MAPS Special Issue: Land Atmospheric Interactions Land-Surface Scheme Validation using the Oklahoma Atmospheric Surface-Layer Instrumentation System (OASIS) Program and Oklahoma Mesonet Data: Preliminary Results J. A. Brotzge and D. Weber Center for Analysis and Prediction of Storms, Norman, Oklahoma 1. Introduction This paper describes the initial use of Mesonet and OASIS data for use in verifying the Interactions Soil Biosphere Atmosphere (ISBA) land-surface model (LSM). Previous routine, real-time measurements of surface fluxes and soil moisture were limited to field projects with short, high-intensive observational periods. Field programs such as HAPEX-MOBILY (Andre et al. 1986), the First ISLSCP Field Experiment (FIFE; Sellers et al. 1988), and MONSOON 90 (Kustas et al. 1991) provided rich data sets for advancing knowledge of land-surface interactions. However, these field experiments were limited in size and scope, and did not allow for long-term (multi-seasonal) nor large-scale (regional to statewide) estimates of surface and ground parameters. The notable difference with the present study is the completeness of the data collected in addition to the spatial and temporal extent of the new data set. As described within, the available data will allow researchers to scrutinize the numerical frameworks over a complete set of weather phenomena and improve the methods associated with land surfaces schemes. The current research efforts at the Center for Analysis and Prediction of Storms (CAPS) includes a rigorous validation of a number of physics components used by the Advanced Regional Prediction System (ARPS) described by Xue et.al. (2000a). The Oklahoma Atmospheric Surface-layer Instrumentation System (OASIS; Brotzge et al., 1999) is a unique, statewide network sponsored by the National Science Foundation which collects, quality controls and archives radiation, surface fluxes, and soil data in real-time on a continuous basis. OASIS enhanced the existing observational capabilities of the Oklahoma Mesonet to allow for continuous monitoring of the total surface energy budget. OASIS data have been collected and archived from 90 Mesonet sites statewide every 5 to 30 minutes since 1 January 2000. The primary objective of OASIS is to develop a long-term, large-scale data set of fine spatial and temporal resolution against which remote sensing algorithms and numerical models can be verified. This is the first such study to use data from OASIS for model verification. The Advanced Regional Prediction System is a three-dimensional, nonhydrostatic mesoscale model developed by the CAPS at the University of Oklahoma. The ARPS is used as both a forecasting and research tool primarily but not limited to the prediction of convective storms. An accurate land-surface scheme for use in a mesoscale model is critical to forecasting cloud formation and convection. The land surface and soil package of ARPS (Xue et al., 2001b) includes the Interactions Soil Biosphere Atmosphere (ISBA) scheme described by Noilhan and Planton (1989) and Pleim and Xiu (1995). While the ISBA scheme has been extensively tested against atmospheric data (Jacquemin and Noilhan, 1990; Noilhan et al., 1991; Bougeault et al., 1 1991; Mahfouf et al., 1995; Noilhan and Mahfouf, 1996; Xiu and Pleim, 2001), the land surface model (LSM) has not been as thoroughly tested against soil observations. In addition, several recent improvements to ISBA (Noilhan and Planton, 1996; Xiu and Pleim, 2001) have not yet been included into the ARPS scheme. This paper outlines the many challenges in developing, maintaining, and using in-situ data for model validation. Section 2 presents an overview of the Mesonet and OASIS data including the limitations and operational challenges with the collection and representativeness of in-situ data. Instrument limitations and failures, spatial representativeness, and inconsistent closure of the surface energy budget each contribute to measurement error and uncertainty. In addition, model sensitivity to initialization and surface heterogeneity makes model verification even more difficult. Each of these problems is discussed, and preliminary results from model validation of ARPS using OASIS and Mesonet data are provided in Section 3. 2. Mesonet and OASIS Data Overview The Oklahoma Mesonet (Brock et al., 1995) provides the infrastructure upon which the OASIS is built. Approximately 90 Mesonet sites have been enhanced to allow net radiation, ground and sensible heat fluxes to be estimated directly (Fig. 1). Latent heat flux is estimated as the residual of the surface energy budget. Skin temperature and soil moisture also are measured directly at the 90 sites. Ten of the 90 OASIS sites, termed “super sites”, have additional sonic anemometry and 4-component net radiometers, which allow more accurate and precise measurements of net radiation and sensible and latent heat fluxes. The ten super sites are equipped with redundant instrumentation and multiple methods of measurement to allow for improved quality control of the data. A super site has been selected in each of Oklahoma’s nine climatic zones, to provide a diverse yet simultaneous set of observations from across the state. 2.1 Instrumentation and measurement procedures The measurement methods used by OASIS are summarized in Table 1. A detailed description of the sensors is found in Brotzge (2000). What follows is a brief summary of the assumptions and unique problems associated with each measurement method. This review is used foremost to develop an approximate error estimate associated with the data set. Second, problems in measurement are important to be understood in context when used for comparison against modeled data. As discussed by Noilhan et al. (1991), model results most likely will not exactly reproduce the surface observations, either because of nonrepresentativeness of the measurements or model deficiencies. Thus, a correct interpretation of the observations and modeled data require a complete understanding of both observational and modeling deficiencies. 