Assessment of a three dimensional model for atmospheric radiative transfer over heterogeneous land cover A. McComiskey Department of Geography, University of California, Santa Barbara Currently: Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder / NOAA Global Monitoring Division P. Ricchiazzi Institute for Computational Earth System Science, University of California, Santa Barbara C. Gautier Institute for Computational Earth System Science, University of California, Santa Barbara Department of Geography, University of California, Santa Barbara D. Lubin Scripps Institution of Oceanography, LaJolla, California

Assessment of a three dimensional model for atmospheric radiative transfer over heterogeneous land cover

A three-dimensional atmospheric radiative transfer model that explicitly represents surface albedo heterogeneity is tested against a one-dimensional model and surface irradiance observations in a polar region where land cover heterogeneity is high. For observations located near high latitude coastlines, the contrast between the highly absorbing ocean and reflective snow or glacier surface creates spatial heterogeneity, or three-dimensional (3D) effects, around the observation site. The resulting effect on total radiation at the sensor should be taken into account when using a radiative transfer model to interpret measurements. This assessment shows that better closure is obtained with a three-dimensional model (≤ 5%) versus a plane-parallel model (≤ 7%) for the case examined here. The importance of the surface 3D effect increases with increasing aerosol or cloud optical depth and with increasing surface albedo contrast. The model described here can be implepented at any surface observation site given the surrounding land cover properties.

Assessment of a three dimensional model for atmospheric radiative transfer over heterogeneous land cover A. McComiskey, P. Ricchiazzi, C. Gautier, D. Lubin 1. Introduction Polar climates have shown more extreme and detectable responses to climate change relative to lower latitude climates and are important indicators of human induced changes to the environment [Holland and Bitz, 2003; Bromwich, 1998]. As harsh conditions limit access to inland polar regions, many observing stations are located near coastlines or aboard ships that remain in ice pack as measurements are made [Shupe and Intrieri, 2004; Ricchiazzi et al., 2002]. As the ice pack freezes, thaws, and drifts in and out around polar coastlines, considerable changes in land cover, and therefore surface albedo, occurs in the area surrounding an observation site. In these studies and others, modeling the radiative and climate forcing of atmospheric constituents is dependent, to a first order, on a robust estimate of the surface albedo [e.g., Gautier and Shiren, 1993]. Several 3D radiative transfer models that represent surface albedo heterogeneity around a measurement site have been developed to improve interpretation of surface radiation measurements and retrievals of aerosol and cloud properties at high latitudes [Degunther et al., 1998; Podgorny and Lubin,1998; Ricchiazzi and Gautier,1998]. Previous studies using these models have not included comparisons with 1D model results and measurements of surface irradiance, showing that the extra computational expense of a more complex model is required to simulate observations with greater accuracy. Here we use surface irradiance measurements made at Palmer Station, Antarctica, a coastal site, to assess the performance of a 3D model versus a plane-parallel (1D) model under conditions of varying and highly contrasting land cover. 2. Three-dimensional modeling for land cover heterogeneity The Surface Atmosphere Monte Carlo Radiative Transfer model (SAMCRT) model was developed to simulate explicitly the surface environment surrounding a measurement site at Palmer Station (64.77S, 64.05W) on the coast of the Antarctic Peninsula. A full review of the model is presented in Ricchiazzi and Gautier [1998], and Ricchiazzi, et al., 2002, and pertinent aspects are reviewed here. Traditionally, Monte Carlo techniques simulate the radiance distribution through a statistical average of a large number of photon trajectories which enter though the top of the atmosphere [Marchuk et al., 1980]. These traditional techniques are not designed for the computation of irradiance at a single point on the surface, which is the value of interest for this study. In this case, the area representing the sensor is a very small fraction of the total area of influence on the irradiance reaching that point. The statistics associated with a very large number of photons would be required to realistically simulate the influence on irradiance from the surrounding surface albedo distribution. In this study we use a backward propagation technique [Marchuk, et al., 1980], whereby photons are injected into the atmosphere from a single point on the surface representing the sensor location. The method of expected values, developed by Gordon [1985], is used to determine the contribution of scattering by Rayleigh and aerosol, cloud, and surface events in the sunward direction. Photon trajectories are affected by interactions associated with absorbing gases, molecular or Rayleigh scattering, aerosol and cloud scattering, and surface interactions. Gas absorption in the model is computed using the HITRAN96 database to produce correlated-k fits [Yang et al., 2000]. Scattering events in the atmosphere are handled explicitly with the scattering phase function

