Day 1 Question 1: Feedbacks that complicate climate change prediction

Question 1:

List and describe some of the demographic, atmospheric, oceanic, and land surface feedbacks that complicate attempts to predict future climate change (associated with rising carbon dioxide levels). Explain how scale constrains our ability to detect these feedbacks, model them, and integrate them into predictive scenarios. Devise a Stommel diagram(s) to illustrate the scales at which these feedbacks play out.

Interactions between various parts of the climate system create complex feedbacks and add to the overall complexity of the entire climate system (Rial et al. 2004). Due to these complexities, and the vast number of feedbacks and variables in the climate system, it is nearly impossible to model the future climate scenario with great certainty. Feedback mechanisms can be organized into several categories, including socio-economic, ecological, climatic, and atmospheric.

Among the socio-economic feedbacks, some of which are the most uncertain, are changes in energy use technologies, changes in energy production technologies, and land-use changes. Among ecological feedbacks are the acclimation of plants, soil microorganisms, and algae and plankton to warmer temperatures and high CO2 concentrations, and land cover and vegetation changes with changes in temperature and precipitation. Climatic feedbacks include the air temperature response to higher CO2 concentrations, albedo feedbacks with the response of glaciers, ice caps, and sea ice to climate change, sea level response to climate change, the release of greenhouse gasses from the thawing of frozen peat bogs, and the transfer of heat and CO2 into the ocean. Within the atmosphere, several feedbacks exist, including the mixing of greenhouse gasses, and changes in dust and aerosol concentrations in the atmosphere. Selected feedback mechanisms are illustrated in a time-space diagram below.

Each system has its own set of complex feedbacks and responses, as well. For example, as temperatures rise, and more melt occurs on the surface of an ice sheet, meltwater from the surface can percolate down through crevasses and moulins in the ice to reach the surface. As the water pressure at the base of a glacier increases, the glacier may actually float on the water and slide along the now lubricated base. This enhances the flow and dynamics of the glacier, but cannot easily be predicted.

As glaciers melt and contribute more freshwater to the oceans, other changes can occur. Higher sea levels allow water to seep further beneath tidewater glaciers, also allowing these to slip and surge. These feedbacks are extremely important, since when glaciers melt, they leave behind a material which is always less reflective than the snow or ice surface itself. Consequently, warmer conditions tend to lead to melting of snow and ice, which in turn leads to more radiation being absorbed by the surface, which ultimately leads to more snow and ice melt. This is an example of a positive feedback, often referred to as the albedo feedback.

Furthermore, glaciers contribute freshwater to the oceans, particularly in the North Atlantic, where temperature and salinity drive the thermohaline circulation, or THC, the major conveyor of heat and water in the world’s oceans, and therefore a major driver of global climate. As freshwater enters this system in the North Atlantic, the water becomes more buoyant and sinks more slowly. It is that sinking that drives the entire conveyor. Therefore, a freshening of the North Atlantic is expected to cause the THC to slow down. This would then drive less warm water northward along the Gulf Stream current, shifting Europe into a much cooler climate. This cooler climate could, potentially, allow for the flourishing of alpine glaciers, adding to the planet’s reflectivity, and ultimately dampening the effects of fossil fuel consumption.

On the other hand, the slowdown of the THC also leads to less upwelling of nutrients in the ocean. This ultimately leads to less benthic plant life, which is an enormous carbon sink. Such complexities and non-linearities in the climate system preclude the certain prediction of future climates.

Figure 1: Time-space Stommel diagram of some feedbacks in the climate system.

Day 1 Question 2: Water resource issues with climate change

Question 2:

First, describe some of the anticipated changes in the distribution and amounts of snowfall and rainfall under a CO2-enriched atmosphere. Second, identify two locations where you might anticipate (or presently observe) large changes in municipal water supplies derived from melting snow and ice. What kind of a signal might be apparent in stream gauge data to indicate these changes? (Assume you have a near perfect data set with information about daily, monthly, and annual discharges from the past 50 years). Are all urban areas that are dependent upon meltwater going to have similar signals? How might this signal vary geographically and why?

Introduction

An abundance of recent evidence suggests that 20th-century warming was unprecedented in the late Holocene and 21st-century warming is projected to be much greater than the natural variability exhibited by the climate system over the past 1000 years (Dyurgerov and Meier, 2000; IPCC, 2001, Thompson, 2000). Observations have identified a general thinning and retreat of glaciers globally (e.g. Dyurgerov and Meier, 2000; Oerlemans, 1994; Thompson et al., 2000). The retreat of the termini of alpine glaciers worldwide contributes to the growing evidence of rapid global climate change (Thompson et al., 1993). In this warming climate, with the rapid disintegration of glaciers, freshwater availability is expected to become a forefront issue for many regions.

Anticipated changes in precipitation in a warming climate

In a warming climate, as predicted by General Circulation Models (IPCC, 2001), the hydrological cycle will be enhanced, and the current distribution of precipitation will change significantly in regional distribution and seasonality (Benniston, 2003). Model results suggest higher evaporation rates and greater proportions of liquid precipitation in the coming decades. These physical and seasonal differences will affect soil moisture, groundwater reserves, and the frequency and intensity of droughts and floods. The most likely precipitation-related change with an enhanced greenhouse effect is more intense precipitation events happening over many regions (IPCC, 2001).

The increased evaporation rate, coupled with an expected increased water holding capacity throughout the atmosphere will also lead to an increase in the water vapor concentration throughout the atmosphere. This is important as water vapor is the most important and strongest of all greenhouse gasses, and may lead to intensified warming. As some of this water vapor condenses into clouds, the net effect on Earth’s radiation balance is unknown. High, thin clouds are transparent to solar radiation, but trap in Earth’s thermal radiation, and have a warming effect like other greenhouse gasses. On the other hand, low, thick clouds reflect solar radiation and cool the atmosphere.

Generally, precipitation is expected to increase over the globe, although this increase will not be uniformly distributed spatially or temporally; a significant portion will be caught by the oceans. While there has been a statistically significant 2% increase in precipitation over land in the last 100 years, precipitation over the U.S. has increased by 5 to 10%. In the last 50 years, however, China’s precipitation totals have decreased (IPCC, 2001).

Bhaskharan and Mitchell (1998) found that in a warming climate, the south Asian monsoon region will change, bringing about increased precipitation in the eastern regions and decreased precipitation in the west. This has tremendous implications for the western China, where semi-arid conditions already push the limits of the available water.

These changes come about in a climate of expanding human population and activities. With our growth and industrialization, we have become increasingly reliant upon water, including for the discharge of our wastes. A recent decline in water quality and availability has driven recent concerns for future water availability (Shiklomanov, 2001).

As mountains are the source of over half the world’s rivers, precipitation changes in a single watershed are not only likely to affect the mountain regions where the precipitation falls and is stored in snow and ice, but the more populated lowland areas that critically rely on the watershed as a resource will also be greatly affected (Benniston, 2003). In arid and semi-arid areas, mountains provide up to 100% of the available freshwater to the surrounding lowlands (Meybeck et al., 2001). As important as the precipitation regimes in mountains are, many fear that global climate models are too coarse to capture the orographic complexity of mountain regions. Others argue that the overall regional patterns of precipitation in the models will still likely hold in mountain regions.

Snow and ice are important components of the hydrological cycle in mountain regions, particularly the seasonality and amount of runoff (Haeberli and Hoelzle, 1995). Changes in regional snowfall are expected with global warming. In temperate regions, where the snow-pack is close to melting conditions, a small increase in temperature may have a significant change on the local mass-balance, and therefore the meltwater runoff. With enhanced warming, temperate regions will experience an increase in the frequency of rain, and the snowline will rise by an expected 150 m/̊C of warming (Beniston, 2003).

Two locations which might experience large changes in water supply from melting snow and ice

Glaciers are often the most the most important freshwater resources in arid regions. For example, glaciers are critically important sources of freshwater in the hyper-arid coastal regions of Peru in the tropics, and the semi-arid regions of western China in the mid-latitudes. These two areas may be the most important in terms of the hydrological response to glacier retreat, and are currently the focus of new studies regarding the connections between glacier melt and catchment runoff (Evans, 2001).

The meltwater runoff from a glacier depends on the glacier’s mass and energy balances at the surface. The mass balance of a glacier depends on mass gained, “accumulation”, through snow and deposition, and mass lost, “ablation”, through melting, iceberg calving, and sublimation. The amount of melt at the surface is determined by the energy balance. One of the key factors of the energy balance is the reflectivity, “albedo”, of the surface. The amount of sublimation greatly depends on the humidity of the air, an often overlooked variable in glacier mass-balance studies.

In a warming climate, the expected increases in precipitation are not expected to outweigh the increased melting predicted on glaciers except, possibly, in Antarctica (IPCC, 2001). Precipitation is important to a glacier’s mass balance, not just through the direct addition of mass, but indirectly through the enhancing of the surface albedo. Fresh snow has the highest albedo of any natural surface (Paterson, 1994). Adding fresh snow to a glacier surface can temporarily shut down melt as excess energy is reflected away. This effect can be up to ten times more important than the effect of the mass directly added by precipitation. Since the albedo of the surface does not depend on the thickness of the snow layer, this albedo effect depends more on the timing and frequency of snowfall, rather than the total amount. In contrast, as a greater balance of liquid precipitation is expected on temperate glaciers, the rainfall tends to add tremendous energy to the glacier surface, and in the case of rain on snow events, can significantly reduce the albedo.

Peru

Coastal Peru is a hyper-arid desert, yet it sustains a population of approximately 28 million people (CIA World Fact Book, 2005). This is possible only through the heavy reliance on glacier meltwater originating high in the Andes. Peru has more than its share of glaciers, being home to 70% of the world’s tropical glaciers (Kaser et al., 1996).

