At What Scales Does Landscape Heterogeneity Impact Climate?
Roni Avissar
Rutgers University
New Brunswick, New Jersey
At What Scales Does Landscape Heterogeneity Impact Climate?
Roni Avissar
Rutgers University
New Brunswick, New Jersey
Avissar discussed the impacts of landscape heterogeneity on atmospheric turbulence, mesoscale circulation, clouds, and precipitation and presented an approach for representing these effects in global models. Atmospheric effects from landscape heterogeneity are examined at three scales: the microscale, defined as up to 2 kilometers, the large scale, above 2000 km, and the mesoscale, which is between the microscale and the large scale.
To understand the effects of landscape heterogeneity on the atmosphere and recognize which are the most important processes and interactions, we begin by asking which land-surface parameters are of greatest importance for atmospheric models. The method used to answer this question is the Fourier Amplitude Sensitivity Test (FAST) with Land-Atmosphere Interactions Dynamics (LAID). FAST is a very efficient and powerful technique for determining the relative contribution of the distribution of individual input parameters (assuming that their exact value is unknown) to the variance of the model output. By simultaneously varying all parameters according to their individual probability density functions, the number of computations needed is very much reduced by this technique.
This technique was used to obtain the most important land-surface characteristics for forcing climate. Results of this analysis indicate that for vegetated land, stomatal conductance and surface roughness are the two most important characteristics for forcing the atmosphere. For bare land, soil-surface wetness and surface roughness are most important. Leaf area index (LAI) is another important parameter as it indicates the relative amounts of vegetation and bare ground. Albedo can also play a significant role in certain circumstances. It is important to point out that this analysis does not offer a priority list for all situations. Different parameters are important under different atmospheric conditions. In addition, this analysis assumes that parameters are independent, whereas in reality, they are related.
For vegetated land, stomatal conductance and surface roughness are the two most important characteristics for forcing the atmosphere. For bare land, soil-surface wetness and surface roughness are most important.
Microscale Heterogeneity Patchiness
What is the impact of the spatial variability of the most important land-surface parameters on land-surface heat fluxes? The method used to address this question was the Patchy Land-Atmosphere Interactions Dynamics (PLAID) technique of Avissar and Pielke (1989). In this "mosaic approach," different subgrid surface types are treated as different "tiles" in a grid "mosaic." This assumes that the horizontal fluxes between different types of vegetation in the landscape are not important. Each system is resolved independently, and then an areal average of these is used to calculate the fluxes that come from the grid. This method is only useful for scales up to 5 to 10 km; at scales larger than this, the atmospheric boundary layer dynamics are affected by land-surface heterogeneity, and this method may not be correct. The major conclusion from this work is that except for albedo, which relates more or less linearly to the surface heat fluxes, the spatial variability of the four other parameters (stomatal conductance, surface roughness, leaf area index, and soil-surface wetness) should be considered to avoid significant errors; using a mean value, as opposed to using the entire distribution, introduces errors as large as 100 watts per square meter.
Microscale Heterogeneity: Large-Eddy Simulations and Lidar Observations
Shifting the focus from small eddies to the effects of large eddies, Avissar discussed the impact of microscale spatial variability of surface sensible heat flux and topographical features on the atmospheric boundary layer. He described Large-Eddy Simulations (LES), a three-dimensional model used at a very fine resolution (on the order of 100 meters) to resolve the large eddies of turbulence. It is well known that the most energetic turbulent eddies have a typical size on the order of 800-1000 meters, and this type of model is designed to resolve these eddies. This is a powerful technique but the high resolution makes it impractical to run more than one diurnal cycle. The resolution is very fine at the ground surface (5 meters) and becomes coarser further away from the surface, so at about 200-300 meters above the surface, the resolution is about 100 meters.
In this study, the LES version of the Regional Atmospheric Modeling System (RAMS) is used. Avissar emphasizes the importance of this work because it indicates when the spatial heterogeneity triggers processes or mechanisms that make it incorrect to assume horizontal homogeneity. The key conclusion from this study is that as long as the characteristic length scale of heterogeneity is smaller than 5 to 10 km, and the topographical features are smaller than about 200 meters, there is no significant impact on the mean characteristics of the convective boundary layer.
