EOS, Transactions, American Geophysical Union, Vol. 80, No. 19, May 11, 1999, Pages 213, 220-221.

Radar Remote Sensing Proposed for Monitoring

Freeze-Thaw Transitions in Boreal Regions

New research is finding that satellite-based radar remote sensing techniques are particularly well suited for quantifying the transition of remote boreal regions from a frozen to a thawed condition. The implications for studying global warming are far reaching. If the timing or areal extent of this freeze/thaw transition were to change significantly, measurable changes in boreal climate, hydrology, and biogeochemistry would result.

Abrupt transition from frozen to thawed conditions occurs each year over roughly 50 million km² of the Earth's remote terrestrial surface at latitudes above 40^o North. Radar remote sensing works well to capture this transition because of the way electromagnetic radiation at radar wavelengths interacts with polar water molecules in solid and liquid states. Also, radar has the substantial advantages at high latitudes of both penetrating through clouds and not requiring solar illumination of the land surface.

Recent atmospheric general circulation model (GCM) projections indicate an average global warming of the lower troposphere of 1-3.5^o C during the next century, depending on the greenhouse gas emissions scenario used [Intergovernmental Panel on Climate Change (IPCC), Houghton et al. 1996]. The IPCC adds that "all model simulations show . . . a maximum warming in high northern latitudes in winter." Evidence that high latitude warming has begun is accumulating. Analysis of local weather records across Alaska suggests that the growing season has been extended by over 14 days in the last 50 years by increasing spring temperatures (Figure 1).

High latitude warming may have amplified implications on global change rates because the impacts will not necessarily be linear. The state transition of the land surface from a frozen to a thawed condition is abrupt near 0^oC and initiates a number of terrestrial processes that are nearly dormant during frozen conditions. This state transition represents the closest analog to a biospheric and hydrologic on/off switch existing in nature, profoundly affecting surface meteorological conditions, ecological trace gas dynamics, and hydrologic activity.

Ku-band (2.1 cm) 25 km resolution NASA Scatterometer (NSCAT) data were applied to examine the feasibility of spaceborne radars for operational mapping of circumpolar freeze/thaw cycles for the northern high latitudes [McDonald et al., 1998]. NSCAT mosaics were used to derive landscape scale freeze/thaw maps of Alaska during spring 1997. These were compared with interpolated surface temperature records (Figure 2).

The NSCAT temporal response shows a 3 to 5 dB shift in measured backscatter, well correlated with landscape springtime thaw processes. The NSCAT-based maps clearly show the progression of the landscape from an initial frozen to a thawed state. Field measurements of surface temperatures reveal that the scatterometer responds to a combination of vegetation, soil and snow thawing. The large range of seasonal and interannual variability of the boreal frozen surface and its broad-ranging impact on regional and even global climate, hydrology, and biogeochemistry would suggest that regular, accurate monitoring be a priority in global change research. However, the density of reporting surface weather and hydrologic stations in sparsely populated high latitudes is very low, on the order of 1 per million km², especially in Canada and Siberia. Current optical satellite data suffer from cloud contamination and seasonal illumination problems, while passive microwave data have a very coarse (10-25 km) spatial resolution. At present no adequate monitoring system is in place to document high latitude climate change.

Boreal Ecological Principles

As early spring air temperatures rise above freezing, a point is reached where the surface soil or snowpack reaches 0^oC and thaws, with ice changing to liquid water (Figure 3). This onset of the spring freeze/thaw transition period initiates snowmelt, which immediately accelerates runoff and stream discharge. Ecosystem responses are equally rapid, with soil heterotrophic respiration (CO₂ flux from soil decomposition activity) and photosynthetic activity of evergreen trees also accelerating with the new presence of liquid water. The thaw transition period can last 1-8 weeks, as the snowpack continues to melt away, keeping soil thawed but with temperatures near 0^oC.

The final step of this state transition is complete when the snowpack is depleted. The surface albedo or reflectivity abruptly drops from 50-80% over snow to 10-20% over vegetation. Absorption of this additional net solar radiation causes surface soil and air temperatures to rise dramatically, often 3-5^oC in a week. Increasing air temperatures, soil temperatures, and day length finally induce vegetation out of dormancy and active growth of new leaves and shoots begins. This explanation of the progression of ecosystem processes does, however, oversimplify the temporal and spatial inconsistencies of the real world.