Being located in the tropics holds special importance for the glaciers of Peru. The tropics comprise nearly half the surface area of the Earth, are home to 70% of its population, and receives the bulk of solar insolation. Excess solar energy is then transported to the rest of the globe. Accordingly, characterization of climate change in the tropics is of extreme importance.

Furthermore, temperature in the tropics varies less horizontally than outside the tropics because the Coriolis effect is weakest there. In the tropics, pressure differences caused by temperature variations are eliminated by redistribution of mass that the Coriolis effect tends to limit outside the tropics. Although temperature is not horizontally homogeneous, it is much more uniform than any of the other meteorological variables which effect tropical glacier mass balance, including precipitation, humidity, and cloudiness. Accordingly, a secular retreat of tropical glaciers suggests a shift in temperature, rather than in some other atmospheric characteristic.

Strong retreat rates of tropical glaciers have been reported for recent decades from the high Andes and elsewhere in the tropics (Kaser, 1999; Peterson and Peterson, 1994; Ramirez et al., 2001; Thompson et al., 2002). In fact, all small glaciers and ice caps outside of the polar regions are expected to be completely absent by the end of this century (IPCC, 2001). The melting of these glaciers may account for as much as half of the expected rise in sea level over the next 100 years, the rest of which will be caused by thermal expansion (IPCC, 2001). Since glaciers act as natural reservoirs that provide stored water to people in the region for hydroelectricity and agriculture during the dry season (Paterson, 1994; Thompson, 2000), the demise of these ice caps has the potential to create significant social and environmental repercussions. The glaciers of Peru provide critical water resources to the hyper-arid coastal lowlands and high altiplano. They are also the source of glacier-related avalanches and floods that have leveled towns and caused over 35,000 recorded deaths (Williams and Ferrigno, 1999).

Although there is significantly more ice in high latitudes, tropical ice fields are temperate, generally existing closer to melt-threshold conditions and, accordingly, relatively small climate changes may significantly affect their mass balance (Martinson et al., 1998). The tropics host a number of climatic phenomena that greatly affect humans, including the El Niño - Southern Oscillation, the Inter-Tropical Convergence Zone, and the Asian monsoon, are characterized by relatively homogeneous thermal conditions, and are not significantly affected by migratory synoptic disturbances. Accordingly, fluctuations on tropical glaciers may be more directly related to secular climate changes (Kaser and Georges, 1999). These factors make tropical glaciers exceptional indicators of recent environmental changes and the effects of these changes on natural systems (Oerlemans and Fortuin, 1992).

Tropical glaciers have received much recent attention as their importance as water resources and as early indicators of climate change has recently been recognized (Georges, 2004). These glaciers are also of tremendous concern on a regional scale, as Peru exhibits a long, hyper-arid annual dry season during which time melt runoff from glaciers is a critical resource for the economic stability of the country. Peru relies almost entirely on this meltwater for its electricity production, which is over 90% hydroelectric, its drinking water, and for its agriculture, which is their primary economic activity.

Like the people and animals in Peru, the glaciers there are at the absolute threshold of existence. Accordingly, they are one of the earliest and most sensitive indicators of climate change. Melting glaciers not only pose a threat to water resources, but many rapidly melting glaciers form dangerous proglacial lakes that are prone to sudden outburst floods, especially in the steep, tectonically active terrain found in the Andes (Mark, 2002). Worse yet, the political and economic instability there, and a lack of institutional support precludes any possibility of large-scale engineering solutions.

The present contribution of meltwater to the regional runoff is large and well-documented. Monthly measurements of discharge from the Cordillera Blanca, Peru’s most heavily glaciated area, exhibit maxima in Austral Spring, prior to the maximum in precipitation. This suggests a large contribution from glaciers as the net radiation is high, surface albedo is low, and perhaps most importantly, relative humidity is high. Humidity is an often overlooked variable in glacier mass balance. In a dry environment, much of the excess energy receipt at a glacier surface will be utilized for sublimation, which is 8.5 times less effective as melting.

All other things equal, with global warming, we expect the contribution of glacier meltwater to initially increase. This has been witnessed over the last few decades, and has allowed Peru’s population and economy to grow. Presently, Peru is one of the fastest growing nations in the world. As warming continues, however, the rate of meltwater production will decrease as glaciers diminish in size, placing intense strain on an already limited system.

China

In China, the direction of water resources there is less well known. One study by Mirza (1997) suggests that changes in sub-basin drainage in the Ganges could range from 27 to 116% with a doubling of CO2 concentrations.

Of the changes in precipitation expected in a changing climate, few would be as significant as a shift in the Asian monsoon system. Throughout the Holocene, the Asian monsoon has faltered in times when the North Atlantic climate was cooler (Gupta et al., 2003). These changes have tended to occur abruptly, often coinciding with Dansgaard-Oeschger events. If glaciers continue to melt, driving freshwater into the North Atlantic, the thermohaline circulation (THC) is expected to slow, resulting in a cooler climate for the North Atlantic. This could result in more frequent failure of the Asian monsoon, which would be devastating to much of the world’s population.

There has been a widespread retreat of Himalayan glaciers. Since 1955, glaciers in the Tien Shan mountains have retreated by over 22% (Meier et al., 2003). Interestingly, however, glaciers in the Qilian Shan also retreated while temperatures there decreased and precipitation increased (Liu et al., 2003). This retreat is likely due to changes in the timing and variability of temperatures and precipitation events. This highlights an important aspect of glaciers, namely that glaciers thrive in stable conditions and tend to retreat in more variable conditions. As increased variability is predicted to accompany global warming, glaciers may be doomed even where precipitation increases.

Precipitation in the Qilian Shan is regulated by the Asian monsoon, westerly winds, intense storms in the summer, periodic rains in the spring and fall, and rare events in the winter (Liu et al., 2003). Meltwater from glaciers in the region account for 39 to 56% of the total runoff, and therefore holds great social and economic importance.

Future changes in the timing, type, and amount of precipitation falling in the Andes and in China will have a tremendous impact on the seasonal availability of water to the surrounding lowland populations, and therefore to the economic stability of those nations. As increased pressure on water resources in politically unstable regions can lead to upheaval and armed disputes, climate change has the potential to not only put increased economic pressure on a nation, but can ultimately lead to war.

Geographic variability

The consequences of a changed hydrological cycle vary from place to place, depending on the degree of reliance on runoff as a water resource, the seasonality and type of precipitation relied upon, and the changes these regions experience in their precipitation distribution, seasonality, and type.

The changes in glacier melt runoff in a warming climate do not vary linearly with climate. Instead, they depend also on the size of the glacier, the rate of temperature rise, and other factors which influence the glacier mass balance, including changes in precipitation rates and type with elevation, changes in the elevation temperature profile, changes in the seasonality of temperature and precipitation, and the mass balance distribution with elevation. Typically, however, as climate warms and glaciers retreat, the runoff tends to increase at first, but then decrease as the size of the retreating glacier diminishes (Ye et al., 2003). Larger glaciers tend to retreat faster, while smaller glaciers tend to lose area faster than length. For all glaciers, however, volume exhibits the most significant change in a warming scenario. Smaller glaciers with higher runoff peaks tend to exhibit the most variability in runoff and retreat more quickly than larger glaciers. Also, the faster the temperature rises, the runoff peaks become earlier and higher (Ye et al., 2003).

Water resources in some regions are highly sensitive to the frequency, intensity, and trajectory of tropical cyclones, monsoon systems, and other traveling synoptic patterns. For example, the water supply in southern and tropical Asia, where nearly half of the world’s population live, is greatly influenced by tropical cyclones (Benniston, 2003). These areas may ultimately become more highly dependent on seasonal storms and rainfall patterns in a time when they are predicted to become more variable.

Not all regions dependant on glacial melt will experience water shortages, but that appears to be the prevailing tendency. Even in northern Sweeden, where temperatures, like Antarctica, are low to begin with, studies by Schneeberger et al. (2001) of the Storglaciären in northern Sweeden suggest that increased ablation there will greatly outweigh the expected increases in precipitation.

Bhaskharan, B., and J. F. B. Mitchell (1998), Simulated changes in the intensity and variability of the southeast Asian monsoon in the twenty first century resulting from anthropogenic emission scenarios, International Journal of Climatology, 18, 1455-1462.

Dyurgerov, M. B., and M. F. Meier (2000), Twentieth century climate change: Evidence from small glaciers, Proceedings of the National Academy of Sciences, 97, 1406-1411.

Evans, D. J. A. (2001), Glaciers, Progress in Physical Geography, 25, 428-439.

Georges, C. (2004), 20th-century glacier fluctuations in the tropical Cordillera Blanca, Peru, Arctic, Antarctic and Alpine Research, 36, 100-107.

Gupta, A. K., et al. (2003), Abrupt changes in the Asian southwest monsoon during the Holocene and their links to the North Atlantic Ocean, Nature, 421, 354-357.

Haeberli, W., and M. Hoelzle (1995), Application of inventory data for estimating characteristics and regional climate change effects on mountain glaciers: a pilot study with the European Alps, Annals of Glaciology, 21.

Kaser, G. (1999), A review of the modern fluctuations of tropical glaciers, Global and Planetary Change, 22, 93-103.

Kaser, G., and C. Georges (1999), On the mass balance of low latitude glaciers with particular consideration of the Peruvian Cordillera Blanca, Geografiska Annaler, 81 A, 643-651.

Kaser, G., et al. (2003), The impact of glaciers on the runoff and the reconstruction of mass balance history from hydrological data in the tropical Cordillera Blanca, Peru, Journal of Hydrology, 282, 130-144.

Kaser, G., et al. (1996), Mass balance profiles on tropical glaciers, Zeitschrift für Gletscherkunde und Glazialgeologie, 32, 75-81.

IPCC (2001), Climate Change 2001: The Scientific Basis, Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), 944 pp., Cambridge University Press, Cambridge, UK.