2.1.1 Net radiation Shuttleworth (1991) describes net radiation as one of the most difficult parameters to measure accurately. For this reason, net and four-component radiometers were installed at the OASIS super sites. The net radiometer used by OASIS is the NR-Lite, a recently developed sensor, which is relatively low-cost, is nearly maintenance-free and does not require the use of polyethylene domes which can degrade over time. However, the simplicity of the design permits 2 greater sensitivity to operational errors. Brotzge and Duchon (2000) identified several limitations of the sensor. These problems include poor initial calibration, and a degradation of performance during high winds and low sun angles, and during precipitation. The four-component radiometer, the CNR1, comprises four separate sensors housed in a single unit, allowing incoming and outgoing shortwave and longwave radiation to be directly observed. A heating element, embedded within the body of the sensor, is activated during certain weather periods to minimize the effects of dew, frost or precipitation. The sensor is much more expensive than the NR-Lite but does not incur the operational problems associated with wind or precipitation. Both radiation sensors are mounted at a height of 2 m. Data from the NR-Lite and CNR1 were collected and archived from the ten super sites (Brotzge, 2000). Comparisons between the co-located sensors revealed minimal differences between them (< 5%). Multiple sensors at a site proved to be valuable in identifying a failed sensor or poor data. 2.1.2 Ground heat flux A combination approach (Tanner 1960) is used to estimate the total ground heat flux and includes separate estimates for the ground flux and storage terms: dT dT G Cdz2 dt dz where [W (m K)-1] is the thermal conductivity, dT [K] is the temperature difference across the plate, dz [m] is the plate thickness, C [J (m3 K)-1] is the soil heat capacity, dz2 [m] is the depth of the soil layer, and dT/dt is the temporal rate of change in the integrated soil temperature between 0 and 5 cm (Fritschen and Gay, 1979). Two Platinum Resistance Temperature Detectors (PRTDs) and two heat flux plates are installed at each of the 90 sites at a depth of 5 cm. An average of the 2 PRTDs is used to estimate dT/dt; likewise the mean of the two flux plates are used to estimate the first term. The soil heat capacity is estimated at each site as a function of the measured volume fraction of minerals, organic material, and soil moisture. For more details, see Brotzge (2000). Massman (1993) identified several assumptions used in estimating ground heat flux. The specific heat capacity of the soil must be assumed constant in depth and in time, and horizontal heat flow is neglected. The thermal conductivity of the ground flux plate must match that of the soil (Fritschen and Gay, 1979). Because the conductivity of the plates generally do not match that of the soil, Fritschen and Simpson (1989) developed a correction for the soil thermal conductivity as a function of plate conductivity, size, and shape, which has been applied to this data set. Finally, the flux plates themselves may impede the vertical flow of heat and moisture within the soil (Tanner, 1960; van Loon et al., 1998). Air gaps between the soil and plates also can lead to significant measurement errors (Fritschen and Gay, 1979). Significant errors are more likely to occur from large variability in ground flux properties and surface heterogeneity than from instrument error. The effects of heterogeneity are described in greater detail in section 2.2.1, however, significant variability has been observed in soil moisture and soil properties within a relatively small (20 x 20 m2) area (Basara, 2001). A comparison study conducted by Brotzge (2000) revealed differences of nearly 100 Wm -2 between two sets of co-located ground flux measurements. These two sets of measurements 3 were located approximately 100 m apart. Thus, while instrument errors generally are limited to less than 5%, errors due to surface heterogeneity can lead to much greater uncertainty. Using the mean from two heat flux plates and the mean from two PRTDs reduces the error in ground heat flux from spatial heterogeneity. However, the exact magnitude of the error remains unknown. 2.1.3 Sensible and latent heat fluxes The ten super sites also have been equipped with sonic anemometry to directly measure the sensible heat flux, in part to verify the gradient method used by the standard sites. The sonic anemometer is installed at a height of 4.5 m and is mounted to the west of the tower. Data from the sensor cannot be used during or immediately following precipitation. Foken and Wichura (1996) have listed a host of sensor configuration and meteorological problems that can occur when using an eddy correlation technique. For this study, sonic measurements have been corrected for moisture dependence (Schotanus et al., 1983; Stull, 1988) and for tilting of the sensor (T. Horst, personal communication 1999). Each of the 90 standard OASIS sites is equipped with similar cup anemometers and thermistors to monitor vertical gradients in wind and temperature. The vertical gradients of heat and momentum are applied to Monin-Obukhov similarity theory to derive an estimate of sensible heat flux (Brotzge and Crawford, 2000). Brotzge and Crawford identified three major difficulties in estimating gradient fluxes using the Mesonet infrastructure. First, the method is extremely sensitive to instrument errors. Second, the temperature sensors are only naturally aspirated, meaning that during high solar radiation, low wind conditions, radiational heating of the sensors creates a bias in temperature measurements. Third, nearby trees and topography create subtle fetch problems, thus biasing flux estimates. McAloon et al. (2000) and McAloon (2001) have found that the gradient technique is much more reliable during unstable and neutral conditions, and have identified those Mesonet sites where fetch is not a problem. Furthermore, McAloon (2001) has found a significant sensitivity (> 100 Wm-2) to the theoretical constants used. Latent heat flux is measured directly at the super sites using the sonic anemometer and a krypton hygrometer, mounted 10 cm below the sonic anemometer. The latent heat flux measurement has been corrected for oxygen absorption (Tanner et al., 1993), density fluctuations (Webb et al., 1980), and for sonic tilt. Latent heat flux is estimated as the residual of the energy balance (Rn – G – H = LE) at the standard sites. An estimate of the uncertainty of the sonic-derived flux measurements is approximately 5% for the sensible heat flux and 10% for latent heat flux. The measurement uncertainty of the gradient method varies widely according to site fetch and atmospheric stability. 