As long as the characteristic length scale of heterogeneity is smaller than 5 to 10 km, and the topographical features are smaller than about 200 meters, there is no significant impact on the mean characteristics of the convective boundary layer.
For validation of this numerical model, Avissar presented observations from FIFE (the 1980s Kansas field experiment). He presented an image from a scanning LIDAR (a light detection and ranging instrument) which shows backscattering from aerosols in the boundary layer (see Figure 1.3 below). This image reveals the structure of turbulent eddies in the atmospheric boundary layer by showing the aerosols transported by those eddies. Power spectrum techniques and spatial and temporal auto correlations are used to discern if the model properly represents this spectrum of turbulence. The LIDAR image shows that eddies on the scale of 800 meters are those with the most energy in the atmosphere. Auto correlations reveal that the lifetime of eddies in the convective boundary layer is roughly ten to fifteen minutes.
Figure 1.3
This LIDAR image reveals the structure of turbulent eddies in the atmospheric boundary layer by showing backscattering from the aerosols transported by those eddies.
Numerical Experiments
Parameterizations of the land surface are very controversial. An intercomparison experiment revealed that even models based on the same concepts had large differences in results when they used different parameterizations of soil moisture in particular. To avoid these drawbacks and focus instead on the dynamics of the atmosphere, Avissar and colleagues forced a model (RAMS-LES) with the spatial distribution of fluxes (diurnal variations of the latent heat flux, heat conducted from ground, and sensible heat flux) from a network of surface observations. In the area observed, the maximum difference in topography is only 100 meters. Results from three simulations indicate that with topographical features of up to 200 meters, there is no effect on the mean properties of turbulence in the boundary layer. The topography is not important to the strength of the eddies but rather simply serves to anchor them. But with topographical differences of more than 200 meters, the mean properties of turbulence and kinetic energy in the boundary layer are significantly affected by landscape heterogeneity.
With topographical differences of more than 200 meters, the mean properties of turbulence and kinetic energy in the boundary layer are significantly affected by landscape heterogeneity.
There is a need to better parameterize eddies as they decrease in size below 100 meters. The current subgrid-scale parameterization does not do a very good job at dissipating eddies. Such an improvement is under development in Avissar's research group and he is optimistic about its success.
The next step is to use this model to determine at what point the dynamics of the atmosphere are altered by landscape heterogeneity. To do this, the model is forced with different amplitudes and wavelengths of heat fluxes. Results show that the distribution of heat flux in the boundary layer is a straight line in homogeneous terrain. But in heterogeneous terrain, with 20 or 40 kilometer waves present, turbulent kinetic energy is very strong close to the ground surface and also very strong close to the top of the boundary layer. This emphasizes that the eddies are organizing in circulation which has a strong component close to the ground surface and near the top of the boundary layer. Results also indicate that this process is nonlinear, and depends on the intensity of the mean heat flux fueling the boundary layer.
Avissar then showed a wavelet analysis of the results, which is similar to a Fourier analysis but has the advantage of showing the location of the different eddies. The key result of this analysis is that a gap is found between the microscale and mesoscale eddies, emphasizing the possibility of making an objective separation of scales. This has significant implications for the design of atmospheric models and subgrid-scale parameterizations.
In addition, when the mean heat flux is relatively small, the horizontal scale of the wavelength impacts the atmosphere much more significantly than when the mean heat flux is high. The ratio between the contribution to turbulent kinetic energy by buoyancy and by the horizontal pressure gradient seems to be a useful dimensionless quantity for the development of an appropriate parameterization of these effects. The more significant the buoyancy, the more difficult it is for horizontal structure to develop in the boundary layer.
Mesoscale Heterogeneity: Numerical Modeling
Moving to much larger scales, what is the impact of mesoscale spatial variability of surface heat fluxes and topographical features on the atmosphere? In sum, results indicate that mesoscale discontinuities considerably affect the vertical profile of mean atmospheric variables, heat fluxes, clouds, and precipitation. These effects should be parameterized in GCMs. Mesoscale perturbations create atmospheric dynamical processes that can be extremely important.
Eddies are organizing in circulation which has a strong component close to the ground surface and near the top of the boundary layer.