Mark, B. G. (2002), Hot ice: glaciers in the tropics are making the press, Hydrological Processes, 16, 3297-3302.

Martinson, D. G., et al. (1998), Decade-to-Century-Scale Climate Variability and Change: A Science Strategy, 142 pp, Panel on Climate Variability on Decade-to-Century Time Scales, Board on Atmospheric Sciences and Climate, Commission on Geosciences, Environment and Resources, National Research Council, National Academy Press, Washington, D.C.

Meier, M. F., et al. (2003), The health of glaciers: recent changes in glacier regime, Climatic Change, 59, 123-135.

Meybeck, M., et al. (2001), A new typology for mountains and other relief classes: an application of global continental water resources and population distribution, Mountain Research and Development, 21, 34-45.

Mirza, M. Q. (1997), The runoff sensitivity of the Ganges River basin to climate change and its implications, Journal of Environmental Hydrology, 5, 1-13.

Oerlemans, J. (1994), Quantifying global warming from the retreat of glaciers, Science, 264, 243-245.

Oerlemans, J., and J. P. F. Fortuin (1992), Sensitivity of glaciers and small ice caps to greenhouse warming, Science, 258, 115-117.

Paterson, W. S. B. (1994), The Physics of Glaciers, 3rd edition ed., 480 pp., Elsevier, Yarrytown, NY.

Peterson, J. A., and L. Peterson (1994), Ice retreat from the neoglacial maxima in the Puncak Jayakesuma area, Republic of Indonesia, Zeitschrift für Gletscherkunde und Glazialgeologie, 30, 1-9.

Ramirez, E., et al. (2001), Small glaciers disappearing in the tropical Andes: a case-study in Bolivia: Glaciar Chacaltaya (16° S), Journal of Glaciology, 47, 187-194.

Schneeberger, C., et al. (2001), Modeling the response of glaciers to a doubling in atmospheric CO2: a case study of Storglaciären, northern Sweeden, Climate Dynamics, 17, 825-834.

Shiklomanov, I. A. (Ed.) (2001), World Water Resources at the Beginning of the 21st Century, UNESCO Publications, Paris, France.

Thompson, L. G. (2000), Ice core evidence for climate change in the tropics: implications for our future, Quaternary Science Reviews, 19, 19-35.

Thompson, L. G., et al. (1993), "Recent warming": ice core evidence from tropical ice cores with emphasis on Central Asia, Global and Planetary Change, 7, 145-156.

Thompson, L. G., et al. (2000), Ice core paleoclimate records in tropical South America since the Last Glacial Maximum, Journal of Quaternary Science, 15, 377-394.

Thompson, L. G., et al. (2002), Kilimanjaro ice core records: evidence of Holocene climate change in tropical Africa, Science, 298, 589-593.

WGMS (1989), World Glacier Inventory, Status 1988, International Association of Scientific Hydrology (ICSI) - UNEP - UNESCO.

Williams, R. S., Jr., and J. G. Ferrigno (Eds.) (1999), Satellite Image Atlas of Glaciers of the World, United States Government Printing Office, Washington.

Ye, B., et al. (2003), Responses of various-sized alpine glaciers and runoff to climate change, Journal of Glaciology, 49, 1-7.

Day 1 Question 3: How human and physical geographers critique global climate models

Question 3:

Explain the differences in how human and physical geographers critique global climate models.

Introduction

According to Rhoads (1999), while physical geographers do engage in debates of methodology and interpretation of scientific data and results, they seldom engage in philosophical discourse. While I agree with his notion that there is much to gain from engaging in such rhetoric, I, myself, am not often drawn to initiating such discussions. In fact, I tend to steer quite clear of them at meetings, perhaps afraid of the “impenetrable jargon” (Schneider, 2001), unless those involved are among those colleagues who I hold in the highest regard.

Chief among the opportunities for physical geographers to engage in any sort of philosophical discourse is when discussing the validity of models. Models, themselves, are highly interesting tools, which when applied correctly, can yield interesting scientific information. Most modelers are aware of the constraints of models, and are careful in their use and in the interpretation of their results. Others, however, are often guilty of using the results of models and treating them the same they would as real observations.

It is all to often, as Demeritt (2001) and Schneider (2001) agree, that model results, or the conclusions drawn from them, are touted by the media, and the model uncertainties and caveats are often disregarded. While the dangers of this are obvious, the opposite is often true when it comes to policy, as Schneider points out: “if we scientists don’t stop caveating everything so much, they won’t be able to get political support for strong action.”

Generally, I think that physical geographers look critically at models, especially their physics, sensitivities, and their abilities to simulate the real world. On the other hand, human geographers might draw into question the use of the model itself. In simpler terms, the physical geographer asks “Is this the best model, and can I trust its results?”, while the human geographer asks “Is it appropriate to use a model in this case?”

The case of Global Climate Models (GCMs)

GCMs are our best option for analyzing how the Earth’s climate might behave in the future given some forcing, such as increased greenhouse gas concentrations. Yet the climate system is highly complex, and therefore behaves chaotically, and non-linearly. Our climate history includes few instances of slow and gradual change, and instead has been punctuated by frequent abrupt changes. This suggest a highly chaotic and non-linear system. The Earth’s climate system depends on the energy balance and dynamics of the atmosphere, the hydrosphere, the biosphere, especially the interactions that plants, humans, and algae have with the atmosphere and hydrosphere, the cryosphere, and, in the case of volcanic out-gassing and eruptions, the lithosphere. Each of these five spheres on its own is highly complex, but the cris-crossing interactions between each of the spheres adds a whole new level of complexity. As such, trying to capture this whole system, including all of its complexities in a numerical model seems absurd at first.

Computer models, however, are not intended to capture every bit of complexity of the Earth systems. Instead, modelers attempt to boil this complexity down to its simplest and most significant parts. They create a model which can simulate the present climate, or the climate over the last 100 years. Then they run these models ahead another 100 years to get an estimate of what climate might do in the future. When they add certain forcings to the models, like increased CO2 concentrations, one hopes that the difference in the model results can tell us something about how those forcings might affect climate in the future.

There is a philosophical question that is still left up in the air regarding models and their results. If you create a model that represents climate, and you run this model into the future, you may be gaining little insight as to how the actual climate system may behave in the future, as the uncertainties, complexities, and feedbacks involved in the actual climate system extend far beyond what is coded in a model. Furthermore, any forcing behavior exhibited by the model is just that, forcing behavior exhibited by the model. There is no way to know if the actual climate system will behave in the same way.

But that is just it, isn’t it? There is no way of knowing how the climate system will react, but climate models may give us some clues.

Geographers who criticize models, and ask critical questions about the models and their results force the modelers to take stock in their models and to see where improvements can be made, to clarify the certainties of the model, and to illustrate any and all weaknesses and uncertainties. Schneider asserts that modelers tend to do this already, which begs the question “what can really be gained by these deconstructions?”

In conclusion, I assert that Human Geographers have no precise analog to a physical, numerical, or analytical model. They are stuck with statistical and empirical models, which in Physical Geography are often seen as second best. Accordingly, they belay their frustrations and misunderstandings by criticizing our tools and the knowledge we reap from them. How dare they sabotage our techniques? I think they are just jealous!

Rhoads, B. L. (1999), Beyond Pragmatism: The Value of Philosophical Discourse for Physical Geography, Annals of the Association of American Geographers, 89, 760-771.

Schneider, S. H. (2001), A constructive deconstruction of deconstructionists: a response to Demeritt, Annals of the Association of American Geographers, 91, 338-344.

Day 2 Question 1: Model cross validation

Question 1:

A method to obtain an assessment of how well a statistical forecast model will perform when predicting data that is not used in building the model is called cross validation. Explain the method of cross validation for estimating forecast skill of statistical forecast models. Explain why it is important. What are the difficulties and potential pitfalls in implementing such a procedure?

Introduction

Empirical models are based on strong statistical relationships existing between two or more physical properties, usually with the goal of forecasting some physical property into the future based on other related observations. The benefits of empirical models are that, in many cases, the results of simple statistically-based models are often as robust as those obtained from much more complex, physically-based models (Elsner and Schmermann, 1994).

For example, a common empirical model, the positive degree day model, is based on the strong statistical relationships between measured air temperature on a glacier and the rate of melt of the snow or ice surface (Braithwaite, 1995). As air temperature and surface melt are both the integrated result of the total surface energy balance, the success of temperature-based models is not surprising. Temperature-based methods are frequently employed because temperature data are easily obtained, the parameterization schemes are simple yet relatively accurate, and temperature data are easily interpolated between sites (Ohmura, 2001).

When creating a statistical or empirical model from observations or from a set of historical data, it is important to accurately understand how well the model performs in forecast mode. Determining the true forecast skill is one of the most critical aspects of developing such a model. Some of the common techniques for assessing forecast skill, such as the standard regression skill estimate, are systematically biased towards higher skills. One method, which is less biased than the hindcast skill estimates that are typically used, is the cross-validation method (Michaelson, 1987).

Overview of the cross-validation method

The cross-validation method is a statistical procedure for accurately assessing the forecast skill of empirical models. In cross-validation, cases from the dataset are systematically deleted, a new forecast model is derived from the remaining data, and then that model is used to “hindcast” the deleted data (Elsner and Schmermann, 1994). One strength of this method is the ability to diagnose individual datum which are especially influential on the results (Michaelson, 1987).

Potential pitfalls

To accurately assess model skill with the cross-validation method, there are three important conditions which must be met. First, hindcasts must be made using only out-of-sample data. Using in-sample data underestimates the model prediction error by only considering the fit of the prediction rule to the data. Secondly, the prediction rule for each hindcast should be made from only the remaining data and not include the datum that is being predicted. This is an essential condition of true cross-validation. Finally, there may be no statistical auto-correlation, or serial correlation, between the omitted data and the remaining data used for the prediction.