2.1.4 Soil moisture and temperature Soil matric potential is measured at the 90 OASIS sites at the depths of 5, 25, 60, and 75 cm where possible. The soil matric potential is measured using 229-L heat dissipation sensors. The soil water potential is converted to soil water content, based upon soil properties (Basara, 1998). Basara and Crawford (2000) identified a problem with preferential water flow seeping down along the sensor cables to the lowest depths after some heavy rain events. Nevertheless, these instances were rare and were quality controlled. Soil temperatures also were measured from the 229L sensors located at the depths of 5, 25, 60, and 75 cm. 4 The surface skin temperature is measured at all sites using infrared sensors mounted at an angle of 45 and installed at height of 2 m. Recent quality control of the data has shown that dust and spider webs can cover the optical sensor, reducing the accuracy of the measurement. 2.1.5 Quality control All data were corrected and quality controlled as described by Brotzge et al. (1999). Standard Mesonet data were further quality assured by a series of automated and manual checks (Shafer et al., 2000). Redundant instrumentation allowed most missing flux data to be replaced; 4-component net radiation estimates from the CNR1 were replaced by data from the NR-Lite when needed, and sensible heat flux estimates from the sonic anemometer were replaced by gradient profile estimates. When any one of the four components of the energy budget were missing, the residual was used. 2.2 Challenges associated with model verification 2.2.1 Surface heterogeneity A major problem in evaluating model performance when using single site, in-situ data is the impact of surface heterogeneity. The natural variability in topography, vegetation, and soils decreases the representativeness of an observation. The degree of this variability is dependent upon the measurement type. Atmospheric parameters are well mixed due to advection and turbulent mixing and exhibit significant correlations across large-scale regions. Surface and soil parameters such as vegetation, soil moisture and soil type, exhibit much less correlation across a pixel that represents several hundred square meters. Basara (2001) quantified the variability of soil moisture within a 20 m x 20 m area during the summer of 1999. Over 2,700 samples were collected between 1 June and 12 August at 12 random locations within the sample area up to a depth of 80 cm. Basara found the greatest variability nearest the surface immediately following rainfall events; the greatest homogeneity was observed during extended dry periods (Fig. 2). Next, Basara (2001) quantified the impact of the observed soil moisture and soil texture variability upon modeled fluxes as estimated by the Oregon State University (OSU) 1-D PBL scheme used by the operational ETA model. Basara found a significant impact; for example, a change in soil texture from clay loam to silt loam increased latent heat fluxes by over 350 W m-2. Ek and Cuenca (1994) and Cuenca et al. (1996) examined the effects of modeled changes in soil water content and soil texture on the OSU PBL scheme and found similar results. If natural small-scale variability exists across a large-scale region, then data from a single site cannot be used for validation of a model grid area. To verify which variables could be used for model validation, Brotzge and Richardson (2002) estimated the spatial correlation among all 90 OASIS sites for most measured parameters. Data archived from all of 2000 were included in the study, which comprised approximately 15 million observations. The results, summarized by Figures 3a and 3b, showed that atmospheric parameters were much greater correlated than were surface and soil variables. Nevertheless, all variables showed decreasing correlation with decreasing distance even among those variables influenced by surface heterogeneity (Fig. 3b). This heterogeneity is expected to vary with season, however, as stratiform rains give way to convective rainfall during the spring and summer periods (Fig. 4), increasing the spatial 5 heterogeneity in soil moisture. Overall, the results from Brotzge and Richardson agree with Shuttleworth (1991), that single point, in-situ data can be used for both calibration and diagnosis of a numerical model if the data represent a large, uniform region and are collected over an extended period of time. 2.2.2 Closure of the surface energy balance A second major problem with measurement of near-surface parameters is the difficulty in closure of the surface energy balance. The four components of the surface energy balance rarely are found to be closed (i.e., Rn – H – LE – G = 0) when measured independently (Nie et al., 1992; Stannard et al., 1994; Twine et al., 2000). Such data cannot be used as “truth” against which model data is to be compared. Brotzge (2000) quantified the closure of OASIS data at the ten super sites and found significant annual variations in closure which varied with the magnitude of the latent heat flux. Closure was observed closest to 100% during extended clear, dry periods. Diurnal cycles indicated some nonclosure during the early morning hours. Examination of all data indicated no significant problems with measurements of net radiation or sensible heat fluxes; ground heat fluxes were in general too small to account for significant nonclosure. For this model evaluation study, latent heat flux was estimated as a residual to ensure closure of the surface energy budget. 