v is the von Karman constant; u * is the friction velocity (m•s −1 ); f q is the integral of profile function for moisture, which is associated with the thermodynamic roughness [25].q m is the specific humidity of the air (kg•kg −1 ); and q sat is the saturated specific humidity of the water surface (kg•kg −1 ), which is associated with the aerodynamic air temperature at the surface.e aerodynamic air temperature at the road surface is difficult to observe and simulate.In the IUM, the road water temperature is used to parameterize the potential evaporation.Road surfaces are treated as shallow lakes, and the lake model [26][27][28] in the CoLM is simplified to compute the road water temperature.e simplified shallow lake model can be described as follows: where T w is the road water temperature (K), k w is the thermal conductivity of water , and � is the net heat flux from the atmosphere (W•m −2 ), which is calculated as follows: where S n,w is the absorbed solar radiation (W•m −2 ), L n,w is the net longwave radiation (W•m −2 ), and H w is the sensible heat flux (W•m −2 ).e sensible heat flux for lake surface is calculated as follows: where c p is the specific heat of dry air (J•Kg −1 •K −1 ); T w is the lake temperature (K); θ a is the air potential temperature at reference height (K); and r ah is the aerodynamic resistance for sensible heat flux between the atmosphere at reference height and the surface (m•s −1 ), which could be calculated as follows: where f h is the integral of profile function for heat, which is associated with the thermodynamic roughness [25].

Road Sites in Beijing.
Six road weather station sites were chosen to validate the IUM.Each of these sites is located either on a major expressway that links Beijing to other cities in China or on a main road within Beijing. Figure 1 is the land cover classification of Beijing and the locations of the six road sites.e urban land use and land cover (LULC) is classified based on the 30 m resolution Landsat ematic Mapper (TM) images.LULC is classified as seven types that is high density urban, low density urban, water body, barren, forest, cropland, and grassland.e local climate zone (LCZ) [29] of Wuyuanqiao and Muxiyuan is 1 (compact high-rise); the LCZ of Bajiao, Dayangfang, Lugouqiao, and Wenyuhe is 2 (compact midrise).

Observation in HEIFE.
e desert site in Heihe Basin Field Experiment (HEIFE) [30] is used for validating the whole layer evaporation scheme.HEIFE is an international land surface observational experiment which organized by World Climate Research Programme (WCRP) and International Geosphere-Biosphere Programme (IGBP).It is the first observational experiment for arid regions with complex surface in the world.Desert meteorological station is built in the south boundary of Badanjilin desert and it is 3 km from oasis.e longitude and latitude of the station are 100.15E and 39.383 N, respectively.e land surface category of desert station is barren.[31] were chosen for comparing the surface layer and whole layer soil evaporation schemes.ese seven river basins were the Yellow River basin, the Nile River basin, the Niger River basin, the Indus River basin, the Tarim River basin, the Amu River basin, and the Syr River basin (Figure 2).e main characteristics of these seven basins are summarized in Table 1.Tarim basin is an extremely arid basin; the main LULC is barren land and accounts for more than 60% of the land surface.Amu and Syr basins are also arid basins, and the main LULC is grassland.e bare fraction of the Nile and Niger basins also exceeds 30%.

Seven River Basins. Seven river basins
e bare fraction of the Yellow and Indus basins is relatively low.

Model Configuration 2.5.1. Model Configuration for the Road Sites in Beijing.
e same as that in the CoLM, soil is also divided into 10 layers.For layers 1-5, the material was set as asphalt; the road surface roughness length was set as 0.01 m; the albedo was set as the darkest soil, which in the IUM is 0.05 for visible solar radiation and 0.1 for near-infrared solar radiation; the density, heat capacity, and heat conductivity were set as 1800 kg•m −3 , 1760000 J•m −3 •K −1 , and 0.756 W•m −1 •K −1 , respectively [32,33], and the depth of these five layers was approximately 0.212 m.For layers 6-10, the material is set as soil.Because no evaporation occurs when the impervious surface is dry, we chose a period with rainfall to perform the validation.e validation time was from the 8 th to the 12 th of July 2010.We utilized the 5minute observations of the precipitation, wind speed and direction, near-surface relative air humidity, and nearsurface air temperature data from ROSA ™ road weather stations manufactured by Vaisala Corporation to drive the IUM.e near-surface air pressure, downward solar radiation, and downward longwave radiation data are originated Advances in Meteorology from the Global Land Data Assimilation System (GLDAS) [34].e GLDAS data were interpolated temporally using the cubic spline method from 3 hrs to 5 minutes, and for the downward solar radiation the solar zenith angle was also used to ensure that there was zero downward solar radiation at night.e spatial resolution of the GLDAS data is 0.25 degrees.To validate the accuracy of the GLDAS data, it was compared with the observed data in Chinese Academy of Sciences (CAS) 325-meter-high Meteorology and Environmental Observation Tower (hereafter called 325 m tower), which is located in downtown Beijing.e longitude and latitude are 116.3708E and 39.9744 N, respectively.e

Model Configuration for Bare Soil Surface.
As for bare soil surface, the model settings of IUM are all the same as those in CoLM [20].e validation time period for desert station is from 0 : 00 1 August 1991 to 17 : 00 20 August 1991, Beijing time.e time step for validation is 30 minutes.e atmospheric forcing data are all from the observation of desert site.
e observed latent heat flux data were used for comparison; it was measured by using the eddy covariance (EC) technique.