Concluding remarks

Models provide an ideal testing ground for our understanding of important physical processes (Cassano et al., 2001). They range widely in their complexity and application. Empirical models are becoming more common as the algorithms and technology to implement them are becoming more readily available. Accordingly, consistent and accurate means of assessing these models need to be established.

Cassano, J. J., et al. (2001), Evaluation of Polar MM5 simulation of Greenland's atmospheric circulation, Journal of Geophysical Research, 106, 33,867-833,889.

Elsner, J. B., and C. P. Schmertmann (1994), Assessing forecast skill through cross validation, Weather and Forecasting, 9, 619-624.

Michaelson, J. (1987), Cross-validation in statistical climate forecast models, Journal of Climate and Applied Meteorology, 26, 1589-1600.

Ohmura, A. (2001), Physical basis for the temperature-based melt-index method, Journal of Applied Meteorology, 40, 753-761.

Day 2 Question 2: Hurricane intensity debate

Question 2:

The very active 2004 and 2005 Atlantic hurricane seasons have helped fuel the debate about global warming and hurricanes. Summarize the debate. Point to problems with each side's opinion. Pick a side and defend it. Explain how we might address the question with the available data. Be creative.

Introduction

Presently a debate rages on as to whether the recent upswing in hurricane intensity is within the natural variability of the climate system or being driven by the secular warming of the Earth’s atmosphere and oceans. One group of scientists feel that the observed increase in tropical cyclone intensity around the globe is being driven by warmer sea-surface temperatures (SSTs) which are a direct result of the atmospheric loading of carbon dioxide gas by humans. Another group of scientists are arguing that natural cycles of ocean circulation affect the amount and intensity of tropical cyclones in the Atlantic basin (hurricanes). A third possibility is that both of these explanations are valid, and that the recent strengthening of storms is part of a natural cycle that is being enhanced by global warming. I will present all three of these possibilities, strengths and weaknesses in their arguments, and then present my own opinion in the concluding remarks.

Background

Tropical cyclones, better known in the United States as hurricanes, which is the name given to tropical cyclones that form in the North Atlantic, Carribean Sea, Gulf of Mexico, or Eastern Pacific, may be thought of as convective engines of latent heat, or Carnot cycles, as a first approximation (Emanuel, 1987). In this simplified view, global warming can theoretically alter the maximum potential intensity (MPI) of tropical cyclones through escalating the surface energy balance and the upper-level cold air return, both of which intensify the cyclone (Henderson-Sellers et al., 1998).

Over the past 30 years, tropical SSTs have been steadily rising. This is generally attributed to anthropogenic global warming (Emanuel, 2005). In this same period, the number and percent of tropical cyclones intensifying to categories 4 and 5 increased (Webster et al., 2005). The relative increases in the number of category 4 and 5 storms for the 15-year periods 1975-1989 and 1990-2004 are shown in Figure 1 for different ocean basins, while the percent of cyclones reaching these intensities are shown in Figure 2 for the same basins and periods. Interestingly, while the number of strong tropical cyclones has increased in all basins, the absolute number of storms (not shown) has decreased in all basins except the North Atlantic since 1995 (Webster et al., 2005). In the last 10 years, there has been an increase in the frequency and intensity of hurricanes in the North Atlantic, while over the last 30 years, there has been a 50% increase in the intensity and duration of hurricanes with no increase in the frequency (Emanuel, 2005).

Figure 1: Changes in the numbers of tropical cyclones reaching category 4 or 5 for the 15-year periods of 1975-1989 and 1990-2004 for different ocean basins (adapted from Webster et al., 2005).

Figure 2: Changes in the percents of tropical cyclones reaching category 4 or 5 for the 15-year periods of 1975-1989 and 1990-2004 for different ocean basins (adapted from Webster et al., 2005).

Potential causes

1. Global Warming

Global climate models (GCMs) reveal a significant increase in the potential intensity of hurricanes with anthropogenic warming (Emanuel, 1987; Henderson-Sellers et al., 1998). This leads to the natural prediction that hurricane intensities should increase in the future. Given the current estimated warming in the tropics of 0.5°C, the resulting changes in hurricane intensities should not yet be detectable (Emanuel, 2005). There is no consistent link between hurricane frequencies and global warming, nor has any long-term trend in hurricane frequency been established (Emanuel, 2005).

Still, an obvious and strong relationship (r2 = 0.69) exists between SSTs and the total power dissipated (power dissipation index, PDI) by hurricanes each year (Emanuel, 2005), as shown in Figure 3. The strong rise in tropical SSTs over the last 30 years, which has been attributed to anthropogenic warming, is well-correlated with the near doubling of hurricane PDI in the same period. That evidence, coupled with the well-known relationships between SSTs and hurricane intensities (e.g. Emanuel, 1987), suggests that anthropogenic warming is at least partially responsible for this observed increase in the destructiveness of hurricanes.

Figure 3: Annual PDI for the North Pacific and North Atlantic basins compared to annually averaged tropical SSTs (source: Emanual, 2005).

Further evidence for the relationship between hurricane intensity and global warming comes from modeling studies. Knutson and Tuleya (2004) demonstrate that in a CO2-enriched atmosphere, the number of intense tropical cyclones predicted by circulation models increases, while the number of storms peaking out at weaker intensities decreases. They get similar results irrespective of the model and the sub-grid-scale convective parameterization scheme they use. Their results are summarized in Figure 4, which shows histograms of hurricane intensity results from all 1296 experiments. The open circles represent the control runs, while the filled circles represent the high CO2 cases. Note that there is no increase in hurricane frequency with high CO2, but the frequency of more intense storms does increase. This matches what is presently observed (Emanuel, 2005).

Figure 4: Hurricane intensity (mb) frequencies from 1296 modeling experiments. Open circles represent the control run, while filled circles represent the high CO₂ experiments (source Knutson and Tuleya, 2004).

Knutson and Tuleya attribute the consistent increase in hurricane intensity to a correlated increase in convective available potential energy (CAPE), despite the expected and observed cooling of the upper troposphere. These results coincide with observations made by Gettelman et al. (2002) which show increased CAPE measured from radiosonde data in the tropics at most of the 15 stations monitored, especially since 1975. Similarly, their modeled increase in precipitation intensity from the storms coincides with GCM simulations (e.g. IPCC, 2001) and the understanding that warmer air has a higher saturation vapor pressure. The relationship between temperature and saturation vapor pressure is dictated by the Clausius-Clapeyron equation (Oke, 1987; Stull, 1988), but this relationship is non-linear and may not include positive feedbacks which occur with the release of latent heat in convective storms (Trenberth et al., 2003).

The robustness of Knutson and Tuleya’s results lends great support for the notion that global warming can cause tropical cyclones to intensify, but it cannot show a unequivocal link to the current climb in hurricane strength. Moreover, Henderson-Sellers et al. (1998) indicate that modeling results which simulate tropical cyclones must be viewed carefully, as GCMs have too coarse a resolution to realistically capture the factors which dictate cyclone intensity. While making the causal jump between warmer SSTs and more intense hurricanes seems logical, and the balance of observations support this relationship, Trenberth (2005) concedes that a definitive linkage is not presently possible. Likewise, Peilke et al. (2005) argue that due to the complexity of the climate system, drawing a certain relationship between global warming and tropical cyclone intensity is a bit precocious before a clear relationship is demonstrated and widely accepted.

2. Natural Cycles

While the connections between global warming and increased hurricane intensity are clear to some, others argue that this upswing in storminess is likely part of a 60-70 year cycle which alters the strength with which ocean currents cycle warm water and heat around the globe. These researchers believe that we are now in the fast-flowing, or warm mode of this cycle, which has been referred to by some as the Atlantic Multidecadal Oscillation (AMO; Kerr, 2000). In this fast phase, warm Atlantic SSTs and low vertical wind-shear favor tropical cyclogenesis.

The AMO was first recognized by Delworth and Mann (2000), who showed that multidecadal oscillations in temperature proxy data extend back through the entire 330 year record. Using two independent coupled ocean-atmosphere models, they were able to produce comparable fluctuations which involved the oscillation of the speed of the thermohaline circulation (THC) in the North Atlantic. The THC is the ocean conveyor belt in which a continuous flow of warm surface waters is drawn from the tropical waters northward into the North Atlantic. As it cools near Greenland, and becomes saltier due to salt loading from sea ice, the water becomes more dense and sinks. The cool waters then flow deep in the ocean back to the original sources and rise. It is the sinking in the North Atlantic, we believe, that controls the speed of the conveyor.

We now understand that the AMO is a natural cycle in the speed of the THC (Knight et al., 2005) driven by years of cool winds off Canada cooling the waters in the North Atlantic, causing them to sink faster, driving the ocean conveyor faster. This faster conveyor then drives warmer waters into the North Atlantic, leaving cooler waters in the South Atlantic. These conditions drive easterly waves off of Africa, provide warmer SSTs in the North Atlantic, and lessen vertical wind-shear between the upper- and lower-troposphere, all of which are conditions favorable for tropical cyclone formation (Goldberg et al., 2001).

Goldberg et al. (2001) argue that the observed multidecadal variability in hurricane activity is greater than what is predicted from global warming. They further indicate that while tropical SSTs have increased gradually over the last 100 years, Atlantic hurricane activity has exhibited multidecadal cycles. Henderson-Sellers et al. (1998) point out that while there have been several studies of the potential effects of global warming on the frequency of Atlantic hurricanes, the results do not agree. Furthermore, it can be argued that the present active hurricane epoch (1995-present) only appears more active than the previous active epoch (1926-1970) because of more advanced monitoring systems including satellite technology.