3. Verification of the ISBA scheme A series of controlled runs were made with ARPS in a 1-D vertical column mode. Stretching was applied to the 33 levels of the vertical column, and the 1.5 turbulent kinetic energy scheme by Sun and Chang (1986) was used. Model runs were initialized using observed NWS soundings, Mesonet atmospheric data, and soil moisture and radiation data from OASIS. Vegetation and soil parameters were initialized to match estimated values from the observation site. The model was then forced every 10 minutes with observed net and solar radiation, air temperature, and relative humidity. . This configuration allowed only the soil model variables to vary during the simulation. These tests represent the first step in a series of tests aimed at validating the soil and boundary layer schemes in the ARPS. A comprehensive data set was compiled for this study from one year of OASIS data archived and collected from a single OASIS super site located at Norman, Oklahoma (NORM). NORM (Lat. 35 15’ 20”; Lon. 97 29’ 0” ) is characterized by flat terrain (~ 0.0 slope), and short grasses. Standard atmospheric Mesonet and OASIS flux data were collected every 5 minutes; soil data were collected every 30 minutes. Atmospheric data included air temperature, relative humidity, atmospheric pressure, rainfall, and wind speed and direction. Surface parameters (what is a parpameter?) included incoming and outgoing shortwave and longwave radiation, sensible and latent heat fluxes, ground heat flux, skin temperature, and soil moisture and soil temperatures at 5, 25, 60, and 75 cm depths. Snowfall depths were included into the data set as estimated at a nearby National Weather Service (NWS) office (approximately 3.03 km distant). Vertical soundings from 00 and 12 UTC from the NWS were archived and collected as well. Vegetation data were obtained from bi-weekly NDVI values from which the appropriate 1 km x 1 km pixel was extracted to coincide with the site location. These estimates were interpolated to daily values. Soil information for the Norman site was provided by the Oklahoma Climatological Survey. 6 Several clear days were chosen from the year data set to provide a precursory review of the performance of the ISBA scheme. Data from 20 May, 2000, represented a synopticallyquiescent spring day characterized by warm temperatures (maximum temperature of 28 C), a moderately moist soil and vigorous vegetation growth (NDVI = 0.61). High pressure also dominated during the 1 – 3 August, 2000, collection period. This data represented a dry-down period with similar soil wetness but hot air temperatures (maximum temperature of 37 C) and stressed vegetation (NDVI = 0.5). 3.1 The ISBA land-surface scheme The ISBA scheme is a “force-restore” method (Deardorff, 1978) and for simplicity is limited to five prognostic equations for soil temperature and moisture (Noilhan and Planton, 1989). These five prognostic equations are: Ts 2 CT (Rn LE H) T T t s 2 (1) T2 1 T T t s 2 (2) wg C C 1 Pg Eg 2 wg wgeq t w d1 (3) w2 1 P Eg Etr t w d2 g (4) Wr vegP Ev Etr Rr t (5) where the five prognostic variables include the surface skin temperature (Ts, K), the mean-layer soil temperature (T2, K), the canopy wetness (Wr), the soil surface wetness (Wg), and the meanlayer soil moisture (W2). These time-dependent parameters are forced by the net radiation (Rn), precipitation (P), bare soil (Eg) and evapotranspiration (Etr) in the form of latent heat flux (LE), sensible heat flux (H), and surface runoff (R). The surface temperature and moisture are restored to equilibrium by heat and moisture sources from the soil layers below. The time scale at which these variables act are prescribed a priori in the form of a time constant (set to one day) and the soil-layer depths (d1 = 0.1 m; d2 = 0.9 m). Furthermore, the interaction between the soil and atmosphere varies as a function of the fractional vegetation cover (veg), vegetation type, soil type, heat capacity, and thermal conductivity. These properties are specified by the diagnostic variables C1, C2, and CT. These five equations provide memory from the land surface to the atmospheric system. 3.2 Model sensitivity 3.2.1 Vegetation cover 7 Jacquemin and Noilhan (1990) conducted a detailed sensitivity analysis of the ISBA scheme. They quantified the influence of soil texture, moisture, and vegetation cover upon the progged values of surface fluxes. In summary, Jacquemin and Noilhan found that for daily time scales, vegetation cover (veg) was the most sensitive parameter; initialization of the deep-layer soil moisture, w2, was most important for longer time scales. For a mesoscale model such as ARPS, proper characterization of the vegetation varies as a function of grid resolution. Currently in ARPS, vegetation cover is specified as a function of vegetation type, which remains constant throughout the year. Nevertheless, vegetation cover cannot be readily known. Likewise, estimation of vegetation cover from an observational site is also rather subjective. From the OASIS site at Norman, a vegetation cover of 75% was estimated visually. Due to the uncertainty in this term and its importance to the ISBA scheme, model data from 20 May were examined by varying the vegetation cover (Fig. 5). Observed NDVI was 0.61. The vegetation cover was varied from 0% to 100%. The observed soil water content (m3 m-3), plotted in bold, was nearly constant during the 24 hour period. Model results show significant (> 0.10) diurnal oscillations with fractional bare soil; only with a complete vegetation cover do model results approach those of the observations. Thus, even a visual estimate at the site may not lead to the most correct model results. The diurnal cycle of the observed soil water content at 5 cm was opposite in phase to model values, and was counter-intuitive to what