Climatefield downscaling for massbalance modelling

There are unique challenges of scale for efforts to estimate climate change impacts on the world's glacial systems. For mountain glaciers, regional-scale icefields and ice-sheet margins, temperature and precipitation fields need to be downscaled to the scale of relevance for glacier mass balance, of the order of 1-10 km. A full treatment of energy balance on glaciers and icefields also requires wind, humidity and radiation balance terms at the 1-10km scale. This poses a significant challenge at the resolution of current operational climate models, ca. hundreds of kilometres.

Pollard & Thompson (1997) discuss many of the issues and strategies involved in downscaling of AGCM climate fields for glaciological modelling. The main challenge in simulating mass balance is adequate resolution of the topographic detail. Altitudi-nal controls on temperature govern both surface melt rates and

P (mm) 7000 6000 5000 4000 3000 2000 1000 0

rT 1.00

a Global gridded data a Global gridded data

-50 -40 -30 -20 -10 0 10 20 30 Mean annual temperature (°C)

rT 1.00

4000

3000

2000

1000

c Canadian station data

-20 -15 -10 -5 0 5 10 15 Mean annual temperature (°C)

Figure 32.2 Relationship between annual gauge-corrected precipitation and mean annual surface temperature. (a) Global gridded observational climatology (Legates & Willmott, 1990a,b), for all data points in 0.5° by 0.5° latitude-longitude grid cells. (b) Zonally averaged linear correlation coefficient rTP for surface temperature and precipitation. (c) Precipitation versus mean annual temperature for all available 1971-2000 climatic average station data in Canada (N = 1170), illustrating the relationship for a mid- and high-latitude continental environment (Environment Canada, 2003).

the critical transition between liquid and solid precipitation. If subgrid topographic distributions are known, atmospheric (temperature-elevation) lapse rates can be used to estimate surface temperature at a mosaic of subgrid cells (e.g. Thompson & Pollard, 1997; Giorgi et al., 2003) or over a statistical distribution of subgrid terrain elevations (Walland & Simmonds, 1996; Marshall & Clarke, 1999). Surface temperature lapse rates (verti cal gradients of surface temperature) are governed by local surface energy balance rather than free-air processes, and are typically less than free atmosphere lapse rates. Surface lapse rates also vary spatially and temporally, so there are significant uncertainties in temperature downscaling from direct altitude adjustments. Simple statistical or dynamical methods are needed to estimate surface lapse rate variability and its controls in different environments.

In addition to the direct effects of altitude on temperature, spatial accumulation rate patterns have a complex dependence on orographic forcing. The intrinsic spatial variability of precipitation is typically much less than the resolution of either modelled or reanalysed climatology, with significant spatial gradients of snow accumulation on scales of hundreds of metres to several kilometres (e.g. Koerner, 1977; Erxleben et al., 2002). At regional scales, this complexity in snow accumulation patterns is largely governed by the interactions of topography with synoptic systems. For instance, orographic forcing of frontal cyclonic systems is responsible for most of the moisture deposition on the Greenland Ice Sheet (Chen et al., 1997). At smaller scales, spatial snowpack gradients are associated with wind redistribution and local topographic influences.

This introduces the requirement to downscale precipitation to subgrid scales. In practice, the most common approach is to distribute resolvable (i.e., frontal) AGCM grid-scale precipitation uniformly across subgrid terrain features (Giorgi et al., 2003), although it is recognized that subgrid orographic controls will give highly non-uniform distributions in complex terrain. Physically based statistical or dynamical models of precipitation variability need to be applied to generate a more detailed approximation of spatial distributions (e.g. Daly et al., 1994).

Mesoscale climate and forecast models are now approaching the resolution of interest for regional topographic detail and precipitation processes (Giorgi et al., 1994, 1996, 2003; Colle et al., 1999; Leung & Qian, 2003). Regional-scale models simulate atmospheric dynamics over a high-resolution domain that is nested within a global-scale model, with boundary conditions coming from the global model. Studies of the Pacific northwest USA by Colle et al. (1999) explored grid cells as fine as 4km, but with little improvement in precipitation forecasts relative to simulations at 12-km resolution. This points to the fact that model performance is limited by physical process understanding as well as resolution of topographic detail and land-surface heterogeneities. Parameterizations of subgrid physics need to be retuned at different resolutions (Giorgi & Marinucci, 1996), and parame-terizations may not be generalizable to different regions.

Limits to resolution-derived improvements are also expected in association with model nesting strategies, the coarse resolution of boundary forcing in nested regional models, and the lack of mesoscale data that is available to constrain and test mesoscale models. Leung & Qian (2003) explored two-way nesting in regional climate model experiments over complex terrain, with promising results in simulation of precipitation distribution at 13-km resolution. They found marked improvements over simulations at 40-km resolution, although modelled snow distribution at both 40- and 13-km scales is poor. Climate models may require additional subgrid-scale precipitation physics to give improved, dynamically based estimates of precipitation in complex terrain (e.g. Hulton & Sugden, 1995; Chen et al., 1997; Roe, 2002).

Mesoscale model development is an active research frontier, with significant progress to be expected on this front. As subgrid physical processes become better understood and modelled, dynamical downscaling may become the standard for simulating the mass balance of alpine glaciers and icefields. It is nevertheless practical to assume that global-scale climate models will not reach the necessary resolution for some phenomena of interest in the foreseeable future (e.g. kilometre-scale hydrological and land-surface processes; mountain glacier distributions). A combination of physically based downscaling strategies and regional/highresolution climate models therefore needs to be pursued.

Was this article helpful?

0 0

Post a comment