3. Could it be both?

One final explanation for the upswing in tropical cyclones is a combined effect of the natural AMO cycle and global warming. This point is illustrated in Figure 5, a simple cartoon-like graphic depicting the smoothed global temperature record of the 20th century (green), which is known to correlate with the AMO, a natural 70-year cycle to represent the AMO (purple), and a 0.6°C/century warming trend (orange; IPCC, 2001). This graphic is strictly for illustrative purposes, but demonstrates that natural cycles and global warming may work in concert to produce the recent climatic fluctuations we have observed. In fact, if the purple and orange lines are added, they faithfully create the green line.

Figure 5: Cartoon graph of 20^th century temperatures (heavily smoothed; green), and the break-down of that curve into a cycle with an approximate period of 70 years (representing the AMO; purple) and a long-term trend (representing global warming; orange). This illustrates the possibility that both the AMO and global warming contributed to the observed temperatures of the 20^th century, and presumably to hurricane intensity cycles.

While Figure 5 is strictly meant to illustrate a point, it implores an additional question that has been posed recently by several researchers. The climate system exhibits a number of oscillations on a variety of time scales, ranging from glacial-interglacial cycles of 100ka to the 2 year QBO, and all points in between. As we shift between warm and cold phases of the PDO, for example, some anticipate that the next warm phase will be warmer than the previous one, as each is now riding on the back of a significant global warming trend. If this is true of all cycles, then will the next warm phase of the AMO potentially be even more devastating than any of the previous ones?

What’s really going on?

While there is strong evidence for both global warming and the AMO contributing to hurricane intensities, an additional piece of the puzzle is currently coming into the spotlight. A rapid freshening of the North Atlantic waters over the past 40 years was recently observed (e.g. Dickson et al., 2002). This freshening, predicted by models of global warming (IPCC, 2001), is potentially caused by enhanced precipitation, river runoff, and increased melt of glaciers and sea ice, especially the Greenland ice sheet. A fear of this freshening is that it may cause a slowing of the THC, and therefore of meridional heat transport, especially to Europe (Quadfasel, 2005; Sutton and Hodson, 2005). This freshening may be responsible for a recent slow down of the Atlantic conveyor as observed at 25°N, a common proxy for THC circulation (Bryden et al., 2005).

This latest finding adds support to the anthropogenic forcing of global warming, as climate models have predicted human-induced warming to cause a slow-down of the THC. Additionally, those who argue that natural cycles are responsible for the upswing in tropical cyclone intensity are left to explain how the THC is slowing down when they claim that we are in the fast mode of the AMO. One final note is that the AMO does not explain, through adequate physical principles, how tropical cyclone intensities are increasing around the globe (see Figures 1 and 2, for example).

With all of this evidence, I would have to stand firmly on the side of those who blame global warming for the increase in storminess. As for the notions that natural cycles have persistently forced hurricane intensity and that global warming is not predicted to have such a strong influence on tropical storm intensity with such a modest warming, my response is simply to restate a famous quote:

Due to the complexity of the climate system, and the abruptness of climate changes that punctuate paleoclimate records, we cannot expect climate to behave exactly as models tell us. I believe that we have reached a tipping point in climate and the plethora of changes we are currently witnessing is merely the beginning.

Bryden, H. L., et al. (2005), Slowing of the Atlantic meridional overturning circulation at 25°N, Nature, 438, 655-657. (link)

Delworth, T. L., and M. E. Mann (2000), Observed and simulated multidecadal variability in the Northern Hemisphere, Climate Dynamics, 16, 661-676. (pdf)

Dickson, B., et al. (2002), Rapid freshening of the deep North Atlantic Ocean over the past four decades, Nature, 416, 832-837.

Emanuel, K. (1987), The dependence of hurricane intensity on climate, Nature, 326, 483-485.

Emanuel, K. (2005), Increasing destructiveness of tropical cyclones over the past 30 years, Nature, 436, 686-688.

Gettelman, A., et al. (2002), Multidecadal trends in tropical convective available potential energy, Journal of Geophysical Research, 107, 4606.

Goldberg, S. B., et al. (2001), The recent increase in Atlantic hurricane activity: causes and implications, Science, 293, 474-479.

Henderson-Sellers, A., et al. (1998), Tropical cyclones and global climate change: a post-IPCC assessment, Bulletin of the American Meteorological Society, 79, 19-38.

IPCC (2001), Climate Change 2001: The Scientific Basis, Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), 944 pp., Cambridge University Press, Cambridge, UK.

Kerr, R. A. (2000), A North Atlantic climate pacemaker for the centuries, Science, 288, 1984-1985.

Knight, J. R., et al. (2005), A signature of persistent natural thermohaline circulation cycles in observed climate, Geophysical Research Letters, 32, L20708. (pdf)

Knutson, T. R., and R. E. Tuleya (2004), Impact of CO₂-induced warming on simulated hurricane intensity and precipitation: sensitivity to the choice of climate model and convective parameterization, Journal of Climate, 17, 3477-3495.

Oke, T. R. (1987), Boundary Layer Climates, 2nd edition ed., 435 pp., Routledge, London.

Pielke Jr., R. A., et al. (2005), Hurricanes and global warming, Bulletin of the American Meteorological Society, 86, 1571-1575.

Quadfasel, D. (2005), The Atlantic heat conveyor slows, Nature, 438, 565-566.

Stull, R. B. (1988), An introduction to boundary layer meteorology, 666 pp., Kluwer Publishers, Dordretcht.

Sutton, R. T., and D. L. R. Hodson (2005), Atlantic Ocean forcing of North American and European summer climate, Science, 309, 115-118.

Trenberth, K. E. (2005), Uncertainty in hurricanes and global warming, Science, 308, 1759-1754.

Trenberth, K. E., et al. (2003), The Changing Character of Precipitation, Bulletin of the American Meteorological Society, 84, 1205-1217.

Webster, P. J., et al. (2005), Changes in tropical cyclone number, duration, and intensity in a warming environment, Science, 309, 1844-1846.

Day 2 Question 3: An attempt to explain the Coriolis force

Question 3:

Recently a lecturer in physical geography attempted to explain the Coriolis force. He appealed to the experience of being on a merry-go-round and attempting to walk from its center to the edge. "This is difficult to do because of the Coriolis force." Discuss this explanation.

Apparently this instructor needs a basic course in physics. The reason that it is difficult to walk from the center to the edge of a rotating carousel is centrifugal force. The Coriolis “effect”, on the other hand, is not an actual force, but an effect of your frame of reference. A better analogy to explain the Coriolis effect would be to imagine sitting on one end of the carousel, not knowing that it was rotating, and trying to throw a ball to your friend sitting directly across the center from you. This assumes, of course, that the center of the carousel is as empty as the aforementioned instructor’s head, allowing you a clear shot at your friend. The ball leaves your hand, traveling directly toward where your friend was when you released the ball. As the ball is traveling in the air, however, your friend is rotating away from where the ball eventually lands. To an observer situated in the rotating frame of reference, it appears that ball has curved away from its original path. To an observer watching from above, in a fixed frame of reference, the ball traveled in a straight line. If the merry-go-round is rotating counter-clockwise as seen from above, analogous to the Earth’s rotation as seen from above the North pole, the ball would appear to always veer to the right in the perspective of the thrower. This apparent curving of the trajectory seems to be caused by some unseen force. This apparent force is named for the French scientist Gaspard Coriolis who first described it. It is referred to by some as the Coriolis force, but technically it is more proper to call it the Coriolis effect, since it is not an actual force.

On the rotating Earth, the Coriolis effect depends on the distance traveled and the distance from the Equator. The Coriolis effect is weakest at the equator, strongest at the poles, and it is only significant over large distances. Accordingly, it is humorous that a gift shop which straddles the equator in Kenya advertises that the toilets on the north end of the store flush the opposite direction as those on the south end. This phenomenon is caused solely by the design of the toilets, which dictates the direction that every toilet flushes, and not the Coriolis effect, which (1) does not matter over the small diameter of a toilet bowl, and (2) is ineffective so close to the equator!

Day 2 Question 4: A critique of the comment that weather results from deterministic and stochastic factors

Question 4:

Criticize or defend the often heard comment that weather events (such as rainstorm, etc) result from a combination of deterministic and stochastic factors.

With the globalization of media, every major weather event becomes a widely discussed phenomena. The plethora of such weather events, particularly the onslaught of hurricanes we’ve witnessed in the past two years, begs the discussion of the cause of such events. People naturally wonder if we’ve altered climate by loading the atmosphere with greenhouse gasses to the point that dramatic weather events will become more common and more severe. Such claims by the public and the media then beg scientists to sort out the trend of global warming from the stochasticity, or randomness, of our climate.

One common question raised today is whether single events, like hurricane Katrina or hurricane Rita, were caused by global warming, or made more intense because of global warming. A second question is if events like those are going to become more common in a warming climate. The first of these questions gets to the heart of the topic at hand. The second question, however, is much simpler and should be handled first.

Global Climate Models (GCMs) used to predict future climates almost all agree that in an atmosphere artificially enriched in CO2, stronger hurricanes will become more common, while weaker storms will be less common (Knutson and Tuleya, 2004). Knutson and Tuleya’s study is based on the physical principles of tropical cyclones, and their results yield a deepening of the central pressure and more intense rainfall in a warmer climate.

These results do imply that more intense storms will become more likely, however, it is not appropriate to draw the conclusion that a single event, like hurricane Katrina, was made worse by global warming. In fact, it is impossible to definitively determine whether a single event was worsened by global warming because it is impossible to separate the effects of anthropogenic warming from natural cycles and the stochastic nature of weather on the scales which affect a single storm.

That said, on a larger scale, it is difficult, but not impossible to determine some of the influence that global warming is having on climate. The climate system, though, is extremely complex (Rind, 1999), and therefore behaves non-linearly and chaotically (Rial et al., 2004). This is not to say that weather events are totally random, rather that they may appear random because the factors affecting a single event are so complex and widespread.