might have been expected. Basara (1998) gave an explanation of this rather unexpected diurnal cycle. A combination of thermal and potential gradients act to moisten the 5 cm soil layer during afternoon heating and dry the layer during nocturnal cooling. A potential gradient is induced as upward diffusion balances surface evaporation; the upward diffusion moistens the 5 cm layer during the period of greatest surface heating (at midday). Likewise, at night water is diffused upward from the 5 cm level to the surface, drying the 5 cm layer. Thermal gradients induce a similar pattern in soil moisture. According to Basara, under certain conditions water flows from regions of high to low temperature. Thus, during the afternoon hours of greatest surface heating, water flows from the skin surface down to the cooler 5 cm depth; during the evening, water flow reverses from the 5 cm depth upward to the cooler skin surface. A significant problem in evaluating the surface soil moisture parameter, wg, is that the soil moisture dynamics changes significantly during a diurnal period over the specified penetration depth of 10 cm. Deardorff (1978) originally defined wg as the soil moisture within the top few millimeters of the soil layer. He verified model estimates of wg using observed values collected from the upper 0.5 cm of soil. Therefore, estimates of wg as defined by the ISBA scheme should not necessarily match those observed by the OASIS sensors installed at a depth of 5 cm (see section 3.2.3). 3.2.2 Surface roughness Surface roughness is another input parameter specified a priori by the ISBA scheme as a function of surface type. Actual estimates of surface roughness from the Norman site are not known. Jacquemin and Noilhan (1990) found that variations in surface roughness changed estimates of latent heat flux by less than 10%. Again, because observed values of surface roughness were not known for the site, model sensitivity was tested by computing the top-layer (skin) soil temperature. Skin temperature was estimated by varying surface roughness from 0.01 m to 0.60 m (Fig. 6). Skin temperature data from 20 May varied by as much as 9 C as a function of the roughness length. However, model results matched daytime observations of skin 8 temperature best when a surface roughness of 0.01 m was used. An expected roughness of 0.005 m < zo < 0.05 m was estimated visually from the Norman site (Garratt, 1992). NORM is characterized by flat terrain and short grasses. Nevertheless, model estimates of skin temperature were > 9 C less than those observed. Such differences were likely due to the sensitivity of the soil and vegetation type chosen as representative of the site. A second set of model runs (not shown) assumed a soil type of silty clay. These results showed model estimates of skin temperature during mid-afternoon to within one degree Celsius with a zo of 0.01 m. The sensitivity of soil type to top-layer soil moisture and temperature is examined greater detail in section 3.2.3. The time lag observed at sunrise between the observed and model values are discussed in section 3.4. 3.2.3 Soil type Soil type dictates the ability of the soil to transfer heat and moisture throughout the soil column. The current version of ISBA used by ARPS assigns several soil parameters including the saturated, wilting point, and field capacity volumetric water contents (m3 m-3) and the soil thermal coefficient at saturation (K m2 J-1) as a function of a single soil type. Noilhan and Mahfouf (1996) suggested an improvement to the scheme by recommending that all hydraulic parameters be estimated as a continuous function of the ratio of sand and clay present in the soil. Before implementation of the Noilhan and Mahfouf (1996) recommendation into ARPS, known soil values from the Norman Mesonet site were used to quantify this change. Data from NORM provided an insightful comparison because the soil type changed with depth; the ISBA scheme does not allow soil type to vary with depth. The fractional soil type was measured from each of the four soil moisture measurement depths (5, 25, 60, 75 cm) at the Mesonet site. At the 5 cm level, the fraction of clay and sand in the soil was 25% and 15%, respectively; at the 25 cm level and below, the fraction of clay and sand was approximately 40% and 15%. These two estimates of soil type were tested against discrete estimates of soil type. The upper 15 cm of the soil column was classified as silt loam; the soil was largely clay loam below 15 cm. The influence of the treatment of these hydraulic parameters upon model estimates of surface soil moisture was determined by comparison against observations (Fig. 7a and 7b). The impact of soil type was significant upon both surface soil moisture and temperature. A change in soil type altered the heat and moisture capacities of the soil layer. As shown in Figure 7a, the current scheme, which uses a discrete soil type, indicates a similar behavior in soil moisture to that observed at the 5 cm depth. However, an examination of the top-layer soil temperature shows a significant (> 6 C) underestimation in temperature by the discrete soil type. On the other hand, the continuous formulations underestimated (overestimated) daytime (nighttime) soil moisture, but forecasted daytime skin temperatures within several degrees. Such differences could have a strong impact on daily forecasted minimum air temperatures and PBL development. Most likely, the surface soil moisture, wg, is best represented by the continuous formulations which show large diurnal variability in the wg term. As discussed in section 3.2.1., the top-layer is generally representative of the top few millimeters; the observation of soil moisture is taken at a depth of 5 cm. A comparison of the top-layer soil temperature is likely the best estimate of the performance of the model. 