The land surface may thaw and then refreeze several times before final snowmelt. Spatial heterogeneity in the timing of freeze/thaw processes is also evident regionally, especially in complex topography with variable drainage, landcover, slope, and aspect. Because of both the biophysical importance of the freeze/thaw transition, and the extreme heterogeneity exhibited in this process, satellite-based monitoring is the only possibility for quantifying this transition across the whole circumpolar boreal region.

Boreal Carbon Flux Dynamics

Keeling et al. [1996] found that the annual amplitude of the seasonal CO₂ cycle in the Arctic had increased by 40% since the 1960s. Their interpretation of the seasonal phasing of the atmospheric CO₂ record was that the growing season in the Arctic may now begin 7 days earlier than 30 years ago. Myneni et al. [1997] corroborated Keeling et al.'s conclusion looking at an entirely different data set, the seasonal normalized difference vegetation index (NDVI) record from the advanced very high resolution radiometer (AVHRR) sensor for the period 1981-1991. Myneni et al. concluded that growing seasons of high latitude terrestrial ecosystems may have increased by 12 days during the period of study. Surface temperature records for weather stations with long (>40 years), consistent data sets now show evidence of an increasing trend toward longer growing seasons for northern high latitude regions (Figure 1). Eddy covariance carbon flux measurements have shown enhanced carbon uptake associated with earlier spring thaws in boreal forest stands. Goulden et al. [1998] found that a black spruce stand in northern Manitoba oscillated between being a carbon sink (-0.1 MG C/ha/yr from October 1996 to October 1997) and a source (+0.7 MG C /ha/yr from October 1994 to October 1995). Goulden et al. also found that the flux of CO₂ from the soil increased 10-fold as soil

thawed from -2 to 5^oC. The timing of spring thaw and the duration of the growing season are strongly linked to the carbon balance of boreal and arctic systems. Frolking et al. [1996] evaluated the range of interannual variability of boreal forest growing season length and associated impacts on carbon and water fluxes. They found that earlier spring thaws lead to significant increases in simulated net carbon uptake (Figure 4). Soil temperature simulations from 1976 to 1996 for boreal forest stands show 6-7 week ranges in the timing of snowmelt and soil thaw at 3 cm, equivalent to a year-to-year change in growing season length of 30%. At the Boreal Ecosystem-Atmosphere Study (BOREAS) southern aspen site, the transition of the ecosystem from a carbon source (~+25 kg C/ha/d) to its maximum rate of carbon uptake (-75 kg C/ha/d) occurs over a 10-day period in the spring. At the BOREAS northern and southern black spruce sites, the ecosystem's transition from a carbon source to a sink is coincident with snow melt and soil thaw [Goulden et al., 1998].

Arctic Snow Hydrology Dynamics

The Northern Hemisphere area of snowcover ranges from nearly 50 million km² in midwinter to 4 million km² in August, according to current estimates from coarse resolution satellite data. The interannual range in monthly snowcover can easily be 10 million km², especially during the initial snowpack accumulation in the autumn and snowmelt season in the spring [Robinson et al., 1993]. Recent trends in boreal snowcover also suggest a warming trend. Groisman et al. [1994] found that the mean annual snowcover of the Northern Hemisphere has retreated by 10%, or 2.4 million km², between 1972 and 1992. They noted a clear relationship between El Nino years and reduced snowcover extent. This temporal range in areal extent of snowpack and timing of spring melt has a large impact on seasonal river discharge and on land surface energy exchange characteristics influencing regional and even global weather patterns.

Flooding is often accentuated at high latitudes when rapid snowmelt occurs over frozen soils that cannot absorb the moisture, accelerating runoff rates. Due to the Arctic Ocean's relatively small size, the ratio of freshwater inputs to ocean basin volume is about 10 times the mean global ratio, causing large seasonal variations in salinity. Major rivers in Eurasia discharge about one third of their annual runoff into the Arctic Ocean during the month of peak flow following snowmelt.

Weather Forecasting

Teleconnections exist from boreal land surface biophysics to global weather and climate. Betts et al. [1998] found that models from the European Centre for Medium-Range Weather Forecasts and the National Centers for Environmental Prediction/National Center for Atmospheric Research overestimate springtime latent energy fluxes because they do not account correctly for frozen surfaces.