Rahmstorf et al. (2005) suggest that every weather event is influenced by both stochastic chance and deterministic values. While this explanation is elegant, I disagree. I believe that weather is not random. I believe that everything that happens happens for a reason. Well, not just for one reason, but for a vast and complex set of reasons. This is, essentially, the premise of chaos.

If we take the stochastic nature of weather to be caused by the complexity of the climate system and the chaotic way in which it behaves, than the notion that weather events have a stochastic element is completely true. If we define stochasticity, however, as being simply random chance, than I no longer feel it is part of the equation. On the other hand, what exactly differentiates randomness and chaos?

Rahmstorf, S., et al. (2005), Hurricanes and global warming - is there a connection?, RealClimate, 181.

Rial, J. A., et al. (2004), Nonlinearities, Feedbacks and Critical Thresholds within the Earth's Climate System, Climatic Change, 65, 11-38.

Rind, D. (1999), Complexity and Climate, Science, 284, 105-107.

Day 3 Question 1: The significance of image resolution in remote sensing

Question 1:

A working knowledge of image resolution is critical for understanding the practical and conceptual aspects of remote sensing. Write an essay to discuss the significance of image resolution as a concept that extends across some major aspects of remote sensing. Whenever possible, you are encouraged to use your own examples to make the case.

One of the most critical aspects of any remote sensing application is the spatial resolution of the sensor platform. Satellite sensors vary in pixel resolution from sub-meter to many kilometers. Each remote sensing application must carefully match an adequate spatial resolution to the scale of the phenomenon being studied. Typically, the satellite sensor resolution should be less than half the size of the smallest feature that must be resolved (Jensen, 1996).

The issues of scale in remote sensing share similarities and differences with those of ground-based studies. For example, in most ecological studies, the challenge is to apply conclusions drawn from detailed studies of small patches to broader scales (Turner et al., 2001). In fact, the size of the study area used in ground-based studies often affects the conclusions drawn from the study. This scaling issue also applies to remote sensing applications (Lillesand et al., 2004), however, in remote sensing the challenge is often to apply what is learned over a broad area (the satellite scene) to a smaller, more detailed patch, which can range from a large group of pixels to sub-pixel areas.

Remote measurements in mountainous terrain require similar spatial resolution to that of the topographic features. In most montane areas, this is on the order of tens of meters (Dozier et al., 1987). Satellite data of this resolution are available from Landsat Thematic Mapper, Enhanced Thematic Mapper, SPOT, and several newer commercial satellites.

In some applications, lower spatial resolution can be offset by higher spectral resolution, or the number and size of wavelength bands that the sensor measures. Hyperspectral data, data which have higher spectral resolution than required by the given application, have been employed to detect features much smaller than the spatial resolution of the sensor. For example, Nolin and Dozier (2000) have employed hyperspectral data to detect the grain size of snow. Their technique was based on the unique spectral reflectances of snow of different grain sizes. Painter et al. (2003) also determined snow-covered area from hyperspectral imagery. Similar applications have been used by geologists to locate rare minerals.

Other scale issues haunt the application of remote sensing. Although remotely-sensed data offer broad coverage, for any given study a balance must be struck between spatial coverage and spatial resolution. The surface of the Earth is too large to sense with high resolution imagery. The resulting dataset would be too large to manage. Furthermore, ground-truth data are usually collected from relatively small areas, which raises issues when relating those data to the broader spatial scales of the remotely-sensed imagery. Turner et al. (2001) warn that any conclusions drawn from studies at one spatial scale may not apply to other scales. Finally, there is great trouble when combining datasets with different spatial resolutions. This issue is covered in more detail in question 2.

Jensen, J. R. (1996), Introductory Digital Image Processing: A Remote Sensing Perspective, 2nd edition ed., 316 pp., Prentice Hall, New Jersey.

Lillesand, T. M., et al. (2004), Remote Sensing and Image Interpretation, 5th ed., 763 pp., John Wiley & Sons, Inc., New York.

Nolin, A. W., and J. Dozier (2000), A hyperspectral method for remotely sensing the grain size of snow, Remote Sensing of Environment, 74, 207-216.

Painter, T. H., et al. (2003), Retrieval of subpixel snow-covered area and grain size from imageing spectrometer data, Remote Sensing of Environement, 85, 64-77.

Turner, M. G., et al. (2001), The critical concept of scale, in Landscape Ecology and Theory in Practice: Patterns and Processes, Springer, New York.

Day 3 Question 2: Challenges of image fusion for data extraction

Question 2:

Image fusion has recently emerged as a major research area in remote sensing and image processing. Write an essay to examine the significance of image fusion for information extraction, to discuss the major problems and challenges of image fusion, and to suggest some possible solutions to the problems you identified.

Often times, a single remotely-sensed data source is inadequate for a given study, giving rise to the need for analysis of a mixed set of spatial data (Richards and Jai, 1999). Drawing inferences about a single pixel from multiple data sources is referred to as data fusion. Data fusion is most easily addressed when the data sources are of the same type, and can therefore be handled by the same algorithms and systems. The issue becomes increasingly complex, however, as various data types are added to the analysis.

Some solutions to data fusion issues with various data types include numerically-based quantitative methods and methods based on evidential theory and expert systems. Expert systems even allow for non-numerical data sources. These results of any of these methods, however, is greatly influenced by the accuracy of the registration of the data (Richards and Jai, 1999).

One numerical method for handling classification of mixed data sets is to stack the data and extend the pixel vectors through the spectral and non-spectral data. This is known as the stacked vector approach. One benefit of this methods is that it allows for the use of traditional classification techniques, except for those which rely on training class statistics, such as the maximum likelihood algorithm (discussed further in question 3). Alternatives to the maximum likelihood classifier which work for mixed data sets include the parallelepiped classification (Richards and Jai, 1999).

Numerical solutions like the stacked vector approach lack the ability to handle non-numerical data, yet many of the data in a typical GIS are implicitly non-numerical. Expert systems and evidentially theory methods allow for non-numerical data types, and allow these data to effect the classification or numerical labeling of pixels without being treated in explicitly numerical ways. The links between the numerical and non-numerical data are still handled quantitatively.

Richards and Jai (1999) provide an example of data fusion using different data types. The example involves a multispectral (MS) satellite image with synthetic aperture radar (SAR) data. The MS image alone is not able to distinguish built-up urban areas from areas cleared for construction. The SAR data provides structural information, but no land-cover types. A knowledge-based expert system is then constructed which uses the MS imagery to decompose the pixels into likely classes and then uses the SAR structural information to either further segment those classes, like separating woody vegetation from grassland, or deciding whether a pixel belongs in one class or another, like determining between construction areas and urban areas. These decisions are made on a pixel-by-pixel basis.

Works Cited: 
Richards, J. A., and X. Jia (19996), Remote Sensing Digital Image Analysis: An Introduction, 3rd ed., 363 pp., Springer, Berlin.

Day 3 Question 3: Problems with image classification and optimism for the future

Question 3:

Image classification is probably the most important task for remote sensing. Yet it can be quite challenging for many applications. Over the past three decades, many classification strategies have been developed. Write an essay to discuss the status and problems of image classification. Are you optimistic about the future of automated classifiers applied to the physical and human environments, and why?

Introduction

One of the most useful and often employed tools for extracting land cover information from remotely sensed data is image classification (Jensen, 1996). Classification of multispectral imagery is generally performed by one of several algorithms including supervised, unsupervised, fuzzy classification, band math, or hybrid techniques. Hybrid techniques include any combination of the other techniques, each of which will be described in more detail below.

Supervised classification methods

Supervised classification techniques require the operator to identify pre-determined land-cover classes by selecting image pixels that are representative of each of those classes (Mausel et al., 1990). These representative pixel sets are referred to as training sets. Mean spectral reflectance and standard deviations of each land-cover classes is derived from the training sets. From these training set statistics, various supervised classification algorithms are able to select the most appropriate class for each pixel in the image (Richards and Jia, 1986).

One of the simplest supervised classification methods is the parallelepiped technique. A parallelepiped is an n-dimensional shape with planar surfaces. The parallelepiped algorithm defines each training class as an n-dimensional parallelepiped, where n is the number of spectral bands in the image. The dimensions of the parallelepiped are defined using a threshold specified by the operator and the standard deviations from the mean of each class. Pixels that fall within a single class parallelepiped are assigned to that class. Pixels that do not fall within a class parallelepiped or in regions where two parallelepipeds overlap are not classified. Although simple and efficient, this method has several shortcomings. When too small a threshold is specified, the class parallelepipeds will be small and many pixels will go unclassified. If too great a threshold is chosen, class parallelepipeds will overlap and pixels that fall in the overlapping regions are also not assigned to a class. This is especially problematic for highly-correlated data sets (Richards and Jia, 1986).

The spectral angle mapper (SAM) algorithm also treats each pixel as a vector in n-dimensional space. The spectral signature of each pixel is described by the angle its corresponding vector makes with the axes that define the n-dimensional space of the image. A mean vector for each class is calculated and all pixel vectors are classified based on the angle they make with each class vector (Richards and Jia, 1986). A maximum angle may be defined whereby pixels that are not within that angle of any class are not classified. This algorithm is less sensitive to illumination and albedo effects when used with calibrated reflectances.

Most supervised classification algorithms assume an equal probability of each pixel belonging to each class (Hord, 1982). One of the most popular supervised classification algorithms, the maximum likelihood classifier, is based on Bayesian probability theory (Eastman, 2000), and therefore does not assume equal probabilities for each class. Instead, this algorithm utilizes the mean measurement vector and the covariance matrix from each signature to determine a probability that each pixel in an image belongs to each class. For this reason, the maximum likelihood procedure is considered the most powerful hard classification system in many remote sensing applications, yet it is recommended only when the training sites are well-defined with large sample sizes and are relatively homogeneous (Eastman, 2000).