regulating the trajectory of a scattered photon in the atmosphere. Molecular and aerosol scattering is allowed to occur at any altitude in the atmosphere. Distance along the current photon trajectory to the next scattering event is determined by:

µσ /))(( 01 NzNeRN −−= where N(z) is the total column density, N0 is the column density at the current photon height, σ is the scattering cross-section, and µ is the cosine of the solar zenith angle of the photon trajectory. Much work has been done to address the issue of three-dimensional cloud effects on the radiation distribution in the atmosphere and at the surface [O’Hirok, 1998]. However, Antarctic cloud cover is dominated by low stratus which is well represented by the plane-parallel assumption. To simplify cloud scattering interactions in the model, all cloud effects are lumped into a single set of interactions based on plane-parallel theory and no horizontal transport is allowed within the cloud. Few measurements have been made of cloud microphysical properties in the Antarctic. Saxena and Ruggiero [1990] report a bimodal drop size distribution for mostly liquid clouds near McMurdo Station with mode sizes of 2 and 6-9 µm. Since no such measurements exist for Palmer Station, which is more northerly than McMurdo, liquid phase clouds with an effective radius (re) of 10 µm are specified in the model following a gamma drop size distribution:

)()/(

)(0

/10

0

pReRr

rNRrp

Γ=

−−

where Γ is the gamma function, R0 = re/(p+2), and p is a fitting parameter equal to 7. Mie scattering theory [Wiscombe, 1980] is used to determine cloud radiative properties including the volume extinction coefficient, single scattering albedo, and phase function. The model surface, created from a digital elevation model from the British Directorate of Overseas Survey published in 1963, consists of a 20 km2 grid of 200 m2 cells. Surface scattering is explicit with photon trajectories dictated by the surface bi-directional reflectance distribution function (BRDF). Surface topography (aspect and slope), BRDF, and albedo are specified for each grid cell. Near Palmer Station the surface is dominated by two cover types: ocean and snow. The reflectivity of the ocean surface is very low compared with the snow, with an albedo of approximately 0.05. Because the computed radiation parameters are insensitive to ocean spectral albedo or BRDF, ocean properties are specified in the model as a Lambertian surface with an albedo of 0.05. Snow in the Antarctic is very pure with little in the way of impurities that would decrease albedo. Snow spectral albedo is specified according to Wiscombe and Warren [1980] and with a small effective radius of 100 µm. Cold temperatures and frequent snowfall provides little opportunity for snow contamination and/or metamorphosis, both of which reduce snow reflectivity. SBDART (Santa Barbara DISORT Atmospheric Radiative Transfer model) [Ricchiazzi et al., 1998] is a plane-parallel model that treats atmospheric and surface layers as horizontally homogeneous. This model has been used successfully for atmospheric process studies in polar regions and many other locations around the globe [e.g., Ricchiazzi et al., 1995], but can not resolve the typical patterns of surface reflectance at Palmer Station in the horizontal dimension. The two models will be run in comparison with irradiance measurements made at the surface. In order to verify that the model output is comparable when the prescribed surface conditions are identical, the two models were each run with homogeneous surface albedo representations ranging from 0.05 to