Avissar discussed an upcoming field experiment in Rondonia which he believes will be very important in helping to improve parameterizations of mesoscale effects in GCMs. A satellite image of the study area reveals the fishbone pattern of deforestation and development along roads which is a characteristic pattern of the expansion of human activity in tropical forests. Two types of heterogeneity are apparent: microscale heterogeneity on the order of a few kilometers and mesoscale heterogeneity on the order of 300 km by 200 km. In addition, the natural topography varies from sea level up to 1100 meters. Despite these large variations, GCMs assume a flat domain with one big leaf covering the whole area; thus, one cannot expect the results of such models to fit reality.
In preparation for this field experiment, numerical simulations have been conducted to assess the effects of microscale and mesoscale heterogeneity on clouds and precipitation. The dynamics of circulation in these simulations are based on results from the LES; the only parameterizations used are for cloud microphysics. Five different domains are used to consider combinations of different types of heterogeneity. Results for accumulated precipitation over these different domains after one day indicate that over homogeneous pasture, there is virtually no precipitation. Over homogeneous forest, there is a random distribution of precipitation cells, which is not surprising because turbulence is the dominant forcing mechanism, and it has a random structure. When there is landscape heterogeneity, it strongly affects the distribution of precipitation. At the microscale, landscape organizes precipitation but does not add to the total amount of water. However, mesoscale landscape heterogeneity significantly increases precipitation, which is well organized according to the heterogeneity.
Another important note is that current GCMs parameterize only one of the two important processes simulated in this experiment. They parameterize the effects of the turbulent heat fluxes which are dominant closer to the surface but fail to include mesoscale heat fluxes, which are dominant in the middle and upper part of the boundary layer. This is a major problem, Avissar says, because the effects of the mesoscale fluxes are much greater and must be parameterized to make the GCMs better simulate reality.
Omitting these effects has a very significant impact on the results of GCMs. Account for these heterogeneous effects over all of Earth's land surface is as important as doubling the carbon dioxide concentration in a GCM (on order of 2°C), Avissar says. The landscape causes perturbations to the system that can affect large scale atmospheric circulation patterns. Roger Pielke adds that thunderstorms are a key element of this because they develop based on surface considerations and have a disproportionate effect on climate. The landscape tends to organize the total moisture available and this has huge and far reaching climatic effects. One example of such far reaching impacts, Pielke says, is El Niño, where a local warm ocean anomaly has teleconnections that have climatic effects around the globe.
Accounting for these heterogeneous effects over all of Earth's land surface is as important as doubling the carbon dioxide concentration in a GCM.
An Approach to Parameterizing Mesoscale Effects in GCMs
Similarity theory suggests a four step system to parameterize these effects: 1) identify the variables relevant to the problem; 2) organize the variables into dimensionless groups; 3) gather observations or perform experiments to determine their values; and 4) find an empirical relationship between dimensionless groups. Because this approach is based on empirical relations, experiments are needed to provide appropriate data sets. When the Rondonia experiment is complete, such data will exist, but in the meantime, a model is used to create characteristic landscapes, and quantities are derived from resolved parameters in GCMs. There is a need to develop Soil Vegetation Atmosphere Transfer Schemes (SVATS) to parameterize these effects into GCMs. Avissar believes that there is hope for incorporating mesoscale effects into GCM-scale grids but that it will require close collaboration between atmospheric and land surface researchers.
Conclusions
Complex mesoscale and microscale (turbulent) interactive processes are involved in the development of shallow convective clouds in heterogeneous landscapes. Even though these processes can significantly affect predictions at all time and spatial scales, they are not parameterized in GCMs and other large-scale atmospheric models. A preliminary parameterization was developed based on large-scale atmospheric conditions which are resolved by the large-scale atmospheric model, the variance of surface sensible heat flux, and a characteristic length scale of heterogeneity of the landscape. Research in this area is needed to improve this parameterization. The development of SVATS able to produce such a length scale and variance is needed. The Rondonia experiment discussed above offers a unique opportunity to provide the appropriate data set needed to calibrate and evaluate this type of parameterization.
Reference
Avissar, R. and R. A. Pielke, 1989. A parameterization of heterogeneous land-surface for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev., 117:2113-2136.
There is hope for incorporating mesoscale effects into GCM-scale grids but it will require close collaboration between atmospheric and land surface researchers.
Forward to Next Section//Back to Table of Contents//AGCI Homepage//Comments: agcimail@agci.org