One special case of the maximum likelihood method is the minimum distance to means (or minimum distance) classifier. This algorithm assumes identical and symmetric class distributions, so it is able to classify pixels using the mean measurement vector of each class without calculating a covariance matrix for each (Richards and Jia, 1986).

Unsupervised classification methods

The greatest weaknesses of supervised classification methods are their sensitivity to training site definitions and their assumption of a normally distributed probability distribution function for each class (Richards and Jia, 1986). Unsupervised classification techniques, on the other hand, use clustering techniques to generate classifications by grouping pixels with similar spectral characteristics. The operator then combines and labels the spectral clusters into meaningful land cover classes (Jensen, 1996). This is not always straightforward as clusters sometimes represent mixed classes of land cover as discussed above.

The most common clustering method is the iterative isodata algorithm in which initial candidate clusters are selected and their means are allowed to migrate in the spectral domain, optimizing the classification with each iteration (Ball and Hall, 1965). In each iteration of the algorithm, every pixel is compared to each new cluster mean and assigned to the nearest cluster. The means are then recalculated and the process repeats until the pixel-to-cluster mean distances are minimized or a pre-defined number of iterations is reached. Generally, the time required for class merging and labeling after an unsupervised clustering algorithm is less than the time required to define training sets for supervised classifications (Albert, 2002).

Fuzzy classification methods

Unsupervised and supervised classification methods are considered hard classifiers in that all pixels are forced into discrete groups (Jensen, 1996). In reality, the instantaneous field of view or pixel size of a sensor records the radiation emitted or reflected from a mixture of land covers, which do not typically exhibit sharp geometric boundaries like image pixels. Accordingly, a single pixel is often a mixture of multiple surface types as land covers grade into one another (Lam, 1993; Wang, 1990a,b). Fuzzy classification methods, however, assign a set of probabilities to each pixel based on the likelihood that it belongs to each land-cover class. This information can then be used to determine more precise land-cover classes, including mixed pixel classes (Jensen, 1996). There are several types of fuzzy classification techniques, including linear spectral unmixing, mixed tuned matched filtering (MTMF), and spectral feature fitting (SFF).

Linear spectral unmixing is based on the assumption that the spectral reflectance of a pixel is a linear combination of the unique reflectance spectrum of each material present in the pixel in the proportion in which they cover the pixel area (Menke, 1984). Although this technique is often reserved for hyperspectral imagery, it has been used with multispectral imagery with limited success (Richards and Jia, 1986). Other classification schemes are usually better suited for multispectral imagery (Albert, 2002).

Generally to perform an unmixing, the reflectance spectra of each land-cover type is needed. These can be obtained in situ using a field spectrometer, or training sets can be established using the endmembers, or samples of pure cover type (Richards and Jia, 1986) found from the results of running a pixel purity index (PPI) algorithm. The PPI is performed on a minimum noise fraction (MNF) transformation of the image data (Green et al., 1988). The MNF transform is a two step transformation. First, a principal component analysis (PCA) is performed on the data to decorrelate and rescale the noise. Any band-to-band correlation in the noise is removed and the resulting noise has unit variance. Next, a second PCA transformation is done on the noise-whitened data. The results are the MNF transformed image. The PPI procedure then continually re-projects the MNF transform result onto random unit vectors for a user-specified number of iterations (typically 1×104 to 1×108 times). Pixels at the ends of these vectors are tagged with each rotation. The number of times each pixel is tagged is recorded. Purer pixels tend to be tagged more often (Boardman, 1993; Boardman et al., 1995) and, thus, the highest scoring pixels are taken as potential endmembers. The user than rotates the purest pixels in n-dimensional space to identify endmember clusters. This method is extremely time consuming, both in terms of operator time and processing time (Albert, 2002), but is extremely useful in hyperspectral applications where spectral libraries are not available for all land-cover types.

The MTMF technique is a partial unmixing method that does not require all image endmembers to be defined. The algorithm returns a percent cover image for each defined endmember as well as an infeasibility score for each to help reduce falsely classified pixels. Endmember mixtures can then be compared to their infeasibility scores and pixels that have a high mixture score and a low infeasibility score are confidently classified as that endmember (Albert, 2002).

One additional fuzzy classification technique is the SFF algorithm. This algorithm returns a scale image which is a measure of how well the spectral signature of a pixel matches each training set spectrum. The algorithm also returns a root-mean-square (RMS) error image for each training set. The RMS image can then be plotted against the individual endmember scale images and pixels with low RMS error and high scale scores for a given class can then be assigned to that class (Albert, 2002).

Band math methods

Since the reflectivity of some surfaces depend on solar insolation angles and viewing angles, the pixel vectors created using multiple bands are often better at classifying a pixel than any single spectral band. These techniques included band ratios, and normalized differences, PCA and MNF transforms, among others. A tremendous benefit of these methods is that they typically require very little operator time and are often some of the most accurate methods (e.g. Albert, 2002, Kääb et al., 2002b; Paul, 2002a, b).

Problems with classifications

One of the difficulties in performing an accurate image classification is identifying all of the environmental factors which affect the spectral reflectance of each land cover you are attempting to extract (Jensen, 1996), and how each alters the reflectance patterns. These may include soil moisture and soil type, water turbidity, plant species variations, scattered patterns of atmospheric haze or water vapor, variations in snow grain size or melt water content, plant stress level, canopy structure and biochemistry, variations in outcrop mineral content, and a variety of other factors. When one includes the variety of materials in an urban landscape, the list of factors increases in magnitude.

Selecting homogeneous and representative training sites for supervised and fuzzy classification algorithms is difficult in many cases, and the performance of the techniques relies on an accurate set of endmembers. The alternatives include spectral libraries, which are becoming more common, and endmembers created from PPI analyses. The latter can be quite costly in terms of human-input and computer processing time.

Another difficulty of image classification is that it is often difficult to gauge the accuracy of the classification. This leads to difficulty in selecting the most accurate classification algorithm, especially as some of the more complex algorithms do not necessarily produce the best results (Albert, 2002).

Optimism for the future

New digital image analysis methods for extracting land cover classifications from remotely sensed data are constantly being introduced. Bateson et al. (2000) have encorporated endmember variability in their training sets to improve spectral mixture analyses. Doucette et al. (2001) utilized a neural network to create a self-organizing road map algorithm which extracts elongated road features from multispectral imagery, while others have used edge-detection techniques. Hubert-Moy et al. (2000) present an original parametric classification method for determining land-cover type. Steele (2000) uses a combination of multiple classification techniques to derive a map of land cover. These new technologies, that continue to augment the science of classification, lend hope to the promise of more accurate automated classification systems for the future.

Ball, G. H., and D. J. Hall (1965), A Novel Method of Data Analysis and Pattern Classification, Stanford Research Institute, Menlo Park, CA.

Bateson, C. A., et al. (2000), Endmember bundles: a new approach to incorporating endmember variability into spectral mixture analysis, IEEE Transactions on Geoscience and Remote Sensing, 38, 1083-1094.

Boardman, J. W. (1993), Automated spectral unmixing of AVIRIS data using concept geometry concepts, paper presented at Proceedings from the Airborne Geosciences Workshop.

Boardman, J. W., et al. (1995), Mapping target signatures via partial unmixing of AVIRIS data, paper presented at Proceedings from the Airborne Geosciences Workshop.

Doucette, P., et al. (2001), Self-organized clustering for road extraction in classified imagery, ISPRS Journal of Photogrammetry and Remote Sensing, 22, 347-358.

Eastman, J. R. (2000), Idrisi32 Help Files, edited, Clark Labs, The Idrisi Project, Worcester, MA.

Green, A. A., et al. (1988), A transformation for ordering multispectral data in terms of image quality with implications for noise removal, IEEE Transactions on Geoscience and Remote Sensing, 26, 65-74.

Hord, R. M. (1982), Digital Image Processing of Remotely Sensed Data, , 256 pp., Academic Press, New York.

Hubert-Moy, L., et al. (2000), A comparison of parametric classification procedures of remotely sensed data applied on different landscape units, Remote Sensing of Environment, 75, 174-187.

Jensen, J. R. (1996), Introductory Digital Image Processing: A Remote Sensing Perspective, 2nd edition ed., 316 pp., Prentice Hall, New Jersey.

Kääb, A., et al. (2002), Glacier monitoring from ASTER imagery: accuracy and applications, paper presented at EARSeL Proceedings. LIS-SIG Workshop, Berne, March 11-13, 2002.

Lam, S. (1993), Fuzzy sets advance spatial decision analysis, GIS World, 6, 58-59.

Mausel, P. W., et al. (1990), Optimum band selection for supervised classification of multispectral data, Photogrammetric Engineering and Remote Sensing, 56, 55-60.

Menke, W. (1984), Geophysical Data Analysis: Discrete Inverse Theory, Harcourt Brace Jovanovich, San Diego, CA.

Paul, F. (2002), Changes in glacier area in Tyrol, Austria, between 1969 and 1992 derived from Landsat 5 Thematic Mapper and Austrian Glacier Inventory data, International Journal of Remote Sensing, 23, 787-799.

Paul, F. (2002), Combined technologies allow rapid analysis of glacier changes, EOS, Transactions, 83, 253, 260-261.

Richards, J. A., and X. Jia (19996), Remote Sensing Digital Image Analysis: An Introduction, 3rd ed., 363 pp., Springer, Berlin.

Steele, B. M. (2000), Combining multiple classifiers: an application using spatial and remotely sensed information for land cover type mapping, Remote Sensing of Environment, 74, 545-556.

Wang, F. (1990a), Fuzzy supervised classification of remote sensing images, IEEE Transactions on Geoscience and Remote Sensing, 28, 194-201.

Wang, F. (1990b), Improving remote sensing image analysis through fuzzy information representation, Photogrammetric Engineering and Remote Sensing, 56, 1163-1169.