0.99 at five different wavelengths between 415 and 870 nm. The differences between the models in all cases ranged from 0.002 to 0.004% indicating that the models can be reliably compared. The effect of explicitly modeling surface albedo heterogeneity is presented in Figure 1. Atmospheric transmission at 550 nm, for a solar zenith angle of 45o, and with a standard sub-Arctic summer atmospheric profile is computed while varying the surface albedo. Computations are made with the 1D model (grey) in the same manner as with the 3D model (black). The representation of the surface albedo in the 3D model is such that the portion of the model grid covered by glacier is set to 0.95. The albedo for the portion of the model grid representing ocean is varied at 0.05, 0.25, and 0.75, theoretically representing an increasing sea ice cover. The representation of the surface albedo in the 1D model is an average of the two albedos used in the 3D model based on the fraction of snow/ocean coverage. Atmospheric transmissions are computed over a range of aerosol optical depths (Figure 1a) and cloud optical depths (Figure 1b). The impact of the different surface albedo representations is seen to increase with surface albedo contrast as well as increasing optical depth, up to very optically thick cloud (~120). Differences in transmission up to 1.5 % occur for the aerosol optical depth cases show in Figure 1a. For the cloud optical depth cases shown in Figure 1b, differences of up to 30% in transmission are seen between the two models. For this particular case, the transmission difference translates to 12 W·m-2 for aerosol and 64 W·m-2 for cloud. Differences between the two models will vary as wavelength, solar zenith angle, and atmospheric constituents change. 3. Surface Irradiance and Albedo Observations Measurements of surface irradiance were made at Palmer Station during the austral summer of 1999-2000 using a Multi-Filter Rotating Shadowband Radiometer (MFRSR) [Harrison et al., 1994]. The MFRSR has six spectral channels at 415, 500, 615, 673, 870, and 940 nm, each with a full-width at half maximum (FWHM) width of approximately 10nm. The shadowband allows the instrument to report data for the total, diffuse, and direct-normal components of the surface irradiance in these channels, with measurements made at one-minute resolution. The direct irradiance measurements allow for the determination of aerosol optical depth on clear-sky days. The MFRSR was run continuously throughout the season. Direct irradiance measurements also allow for a solar or in situ calibration of the instrument, considered preferable to lamp calibration, if a sufficient number of clear sky observations days exist from which to obtain Langley regressions. The instrument was stationed in Santa Barbara, CA between 07 July 1999 and 26 August 2000 during which 72 acceptable Langley events occurred. Langley regressions for the spectral calibration of the MFRSR were determined using the objective algorithm developed by Harrison and Michalsky [1994]. Forty-two Langley events occurred before the field season and 30 events occurred afterward. The standard error, a common estimator of uncertainty, is within 2% for all channels at the 95% confidence level. Angular calibrations in the lab in March of 1999 before the field season and in June 2000 after returning from the field shows that the accuracy is generally within 5% of the ideal cosine response. Observations of land cover distribution were made on six days throughout the season between 4 November 1999 and 19 January 2000 using a 180o lens on a digital camera attached to a helium balloon that was flown up to 1 km above the surface observation site. On the ground, the same camera was pointed upwards to obtain an all-sky image each day documenting cloud cover conditions. Examples of both types of images for two days with varying land cover and sky conditions are shown in Figure 2. Information from these observations is used to constrain the models and to provide a qualitative validation of model results. Only one of the six days that included a balloon flight (27 November 1999) was clear enough to obtain aerosol optical depths from the MFRSR observations. Optical depths at the five