This problem can propagate to errors in the 5-day forecasts of up to 5^oC in lower tropospheric temperatures. Not only are near-surface temperatures affected by these budgets, but the influence of land surface energy budgets extends throughout the planetary boundary layer and above. Groisman et al. [1994] estimated that the observed 10% reduction in northern hemispheric spring snowpack from 1972 to 1992 resulted in an average 2.7 W m² increase in surface net radiation that would correspond roughly to a 1.5^oC increase in spring air temperatures at high latitudes, very close to the magnitude observed in Figure 1.

Freeze Thaw Detection via Radar

The microwave backscatter signature of a landscape is controlled by the landscape's structure and dielectric properties. The interaction of an electric field with a dielectric material has its origin in the response of charged particles to the applied field. The displacement of these particles from their equilibrium positions gives rise to induced dipoles that respond to the applied field. In addition, polar materials contain permanent dipoles caused by the asymmetric charge distribution within the molecules themselves. Liquid water consists of highly polar hydrogen-oxygen molecules exhibiting a dielectric constant that dominates the microwave response of natural landscapes. As water freezes, the molecules become bound in a crystalline lattice, and the dielectric constant decreases substantially.

For vegetated landscapes that undergo freeze/thaw transitions, this drop in dielectric constant results in a large backscatter shift. In the late 1980s researchers using truck-mounted scatterometers discovered that radar backscatter from frozen ground and frozen vegetation was 3-4 dB lower than that of thawed ground and vegetation. Freeze/thaw backscatter change was first observed in radar image data from a series of L-band (23.5 cm) aircraft data sets acquired in March 1988 over the Bonanza Creek Experimental Forest, a long-term ecological research site near Fairbanks, Alaska. During this time period, temperatures ranged from unseasonably warm (up to 9^oC) to well below freezing (-8 to -15^oC), where liquid water in the vegetation froze, resulting in a 4 to 6 dB shift in radar backscatter relative to nonfrozen conditions. Similar radar backscatter change detection methods were used to assess spatial and temporal freeze/thaw patterns using the European C-band (5.7 cm) satellite imaging radar ERS-1 temporal series data within the BOREAS southern study region in central Canada [Way et al., 1997]. Radar observed freeze/thaw transitions correlated well with the initiation of vegetation respiration and photosynthesis.

The NSCAT sensor platform ceased functioning in June 1997. Current satellites with radar capabilities can detect freeze/thaw transitions when the signal is processed for that purpose, but do not have a sufficiently high temporal repeat cycle or spatial resolution to provide the regular monitoring suggested here. Design is under way for development of a

proposed new radar sensor with moderate spatial resolution (1 km) and 2-3 day repeat coverage for all northern latitudes above 40 degrees to overcome current limitations in data spatial coverage and repeat frequency. Optical sensors, such as the current AVHRR and the future EOS Moderate-Resolution Imaging Spectroradiometer are superior for tracking vegetation development during the growing season, using algorithms such as NDVI. However, at high latitudes cloud cover routinely limits their true repeat cycle to 10-14 days, and early spring sun angles are too low to provide sufficient reflected solar energy. Consequently, satellite-deployed radar for monitoring trends in boreal land surface processes may have increased scientific and policy significance in the future.

More details, including full season time animations of the results in Figure 2, can be found on the Web (http://frosty.jpl.nasa.gov)

Acknowledgment

This research is funded by the Office of Earth Science of NASA.

Authors

S. W. Running, J. B. Way, K. C. McDonald, J. S. Kimball, S. Frolking, A. R.

Keyser, and R. Zimmermann. For more information, contact S. W. Running, Numerical Terradynamic Simulation Group, School of Forestry, University of Montana, Missoula, MT 59810, USA; E-mail: swr@ntsg.umt.edu.

FIGURES:

Fig. 1. The increase in growing season length over the last 50 years averaged for eight stations in Alaska (Barrow, Bettles Field, Fairbanks, Big Delta, Anchorage, Talkeetna, Yakutat, Juneau) having the longest and most consistent temperature records. Standard surface weather station records were used to compute a growing season defined as the cumulative number of days each year with daytime average temperatures >5^oC.