3.3 Time constant, 9 A time constant, , was chosen in the force-restore scheme equal to one day to represent a typical diurnal cycle (Deardorff, 1978). However, the same time constant is used to represent both the short-term variability of the skin temperature, Ts, and the surface soil moisture, wg, as well as the more slowly varying mean-layer soil temperature, T2. Xiu and Pleim (2001) recommended setting = 10 days to improve seasonal variability. Mean-layer soil temperatures were estimated from ARPS as a function of and plotted against observed data collected from a depth of 75 cm during 1 – 4 August (Fig. 8). These data showed that a longer time constant improved the mean-layer soil temperature estimate by eliminating the diurnal signal. To maintain the diurnal signal in the restoration term of the skin surface and surface moisture equations, a time constant of 10 days cannot be used. Instead, Equation (1) was replaced with: Ts (6) CT G Ts T2 t D where [W m-1 K-1] is the soil thermal conductivity and D is a damping coefficient [m], as described by Dickinson (1988). The damping coefficient represents a penetration depth and varies as a function of thermal conductivity and specific heat capacity. Because D is a function of soil type and moisture, the coefficient may vary with time and space. Results from this modification were tested for both the 20 May (Fig. 9a) and 2 August days (Fig. 9b). Skin temperature was calculated according to Eq. (6) by varying D from 0 m to 1 m. Daytime temperatures varied little among model values and were much cooler than observations. Model estimates differed greatest during the night and early morning hours. Results from the 20 May case showed that a D ~ 0.40 m was closest to observations. Results from the 2 August data indicated a much smaller damping coefficient of D ~ 0.2 m or less. These results indicated a much greater penetration depth during May due to wetter soil and greater thermal conductivity. Likewise, during August a very dry soil limited the thermal conductivity and which lead to a much shallower damping coefficient. 3.4 Delay in model response An examination of Figures 6, 7b, 9, 10, and 11 revealed another significant problem with modeling of the surface skin temperature and surface fluxes by the current ISBA scheme as used within ARPS. Both the 20 May and 1 - 2 August data showed a significantly lagged (> 1 hour) model response to the rapid surface heating of the soil at sunrise. Neither changes to the time response function nor damping coefficient increased this response to early morning heating. Model and observed sensible and latent heat fluxes for 20 May and 2 August are plotted in Figures 10a,b and Figures 11a,b, respectively. These plots of the aerodynamic fluxes also indicated the sluggishness of the model fluxes to morning radiation. Not until after the net radiation increased to greater than zero did the model fluxes respond. The lack of diffuse energy in the radiation code of ARPS prevented the immediate response to the sunrise forcing. The differences between the observed and model (discrete soil type) surface fluxes during 20 May (Figs. 10) and 2 August (Figs. 11) were due to an overestimation of the top-layer soil moisture by the LSM. This overstimate in soil moisture lead to a significant underestimate in skin temperature (Fig. 7b, 9). This underestimate in skin temperature lead to a decreased air-soil temperature gradient, and a decreased sensible heat flux. Likewise, the latent heat fluxes were overestimated due to the overestimated top-layer soil moisture. 10 The model results improved when the continous soil formulations were used. As shown in Figures 7a,b, a drier top-layer is warmer, leading to a greater air-soil temperature gradient, which lead to greater daytime sensible heat flux and lower latent heat flux. These model estimates were much closer to those observed. Note that this configuration allowed only the soil model variables to vary during the simulation. These tests represent the first step in a series of tests aimed at validating the soil and boundary layer schemes in the ARPS. 4. Conclusions A proper evaluation of the ISBA scheme involved a thorough review of the validation data set and the model results. A detailed examination of the data set permitted a preliminary evaluation of the current ISBA scheme used within ARPS. The principal results of our investigation follow. 1.) The sensitivity of the ISBA scheme to input parameters is detrimental to a microscale model such as ARPS due to the uncertainty which exists in choosing the surface parameters. Furthermore, additional uncertainty exists due to the surface heterogeneity within each grid area represented by the input parameter. As demonstrated in Figures 5 and 6, the impact of vegetation cover and surface roughness is significant (> 13 C in Fig. 6). 2.) Changes in the classification of soil type significantly alter the soil heat capacity and conductivity (Fig. 7). A continous formulation proposed by Noilhan and Mahfouf (1996) lead to noticeable differences in soil temperature and moisture (Fig. 7a and 7b), particularly when radiative forcing was minimal. 3.) The observations of soil texture indicated decreased clay fraction with depth. The simplicity of the ISBA scheme does not allow for variations of soil texture with depth. The impact of this assumption was quantified by the 20 May 2000 example (Fig. 7a and 7b). Modeled estimates of skin temperature varied less than 0.5 C, however, the top-layer soil water content varied by as much as 0.4 m3 m-3 during the evening and early morning hours. 4.) This study verified the change in from 1 day to 10 days, as originally proposed by Xiu and Pleim (2001). As shown in Figure 8, the observed mean-layer soil temperature exhibited little diurnal variability; the model results improved by setting equal to 10 days, which eliminated the model-imposed diurnal cycle. 5.) To maintain the diurnal signal in the restoration term of the skin surface, a time constant of 10 days could not be used. Instead, the term was replaced as a function of soil conductivity and a damping coefficient. Results indicated model values of skin temperature improved compared to the observed estimates, but varied according to the chosen damping coefficient (Fig. 9). The damping coefficient varied with soil wetness. 