wavelengths and the spectral snow albedo model discussed above are used to compute surface irradiances with the 1D and 3D models to determine closure for this day. Inputs for the models are listed in Table 1. The surface on this day had a fresh snow cover over the glacier and the ocean was free of sea ice cover (see Figure 2b). Other aerosol properties required to compute surface irradiance include the single scattering albedo (ω0) and asymmetry parameter (g), which are taken from a climatological sea salt aerosol model [d’Almeida et al., 1991] which comprises the bulk of the aerosol in this region. For this range of wavelengths, ω0 is constant at 0.99 and g is constant at 0.75 except at 870 nm where it is 0.76. Computed irradiances from both models are compared with the surface measurements. For all other days when cloud properties are unknown, the two models are used with the MFRSR measured total surface irradiance to retrieve cloud optical depths, which are then compared to each other. 4. Model Assessment Results from the closure exercise for 27 November 1999, presented in Table 2, show that the 3D model provides better closure with the measurements than the 1D model. This is true especially for the shorter wavelengths (415 and 500 nm) where the aerosol optical depth is more significant. A difference of 1-2% in irradiance between the two models at these wavelengths is comparable to the measurement error estimated for surface radiation. A recent study at a mid-latitude continental site where surface albedo is low and relatively homogeneous [Michalsky et al., 2005] shows that closure for direct and diffuse irradiance on clear-sky days for well determined aerosol properties is < 2%. The 3D model provides results for the high latitude coastline that are are more consistent with the most current closure obtained at a lower latitude where surface albedo is not as heterogeneous. For the remaining three days of MFRSR measurements, cloud optical depths are retrieved from each of the models using the measured surface irradiances and observed surface albedo configurations. Comparisons of the retrieved optical depths are presented in Figure 3. Common cloud optical depths in the region of Palmer Station are 15 to 30. In this range, significant differences in a few cases of up to 20% between the models can be seen. The 1D results are often lower than the 3D results indicating that an underestimate of cloud radiative effects may occur when proper surface albedo is not specified. 5. Conclusion For areas of heterogeneous land cover, especially with highly contrasting surface albedo, a three-dimensional radiative transfer model that represents the surface albedo distribution explicitly is compared to a plane-parallel model. Closure with surface irradiance measurements is improved from < 7% to < 5% when the surface albedo is modeled explicitly. Differences between a 1D model and 3D model range from 7% for computed irradiances, 30% for atmospheric transmission, and 20 % for retrieved cloud optical depths. For high latitude climate studies, which often occur near coastlines or aboard research vessels, the effect of shifting sea ice cover may have a significant inpact on how surface irradiance measurement are used to interpret atmospheric properties, including the radiative forcing of clouds and aerosols. A three dimensional radiative transfer model that represents heterogeneous land cover can be used to improve the treatment of atmospheric scattering and absorption processes and surface radiation interactions. Obtaining closure in polar radiation studies at the same level achieved at lower latitudes where land cover heterogeneity is not as pronounced may require the use of a 3D surface radiation model.

Acknowledgements This work was supported in part by the Atmospheric Radiation Measurement Program (ARM) of the U.S. Department of Energy under grant DE-FG03-90ER61062 and in part by the National Science Foundation under grant OPP-9725403.

References Bromwich, D.H., and T.R. Parish, Antarctic: Barometer of climate change (1998), Report to the

National Science Foundation from the Antarctic Meteorology Workshop, 1-7. d’Almeida, G., P. Koepke, E. Shettle (1991), Atmospheric aerosols: their global climatology and

radiative characteristics, A. Deepak Publishing, Hampton, VA. Degunther, M., R. Meerkotter, A. Albold, and G. Seckmeyer (1998), Case study on the influence of

inhomogeneous surface albedo on UV irradiance, Geophys. Res. Lett., 25, 3587-3590. Gautier, C., and Y. Shiren (1993), Role of cloud-surface interactions on the net surface solar flux in

high latitudes: climatic implications, SPIE Atmospheric Radiation, 2049, 24-30. Gordon, H.R. (1985), Ship perturbation of irradiance measurements at sea 1. Monte-Carlo

simulations, Appl. Opt., 24, 4172-4182. Harrison, L., J. Michalsky, and J. Berndt (1994), Automated multi-filter rotating shadowband

radiometer: An instrument for optical depth and radiation measurements, Appl. Opt., 33, 5118-5125.

Harrison, L. and J. Michalsky (1994), Objective algorithms for the retrieval of optical depths from ground-based measurements, Appl. Opt., 33, 5126-5132.

Holland, M. M. and C. M. Bitz (2003), Polar amplification of climate change in coupled models, J. Climate, 21, 221-232.

Marchuk, G., G. Mikhailov, M. Nazaraliev, R. Darbinjan, B. Kargin, and B. Elepov (1980), The Monte Carlo Methods in Atmospheric Optics, Springer Series in Optical Sciences, Springer Verlag, New York.