Model Configuration for Basin
Scale.We used the reanalysis data from the GLDAS to drive the land surface model.ese data included the precipitation, near-surface air temperature, wind speed and direction, near-surface relative air humidity, near-surface air pressure data downward solar radiation, and downward longwave radiation.
LULC data were from MODIS (Figure 3).We used 0.05 °spatial resolution LULC data (MCD12C1) from 2012.First, the LULC data were transformed from the IGBP classification to the United States Geological Survey (USGS) classification scheme using the relationship listed in Table 2. en, the LULC data were upscaled to a 0.25 °spatial resolution using the mosaic scheme [16,35,36].e simulation time periods for the basin encompass the whole year of 2013.e spatial and temporal resolutions are 0.25 °and 3 hrs, respectively.

Results and Discussion
Firstly, to emphasis the importance of ISE on urban land surface modeling, we focus on the simulation improvement of road surface temperatures (RST) in six road sites.As no observational downward solar and longwave radiation data are available for these road sites, we validate the downward solar and longwave radiation data from GLDAS through the comparison with the observation in the 325 m tower.Figure 4 shows the comparison of scattered plots of the downward solar radiation and longwave radiation from observation and GLDAS.Table 3 shows the Biases, Mean errors (MEs), root mean square errors (RMSEs), and correlation coefficient (R) between them.In general, the downward shortwave and longwave radiation from GLDAS have good correlation with those of the observation.e shortwave radiation from GLDAS is larger than that of the observation; the bias is about 20 percent of total mean observation.e longwave radiation from GLDAS is smaller  than that of the observation; the bias is about 6 percent of total mean observation.e total mean bias of the downward radiation from GLDAS is about 23 W•m −2 larger than that of the observation; it approximately accounts for 14 percent of the total downward observation.Figure 5 shows the simulated ISE, observed precipitation, and simulated water depth by the IUM for the six road sites.Precipitation occurred at noon on July 8 of 2010.Waterlogging occurred starting on the afternoon of July 8 and ended on July 12, except at the Dayangfang site, where it ended on July 11.ISE is associated with potential evaporation (road water temperature), drainage, and precipitation; it is almost continuous when the water depth is larger than zero.ISE plays a crucial role in urban land modeling on rainy days.Unfortunately, no observed ISE data are available for validation.
As the observed and simulated RST for all the six sites are similar, to validate the improvement of IUM in urban areas, the mean RST simulations from the CoLM and the IUM and the observations at these six road sites are shown in Figure 6.
Only days with successive ISE were considered for comparison.Figure 6 shows that ISE plays a very important role in RST simulation on rainy days.e biases of the RST between the CoLM and the observations are about 5K at night and 30K in the daytime, while the biases of RST between the IUM and the observations are lower than 2K at night and lower than 5K in the daylight.
e physical mechanism of the improvement could be explained using the urban energy balance equation (equation ( 4)). e RST is associated with the road surface heat flux, net radiation, sensible heat flux, impervious surface evaporation, specific heat capacity, and thermal conductivity.It is inversely proportional to the impervious surface evaporation which could cool down the road surface.e deviation of downward radiation of GLDAS perhaps is another reason of the relative higher RST simulation in the daytime.But the ISE is apparently the more important reason.
en, we validate the improvement of soil evaporation in an arid site by the whole layer evaporation scheme.Figure 7 shows the scattered plots of time series resulting latent heat flux value estimates from CoLM and the whole layer parameterization scheme which used in IUM compared with the observation for the desert site during the validation time period.Table 4 shows the Biases, MEs, RMSEs, and R of the latent heat fluxes simulated by CoLM and IUM compared with those of the observations.After using the whole layer soil evaporation parameterization scheme, the biases (simulation subtract observation) were reduced from −5.06 to −4.48 W•m −2 , the MEs were reduced from 23.8 to 12.6 W•m −2 , the RMSEs were reduced from 45.3 to 25 W•m −2 , and R is increased from 0.75 to 0.80.ese indicate that the inner layer evaporation is very important in arid regions and could not be neglected for CoLM.From Figure 7, it could be noticed that under certain circumstances, the simulated soil evaporation by CoLM is zero.at means in these circumstances, soil evaporation only happens in the inner layers.
Based on the validation of the whole layer evaporation scheme in single site, a sensitivity analysis was implemented in the regional scale.We compared the simulation results of the CoLM and IUM in seven river basins.We calculated the absolute (IUM-CoLM) (Figure 8(a)) and relative (absolute differences divided by the evapotranspiration (ET) estimates from the CoLM) (Figure 8(b)) monthly mean differences of the ETs between the two models.A graduated color scheme for the different basins was used based on the annual ET (areas in red indicate the lowest ET, purple indicates the highest ET). e relative monthly mean differences in the ETs range from −8% to 8%. e absolute monthly mean differences in the ETs range from −0.025 mm•d −1 to 0.015 mm•d −1 .e monthly mean differences in the ETs are inversely proportional to the annual ET.To study the relationship between the inner soil layer evaporation and the other parameters, scatter plots were drawn comparing the absolute value of relative monthly mean ET differences and the monthly mean evaporation (Figure 9 e annual averages of the absolute value of relative monthly mean ET differences were also compared with the bare fraction (Figure 9(f )) for these seven basins.e R were also calculated, showing that the absolute value of relative monthly mean ET differences are inversely proportional to the monthly mean evaporation, runoff, volumetric soil moisture, evaporative fraction, and LAI and are proportional to the degree of the dryness.
e calculated correlation coefficients reflect the significance of these parameters to the inner soil evaporation.e most important parameter to soil inner evaporation is the volumetric soil moisture.Along with the dry of the soil surface, the inner soil evaporation plays an important role in total soil evaporation.e LAI is associated with the fractional vegetation cover, and then associated with the bare fraction and the inner soil evaporation.
e surface runoff is associated with the precipitation, infiltration, and soil water storage.When surface runoff is larger than zero, the soil is saturated, and the inner soil evaporation could be neglected.As the result, the surface runoff differences are also inversely proportional to the inner soil evaporation.e evaporative fraction is the latent heat flux, which is proportional to the soil evaporation, divided by the sensible heat flux plus the latent heat flux.It is also associated with the dryness of soil surface.
e analysis above indicates that the inner soil layer evaporation could not be neglected in arid areas, especially in deserts.Advances in Meteorology

