As discussed in the previous sections, permafrost modeling in NWPMs, CTMs, GCMs and ESMs still has several short-comings. Some of them may be addressed easily as available computer resources increase with the next generations of supercomputers, while others require serious research and data collection efforts.
Increased computational power will permit us to consider more layers and locate the lower boundary of NWPM, CTM, GCM and ESM soil models at deeper levels, i.e. reduce uncertainty related to the choice of the lower boundary condition. This way a greater depth of the soil model does not compromise the required fine resolution in the upper soil that is required to capture the diurnal and seasonal cycle of active layer depth.
As has been shown by Narapusetty and Molders (2006) finite element schemes permit us to better capture the phase and amplitude of soil temperature variations and hence the active layer depth. Currently the computational burden is too high to run GCMs and ESMs for several decades using such methods. Therefore this improvement has to be postponed until the next generations of supercomputer will become available.
The difficulties related to initialization of soil moisture and temperature in NWPMs and CTMs could be addressed by developing a kind of analysis procedure like applied to initialize the atmosphere in NWPMs. Such an analysis method would require to measure worldwide soil temperature and moisture at the same universal coordinated time (UTC) several times a day like is common practice for meteorological data. These soil data would have to be reported and collected at a central place in an agreed upon format like GRIB that is used by the World Meteorological Organization (WMO) for reporting the huge amount of meteorological data. Some kind of interpolation procedure would have to be run to produce a hydro-thermodynamically consistent gridded global soil temperature and moisture dataset based on the latest observations.
Data of soil type distribution exist in various different data sources and must be gathered in a data center to derive a quality assessed and quality assured global gridded dataset.
ESMs that consider random aerosol release from wildfires and wildfire related land-cover changes in the biogeochemical cycles and CTMs that serve for wildfire smoke forecasts in boreal regions could be enlarged to also consider the impact of wildfires on permafrost. Doing so would require assuming a wildfire-related heat source at the top of the soil where currently the wildfire related aerosols are released and later land-cover is changed.
Non-representative network design, low site density, shut-down and/or adding sites to long-term monitoring networks can introduce substantial uncertainty in gridded data (e.g. PaiMazumder and Molders 2009). Therefore it is an urgent need (1) to assess the potential influence of networks on gridded data derived there from, (2) to develop evaluation strategies for application of gridded data from "imperfect" existing long-term networks, and (3) to develop recommendation to improve existing networks and/or design better networks in the future.
To avoid errors from non-representative networks in gridded soil-temperature data some kind of data assimilation could be used. Similar to reanalysis in atmospheric sciences all available soil data plus meteorological forcing data as upper boundary condition in conjunction with physical soil modeling could be performed to provide some kind of reanalysis (e.g., Kalany et al., 1996, Uppala et al., 2005) for soil temperature. This method could consider the various soil types within a grid and the gridded dataset would provide a weighted soil temperature. The weighting would be with respect to the fractional coverage of a given soil-type within the grid area like soil temperatures are typically simulated in GCMs or ESMs.
Was this article helpful?