Michalsky, J.J., G. P. Anderson, J. Barnard, J. Delamere, C. Gueymard, S. Kato, P. Kiedron, A. McComiskey, and P. Ricchiazzi, Radiative Closure Studies for Clear Skies During the ARM 2003 Aerosol Intensive Observation Period, J. Geophys. Res., in review.

O’Hirok, W. and C. Gautier (1998), A three-dimensional radiative transfer model to investigate the solar radiation within a cloudy atmosphere. Part II: Spectral effects, J. Atmos. Sci., 55, 3065-3076.

Podgorny, I., and D. Lubin (1998), Biologically active insolation over Antarctic waters: Effect of a highly reflecting coastline, J. Geophys. Res., 103, 2919-2928.

Ricchiazzi, P., and C. Gautier (1998), Investigation of the effect of surface heterogeneity and topography on the radiation environment of Palmer Station, Antarctica with a hybrid 3-D radiative transfer model, J. Geophys. Res., 103, 6161-6176.

Ricchiazzi, P., A. Payton, and C. Gautier (2002), A test of three-dimensional radiative transfer simulation using the radiance signatures and contrasts at a high latitude coastal site, J. Geophys. Res., 107, 4650-4665.

Saxena, V.K., and F.H. Ruggeriero (1990), Aerosol measurements at Palmer Station, Antarctica, in Contributions to Antarctic Research I, Antarctic Research Series, 1-5.

Shupe, M. D. and J. M. Intrieri (2004), Cloud radiative forcing of the Arctic Surface: The influence of cloud properties, surface albedo, and solar zenith angle, J. Climate, 17, 616-628.

Wiscombe, W. J., and S. G. Warren (1981), A model for the spectral albedo of snow I: Pure Snow, J. Atmos. Sci., 37, 2712-2733.

Yang, S., P. Ricchiazzi, and C. Gautier (2000), Modified correlated-k distribution methods for remote sensing applications, J. Quant. Spectrosc. Radiat. Transfer, 64, 585-608.

Figure Captions Figure 1. Difference in atmospheric transmissions between a 1D model (grey lines) and 3D model (black lines) with varying surface heterogeneity for a range of (a) aerosol optical depth and (b) cloud optical depth. Surface albedo is shown in the key as the 3D glacier albedo, 3D ocean albedo and (1D albedo).

Figure 2. Sky cover and surface albedo balloon images for (a) 9 Nov 99 and (b) 27 Nov 99. Figure 3. Retrievals of cloud optical depth from 1D (grey) and 3D (black) models.

Tables

Table 1. Aerosol and Surface Properties for 27 Nov 99 used to model surface irradiances. Taua 1D αs

3D αs (land)

3D αs (ocean)

415 nm 0.061 0.517 0.983 0.05 500 nm 0.040 0.520 0.990 0.05 615 nm 0.032 0.508 0.966 0.05 670 nm 0.030 0.501 0.952 0.05 870 nm 0.026 0.457 0.864 0.05

Table 2. Measument/Model Comparison for 27 Nov 99 (W·m-2). MFRSR 1D 3D 415nm 1095.8 1134.8

(3.6 %) 1116.4 (1.9 %)

500nm 1258.3 1340.9 (6.6 %)

1324.3 (5.2 %)

615nm 1093.5 1105.2 (1.1 %)

1100.9 (0.7 %)

670nm 1020.8 1019.0 (-0.2 %)

1014.3 (-0.6 %)

870nm 647.7 631.1 (-2.6 %)

629.4 (-2.8 %)

Figure 1. Difference in atmospheric transmissions between a 1D model (grey lines) and 3D model (black lines) with varying surface heterogeneity for a range of (a) aerosol optical depth and (b) cloud optical depth. Surface albedo is shown in the key as the 3D glacier albedo, 3D ocean albedo and (1D albedo).

Figure 2. Sky cover and surface albedo balloon images for (a) 9 Nov 99 and (b) 27 Nov 99.

Figure 3. Retrievals of cloud optical depth from 1D (grey) and 3D (black) models.