Conclusions
In this paper, the IUM was tested at one desert site, at six road sites, and in seven basins in order to validate the simulation of evaporation from impervious surfaces and in arid regions.
ISE is associated with the potential evaporation (road water temperature), drainage, and precipitation; it is almost continuous when the water depth is greater than zero.ISE plays a crucial important role in RST simulation on rainy days.e ISE parameterization scheme is one of the key points in urban water balance and water cycle research, which is not fully researched compared with the research of the energy balance.ISE is also extremely important in RST simulation in raining days, so it is very important in urban heat island (UHI) and urban climatic effect research, especially in rainy cities. e imperious surface water depth is the key parameter in urban waterlogging application.
Inner layer soil evaporation is important and cannot be neglected in arid and semiarid areas.At extremely arid areas, such as deserts, the inner layer soil evaporation has the same order of magnitude as the surface evaporation.e area of arid regions is very large in our world, and they are vulnerable to climate change and human activities.So the whole layer evaporation will be used as an important tool in water resources, water cycle, and climate change research in arid areas.
In the future, the whole layer evaporation and the ISE parameterization scheme will be fully validated in more sites and regions and in different time periods.A global validation was implemented to evaluate the relationship between the differences in the soil evaporation simulation with the bare fraction.
e IUM will be coupled with the numerical

Figure 1 :
Figure 1: Land cover classification of downtown Beijing and the locations of six road sites.

Figure 4 :
Figure 4: Scattered plots of downward solar radiation (a) and longwave radiation (b) from observation and GLDAS.

Figure 5 :
Figure 5: Simulated (a) impervious surface latent heat flux, (b) observed precipitation, and (c) water depth for the six road sites.

Figure 7 :
Figure 7: Scattered plots of time series resulting latent heat flux value estimates from CoLM and the whole layer parameterization scheme which used in IUM compared with the observation for the desert site during the validation time period (from 0 : 00 1 August 1991 to 17 : 00 20 August 1991, Beijing time).

Figure 6 :Figure 8 :
Figure 6: (a) Time series of the mean road surface temperature simulated by CoLM and IUM compared with the observation at the six road sites.(b) Time series of the mean road surface temperature differences between the estimations from the CoLM and the IUM and the observations at the six road sites.
Figure Continued.

Table 1 :
Main characteristics of seven basins.

Table 2 :
Land use and land cover (LULC) mapping between the IGBP and the USGS classifications.

Table 4 :
Biases, mean errors (MEs), root mean square errors (RMSEs), and correlation coefficient R of the latent heat fluxes simulated by CoLM and IUM compared with those of the observations.