Fig. 2. Comparison of maximum air temperature with freeze/thaw state maps derived from 2-day NASA Scatterometer (NSCAT) composite mosaics for 4 days in March and April 1997. The surface temperature maps were spatially interpolated with measurements from 72 meteorological stations in Alaska. The NSCAT images used a 4-dB shift in backscatter to define the freeze/thaw transition. The bottom four graphs show temporal series of the NSCAT backscatter at four locations along a north-south transect extending from Toolik Lake on the north slope of the Brooks Range to Coldfoot, Alaska, near the northern limit of the boreal forest, through the Bonanza Creek Experimental Forest in the central interior to Denali National Park in the Alaska Range. Vegetation temperature measurements obtained in situ at four sites are compared with NSCAT backscatter extracted from 50-km regions centered at each of the ground stations. The broken vertical lines mark the times initiating the 2-day NSCAT composites. Comparison of in situ data with normalized backscatter reveals that the scatterometer observes a combination of vegetation and snow thaw as conditions in the Alaskan Interior progress from a short warming trend, through a brief re-freezing transition, and to the final initiation of springtime thaw and progression to a snow-free state.

Fig. 3. Summary conceptual diagram of the temporal sequence of surface biophysical processes that occurs during the freeze/thaw state transition period in boreal ecosystems.

Fig. 4. Examples of the interannual variability in boreal surface thawing. a) Snowpack depth measured at the Thompson, Manitoba, airport weather station (operated by Canadian Atmospheric Environment Service) for a spring with an early melt date (1980), an intermediate melt date (1994), and a late melt date (1983). b) Daily net ecosystem exchange (NEE) of carbon dioxide simulated for a black spruce forest stand near Thompson, Manitoba [Frolking et al., 1996], for the same 3 years. Positive NEE represents carbon uptake by the forest; negative NEE represents net ecosystem respiration. In all 3 years there is low ecosystem respiration until the snowpack melts. Following snowmelt, these evergreen ecosystems rapidly shift from a net carbon source to a net carbon sink. An early spring thaw leads to significant enhancement in annual carbon sequestration by these forests.

References

Betts, A. K., P. Viterbo , A. Beljaars, H-L. Pan, S-Y. Hong, M. Goulden, and S. Wofsy, Evaluation of land-surface interaction in ECMWF and NCEP/NCAR reanalysis models over grassland (FIFE) and boreal forest (BOREAS), J. Geophys. Res., 103, 23,079-23,085, 1998.

Frolking, S., et al., Modelling temporal variability in the carbon balance of a spruce/moss boreal forest, Global Change Biol., 2, 343-366, 1996.

Goulden, M. L., et al., Sensitivity of boreal forest carbon balance to soil thaw, Science, 279, 214-216, 1998.

Groisman, P. Y., T. R. Karl, and R. W. Knight, Changes of snow cover, temperature and radiative heat balance over the Northern Hemisphere, J. Clim., 7, 1633-1656, 1994.

Houghton, J.T. et al. (eds.) Climate Change 1995: The Science of Climate Change. Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge. 1996.

Keeling C. D., J. F. S. Chin, and T. P. Whorf, Increased activity of northern vegetation inferred from atmospheric CO₂ measurements, Nature, 382, 146-149, 1996.

McDonald, K. C., J. Kimball, R. Zimmermann, J. B. Way, S. Running, and S. Frolking, Mapping Seasonal Freeze/Thaw Processes in Alaska with NSCAT, Paper presented at the AGU Fall Meeting, San Francisco, Calif., Dec. 6-10, 1998.

Myneni R. B., C. D. Keeling, C. J. Tucker, G. Asrar, and R. R. Nemani, Increased plant growth in the northern high latitudes from 1981-1991, Nature, 386, 698-702, 1997.

Robinson, D. A., K. F. Dewey, and R. R.Heim, Global snow cover monitoring: An update, Bull. Am. Meteorol. Soc., 74, 1689-1696, 1993. Way, J. B., R. Zimmermann, E. Rignot, K. McDonald, and R. Oren, Winter and spring thaw as observed with imaging radar at BOREAS, J. Geophys. Res., 102, 29,673-29,684, 1997.