6.) A delay in the model response of the sensible and latent heat fluxes was detected when compared against the observations of May (Fig. 10) and August (Fig. 11). The most likely cause of such a lag is the lack of diffuse radiation within the model. 11 This work will be expanded in the near future to quantify differences between the model and observed fluxes on a seasonal and annual basis. In addition, a variety of vegetation and soil types will be tested using data from other OASIS sites across Oklahoma. Furthermore, the spatial interaction of surface fluxes, soil wetness, and vegetation will be examined across regional scales using the 90 standard flux sites. Such validation of the LSM against observed surface variables is anticipated to lead to further improvements in the ISBA scheme, and which will ultimately lead to improved prediction of air surface temperature and moisture within ARPS. Acknowledgments. This research was made possible in part by NSF Grant 125-5645 and the Williams Energy Corporation Grant xxx-xxxx REFERENCES Andre et al. 1986: HAPEX MOBILHY data Basara JB (2001) The value of point-scale measurements of soil moisture in planetary boundary layer simulations. PhD dissertation, School of Meteorology, University of Oklahoma, Norman. pp 225 Basara JB, Crawford TM (2000) Improved installation procedures for deep-layer soil moisture measurements. J. Atmos. Oceanic Technol 17: 879 – 884 Bougeault P, Noilhan J, Lacarrere P, Mascart P (1991) An experiment with an advanced surface parameterization in a mesobeta-scale model. Part I: Implementation. Mon Wea Rev 119: 2358 2373 Brock FV, Crawford KC, Elliott RL, Cuperus GW, Stadler SJ, Johnson HL, Eilts MD (1995) The Oklahoma Mesonet: a technical overview. J Atmos Oceanic Technol 12: 5-19 Brotzge JA (2000) Closure of the surface energy budget. PhD dissertation, School of Meteorology, University of Oklahoma, Norman. pp 208 Brotzge JA, Crawford KC (2000) Estimating sensible heat flux from the Oklahoma Mesonet. J Appl Meteor 39: 102 – 116 Brotzge JA, Duchon CE (2000) A field comparison among a domeless net radiometer, two 4component net radiometers, and a domed net radiometer. J Atmos Oceanic Technol 17: 1569 – 1582 Brotzge JA, Richardson SJ (2002) Spatial and temporal correlation among Oklahoma Mesonet and OASIS surface-layer measurements. Preprint, Hydrology Conf., Amer Meteor Soc. 12 Brotzge JA, Richardson SJ, Crawford KC, Horst TW, Brock FV, Humes KS, Sorbjan Z, Elliott RL (1999) The Oklahoma Atmospheric Surface-layer Instrumentation System (OASIS) Project. Preprint, Thirteenth Conf on Bound-Lay Turb. Dallas, TX, Amer Meteor Soc. Cuenca RH, Ek M, Mahrt L (1996) Impact of soil water property parameterization on atmsopheric boundary layer simulation. J Geophys Res 101: 7269-7277 Dickinson RE (1988) The force-restore model for surface temperatures and its generalizations. J Climate 1, 1086 – 1097 Ek M, and Cuenca RH (1994) Variation in soil parameters: Implications for modeling surface fluxes and atmospheric boundary-layer development. Bound.-Layer Meteor 70: 369-383 Foken T, Wichura B (1996) Tools for quality assessment of surface-based flux measurements. Ag and For Meteor 78: 83-105 Fritschen LJ, Gay LW (1979) Environmental Instrumentation. Springer-Verlag, 216 pp Fritschen LJ, Simpson JR (1989) Surface energy and radiation balance systems: general description and improvements. J Appl Meteor 28: 680 – 689 Jacquemin, B, Noilhan J (1990) Sensitivity study and validation of a land surface parameterization using the HAPEX-MOBILHY data set. Bound-Layer Meteorol 52: 93-134 Kustas WP, Goodrich DC, Moran MS, Amer SA, Bach LB, Blanford JH, Chehbouni A, Claassen H, Clements WE, Doraiswamy PC, Dubois P, Clarke TR, Daughtry CST, Gellman DI, Grant TA, Hipps LE, Huete AR, Humes KS, Jackson TJ, Keefer TO, Nichols WD, Parry R, Perry EM, Pinker RT, PJ Pinter, Jr., Qi J, Riggs AC, Schmugge TJ, Shutko AM, Stannard DI, Swiatek E, van Leeuwen JD, van Zyl J, Vidal A, Washburne J, Weltz MA (1991) An interdisciplinary field study of the energy and water fluxes in the atmosphere-biosphere system over semiarid rangelands: description and some preliminary results. Bull Amer Meteor Soc 72: 1683-1705. Mahfouf J-F, Manzi AO, Noilhan J, Giordani H, Deque M (1995) The land surface scheme within the Meteo-France climate model ARPEGE. Part I: Implementation and preliminary results. J Climate 8: 2039 - 2057 Massman WJ (1993) Errors associated with the combination method for estimating soil heat flux. Soil Sci Soc Am J 57: 1198 - 1202 McAloon C (2001) An examination of sensible heat flux using the gradient-profile technique at ten Oklahoma Mesonet sites. MS thesis, University of Oklahoma, 175 pp McAloon, C, Richardson SJ, Brotzge JA, Horst TW (2000) Preliminary results from the OASIS2000 Field Project. Preprint, Fourteenth Conf on Bound-Layer Turbul Amer Met Soc, Aspen, CO. 13 Nie, D, ET Kanemasu, LJ Fritschen, HL Weaver, EA Smith, SB Verma, RT Field, WP Kustas, and JB Stewart (1992) An intercomparison of surface energy flux measurement systems used during FIFE 1987. J Geophys Res 97: 18715-18724 Noilhan J, Mahfouf J-F (1996) The ISBA land surface parameterization scheme. Global and Planet Change 13: 145-149 Noilhan J, Planton S (1989) A simple parameterization of land surface processes for meteorological models. Mon Wea Rev 117: 536-549 Noilhan J, Lacarrere P, Bougeault P (1991) An experiment with an advanced surface parameterization in a meso-B model. Part III: Comparison with the HAPEX-MOBILHY data set. Mon Wea Rev 119: 2392-2413 Paulson CA (1970) The mathematical representation of wind speed and temperature profiles in the unstable atmospheric surface layer. J Appl Meteor 9: 857 – 861 Pleim JE, Xiu A, (1995) Development and testing of a surface flux and planetary boundary layer model for application in mesoscale models. J Appl Meteor 34: 16-32 Schotanus P, Nieuwstadt FTM, De Bruin HAR (1983) Temperature measurements with a sonic anemometer and its application to heat and moisture fluxes. Bound-Layer Meteor 26: 81-93 Shafer MA, Fiebrich CA, DS Arndt, SE Fredrickson, and TW Hughes (2000) Quality assurance procedures in the Oklahoma Mesonetwork. J Atmos Oceanic Technol 17: 474 – 494 Shuttleworth WJ (1991) Insight from large-scale observational studies of land/atmosphere interactions. Surveys in Geophysics 12: 3-20 Stannard DI, Blanford JH, Kustas WP, Nichols WD, Amer SA, Schmugge TJ, Weltz MA (1994) Interpretation of surface flux measurements in heterogeneous terrain during the Monsoon ‘90 experiment. Water Res Res 30: 1227-1239 Stull RB (1988) An introduction to boundary layer meteorology. Kluwer Academic Publishers, Boston, MA, 666 pp Tanner CB (1960) Energy balance approach to evapotranspiration from crops. S Sci Soc of Amer Proc 24: 1-9 Tanner CB, Swiatek E, Greene JP (1993) Density fluctuations and use of the Krypton hygrometer in surface flux measurements. Irrigation and drainage systems: 945-950 Twine TE, Kustas WP, Norman JM, Cook DR, Houser PR, Meyers TP, Prueger JH, Starks PJ (2000) Correcting eddy covariance flux underestimates over a grassland. Agric and For Meteor 103: 279-300 14 van Loon WKP, Bastings HMH, Moors EJ (1998) Calibration of soil heat flux sensors, Ag and For Meteor 92: 1-8 Webb EK, GI Pearman, Leuning R (1980) Correction of flux measurements for density effects due to heat and water vapour transfer. Quart J R Met Soc 106: 85-100 Xiu A, Pleim JE (2001) Development of a land surface model. Part I: Application in a mesoscale meteorological model. J Appl Meteor 40: 192-209 Xue M, Droegemeier KK, Wong V (2001a) The Advanced Regional Prediction System (ARPS) – A multi-scale nonhydrostatic atmospheric simulation and prediction model. Part I: Model dynamics and verification. Meteorol Atmos Phys 75: 161 – 193 Xue M, Droegemeier KK, Wong V, Shapiro A, Brewster K, Carr F, Weber D, Liu Y, Wang D (2001b) The Advanced Regional Prediction System (ARPS) – A multi-scale nonhydrostatic atmospheric simulation and prediction model. Part II: Model physics and applications. Meteorol Atmos Phys 74: 143 – 165 15 Fig. 1. State map displaying 89 OASIS sites (circles and triangles) and ten OASIS super sites (circles). Table 1. Measurement methods used at OASIS standard and super sites. Measurement Technique 16 90 Standard sites Domeless net radiometer, (NR-Lite) 10 Super sites Four-compenent net radiometer, (CNR1) Sensible heat flux (H) Gradient method (Paulson 1970; Brotzge and Crawford 2000) Eddy covariance using sonic anometery Latent heat flux (LE) Residual from the energy balance, LE = Rn – H – G Eddy covariance using sonic anometer and Krypton hygrometer Ground heat flux (G) Combination method, using the mean of two heat flux plates, soil moisture at 5 cm, and average soil temperature between 0 and 5 cm Combination method, using the mean of two heat flux plates, soil moisture at 5 cm, and average soil temperature between 0 and 5 cm Net radiation (Rn) 17 Standard Deviation of Soil Water Content in the 0-5 cm Layer vs. Days Since Rainfall at Norman, OK (1 June 1999 - 12 August 1999) 0.08 Sta ndard Dev iation 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0 5 10 15 20 Days Since Rainfall 25 30 Fig. 2. The standard deviation of soil water content in the 0-5 cm layer plotted as a function of days since the previous rainfall (Basara 2001). Figure includes data collected between 1 June and 12 August 1999 at the Oklahoma Mesonet site in Norman. Figure courtesy of J. Basara. Table 2: Summary of model experiments. 18 Summary of Model Experiments Exp. #1 Exp. #2 Soil Type [days] of Silt loam 10 days Figure 5 10 days Figure 6 10 days Figure 7 Varied, Figure 8, 9 Vegetation % Zo [m] Varied, Function 0 – 100% veg type (.02) 75% Varied, Silt loam 0.01 – 0.60 Exp. #3 75% Function of Continuous veg type (.02) Exp. #4 75% Function function % of Silt loam veg type (.02) 1 - 10 days 19 (a) 1 Correlation (R) 0.5 0 Solar Radiation Air Temperature Relative Humidity Wind Speed Rainfall -0.5 -1 0 20 40 60 80 100 Distance (km) (b) 1 Correlation (R) 0.5 0 Net Radiation Sensible Heat Flux Ground Heat Flux -0.5 -1 0 20 40 60 80 Distance (km) Fig. 3. Estimated correlation between sites plotted as a function of distance. 20 100 1 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Correlation (R) 0.8 0.6 0.4 0.2 0 0 20 40 60 80 Distance (km) Fig. 4. Estimated correlation of rainfall between sites plotted as a function of distance. 21 100 Fig. 5. The surface soil water content, wg (m3 m-3), plotted as a function of time of day during 20 May, 2000. The soil water content is estimated while varying the fractional vegetation cover (veg) from 0.0 to 1.0; observations are plotted as bold circles. 22 Fig. 6. The surface soil temperature, plotted as a function of time of day during 20 May, 2000. The soil temperature is estimated while varying the surface roughness (Zo) from 0.01 to 0.6; observations are plotted as bold circles. 23 Fig. 7. The surface soil water content, wg (m3 m-3) (a) and soil temperature, Ts (K) (b), plotted as a function of time of day for 20 May, 2000. The soil water content and soil temperature are estimated while varying the soil characteristics; observations are plotted as bold circles. 24 Fig. 8. Observations and model estimates of deep-layer soil temperature, t2 (K), plotted as a function of time of day from 1 August to 4 August, 2000. The time constant, , is varied from 1 to 10 days; observations are plotted using large circles. 25 Fig. 9. The top-layer soil temperature, Ts (K), plotted as a function of time of day during (a) 20 May, 2000, and (b) 2 August, 2000. The modeled surface temperature is varied as a function of damping depth, D (m); observations are plotted using large circles. Note the time lag of the model around 1200 UTC. 26 Fig. 10. Observed and model computed (a) sensible and (b) latent heat fluxes, as observed during 20 May, 2000. Observed data are plotted with bold circles. 27 Fig. 11. Observed and model computed (a) sensible and (b) latent heat fluxes, as estimated for 1 August, 2000. Observed data are plotted with bold circles. 28