Day 3 Question 4: Remote Sensing of the Cryosphere

Question 4:

Prepare an essay to discuss the generic procedures for a cryospheric remote sensing project (such as mapping the extent and depth of snow cover, land ice, or sea ice). What are the major pitfalls at each phase, how to deal with them from both theoretical and practical perspectives? Whenever possible, you are encouraged to use your own examples to make the case.

Introduction

Climate fluctuates on all time and spatial scales, and these fluctuations are reflected in the continually varying distribution of ice on Earth (Paterson, 1994). Careful observation and analysis of changes in the extent and volume of ice may contribute to the reconstruction of past climates and aid in the understanding of current and future climate changes. Changes in the mass balance and morphologies of glaciers and ice sheets provide insight into changing climatic conditions on various spatial scales and temporal scales. Accordingly, changes in the Earth’s cryosphere, or frozen portions, and their links to climate, need to be studied on these various scales.

Glacial ice presently covers about 3 % of the Earth’s surface and contains nearly 75 % of its fresh water (Martinson et al., 1998). The mass balance of these ice bodies has a direct influence on sea level. Glaciers are important on regional scales as reservoirs of water for hydroelectricity and irrigation, and can effect humans through sudden events like ice dam collapses, or through glacier advances which can threaten human structures. On a larger scale, glaciers, and especially ice sheets, can modulate sea level and effect climate through the albedo feedback and by effecting the thermohaline circulation, and, in the case of the ice sheets, by posing as a barrier to atmospheric circulation.

The Greenland ice sheet stretches from approximately 60 N to 80 N, covers an area of approximately 1.7×106 km2, has an estimated volume of 2.6×106 km3, and contains roughly 8% of the volume of ice on Earth (Thomas, 1993). The Greenland ice sheet plays a crucial role in the global climate system on scales larger than the ice sheet itself. This large ice sheet directly influences global climate by forming an orographic barrier to atmospheric circulation in the Northern Hemisphere and through the albedo feedback (Barry and Kiladis, 1982). In turn, the mass balance of the ice sheet is influenced by the local climatic environment, particularly the exchanges of mass and energy between the surface of the ice sheet and the atmosphere.

Greenland's contribution to sea level is ultimately controlled by its mass balance. If the ice sheet were to melt completely, it would cause sea levels to rise by 7.4 m (IPCC, 2001). Although there is no evidence that this will happen in the future, recent modeling results suggest that the Greenland ice sheet may have contributed much more to sea level rise during the warm Eemian than the great ice sheets of Antarctica (Cuffey and Marshall, 2000). Repeat remote altimetry of the ice sheet reveal that at present, the ice sheet is thinning rapidly along its margins and is contributing an estimated 0.13 mm y-1 to global sea level rise (Krabill et al., 2000).

With an estimated total volume of 180,000 km3, glaciers and small ice caps, outside Antarctica and Greenland, comprise less than one percent of the ice volume on Earth, yet they play a significant role in the global hydrologic cycle, particularly the transfer of freshwater to the oceans (Williams and Hall, 1993). These smaller glaciers are potentially the most sensitive to climate fluctuations and may contribute significantly to the rise of sea level in the next century (Oerlemans and Fortuin, 1992). This results from their potential to rapidly gain or lose mass, and therefore many of the smaller glaciers are expected to disappear completely by 2100 (IPCC, 2001; Thompson, 2000; Thompson et al., 2000). Careful and continuous monitoring of these ice fields may help reveal small climatic changes that may otherwise go undetected.

Remote sensing of the cryosphere

Satellite imagery is an excellent tool for monitoring isolated ice fields as it allows more extensive and frequent observations. Thus, the response of ice fields to smaller scale climatic perturbations may be inferred. These data have the potential to provide regional and synoptic scale observations (Gurney et al., 1993) that are otherwise unavailable and offer a less expensive alternative to continuous field monitoring. Satellite data are well dated and may eventually provide a decadal- to century-scale record of glacial extent and environmental change. Furthermore, satellite imagery allows for a uniform digital inventory of land ice (Williams and Ferrigno, 1997). Nevertheless, satellite sensors are still limited in their measurement resolution and some variables cannot be observed from these platforms. Therefore, ground-based observations and field research will continue to play an important role in science in the coming years. Rather than eliminating field research, satellite observations offer new ways to address questions and a wider range of questions that may be addressed (Gurney et al., 1993) and may aid in more efficient and productive field campaigns.

Ice-extent changes

The areal extent of an ice cap may be determined through land-cover classification of a satellite image (Albert, 2002). Classification schemes rely on the spectral reflectance signatures of ground features in both short and long wavelengths. Different materials exhibit different spectral signatures which are utilized in grouping features into land cover types. In theory, the unique spectral signatures of snow and ice make them readily distinguishable from other materials in most conditions. In practice, however, the identification of snow and ice is not always straightforward. The albedo of a snow or ice surface changes dramatically with age and dustiness. This factor is especially acute near the margins, where exposed ice becomes dusty and may develop a surficial cover that resembles the surrounding glacial moraines. Additionally, snow and ice are not Lambertian; instead, their reflectance is highly dependent on the slope and aspect of the surface, the zenith angle and azimuth of the sun, and the zenith angle and azimuth of the satellite sensor. Since the geometry of the sun and satellite are effectively fixed throughout the recording of the image, additional variation in the reflectance of ice measured by the satellite may be caused by the slope and aspect of the surface. Generally, steep ice walls and crevasses in the image are heavily shaded and have very low reflectances. Shaded ice has a depressed spectral curve that approaches the spectral curve of water. Another factor that confounds the identification of ice is that edge pixels are likely to be combinations of ice and the surrounding surface material, resulting in a mixed signal. While fuzzy classification schemes can usually help determine the proportion of class cover in edge pixels, the environmental factors generally pose larger errors than mixture problems, and fuzzy classifiers tend to offer little or no advantage over hard classifiers.

Glacier extent mapping from satellite data has become more common and is the focus of the Global Land-Ice Measurement from Space (GLIMS) project. Bayr et al. (1994) used a a ratio image of Landsat TM band 4 to band 5 to delineate glacier area, while Rott (1994) used a band 3 to band 5 ratio image. Paul (2000) found that the band 4 to band 5 ratio technique yielded the best results for glacier mapping, especially in regions with low insolation. Unsupervised classification techniques, specifically the isodata algorithm, were used to map the Southern Patagonia Icefield by Aniya et al. (1996). Supervised techniques, in combination with elevation data was used to map glaciers by Gratton et al. (1990) and by Sidjak and Wheate (1999). Fuzzy classification schemes have also been used in glacier mapping by Serandrei-Barbero et al. (1999). Landsat images have been used to map alpine glaciers with the primary problem being the distinction between moraine-covered ice and ground surface. Albert (2002) analyzed a wide set of classification algorithms for delineation of snow and ice features, with particular consideration for the GLIMS project.

Delineation of ice extent of a single glacier for multiple dates allows for studies of ice-extent changes. However, it is possible for a glacier to advance while it is thinning and losing mass (Meier, 1984). Thus, observations of areal extent may not relate directly to changes in mass balance, and observations of the volumetric change of an ice cap may be a better indicator of mass balance changes (Popovin, 1996). This may limit the use of remote observations to monitor changes in the mass balance of ice fields and determinations of glacier sensitivities to climate change.

Thickness change and motion

Various methods have been developed to generate elevation information over ice fields from satellite sensors. These include synthetic aperture radar interferometry, shape-from-shading algorithms, and radar and laser altimetry. These and similar methods may be employed to reproduce the surface topography of the ice cap at a given time. Mass balance changes may be inferred directly from the change in volume between two observations (Krabill et al., 1999).

Synthetic aperture radar interferometry (InSAR) permits measurement of the relative position of points on the Earth’s surface as well as their displacement over time (Bürgmann et al., 2000; Meade and Sandwell, 1996), and has recently been applied to glaciological studies (e.g. Joughin et al., 2003). Topographic mapping of features may be obtained with sub-meter scale accuracy from radar images taken from slightly different perspectives. Any movement or disturbance of the surface between successive images may be mapped with sub-centimeter accuracy (Bürgmann et al., 2000). Each time the satellite passes over a ground feature it emits a radar pulse and measures the travel time required for that pulse to reach the surface and be reflected back. In this way, the distance between the surface feature and the satellite are measured precisely. On the next pass the distance to the same feature will change slightly due to the new viewing angle and any displacement that has occurred. When the effects of perspective are removed, the phase difference between the two images results in an interferogram consisting of a concentric pattern of interference fringes each representing a displacement of 28 mm (Kerr, 1996).

InSAR provides a systematic way to precisely map the grounding line of tidewater glaciers (Rignot et al., 1997). These methods have been employed to determine ablation rates on individual glaciers in the Arctic and Antarctic (e.g. Reeh et al., 2002). Excess water vapor in the tropics, however, hinders the utility of InSAR for monitoring glaciers there. Atmospheric water vapor may delay radar echos, and this delay may vary between successive images introducing large errors (Meade and Sandwell, 1996). For this reason, InSAR applications have been most suitable in arid or polar regions. Repeat measurements at multiple wavelengths may allow the humidity phase shift to be isolated, although this idea has not yet been tested. If removal of this signal is demonstrated, InSAR may prove to be an ideal tool for monitoring tropical ice caps as it is also capable of peering through thick cloud cover (Clery, 1997).

Similar to InSAR, radar altimetry may determine ice elevations to better than 10 cm. Yet the coarse spatial resolution and retracing problems limit the use of such systems in alpine areas. For these areas, laser altimetry may be better suited (Dozier et al., 1987). Laser altimetry utilizes short pulses of infrared radiation to measure ice sheet and land surface topography, and visible-green light to measure clouds and aerosols. The distance from the feature being measured to the satellite is inferred from the time taken for the laser pulses to travel to these targets and return (Isbell and Jones, 1998). Airborne laser altimetry was used to survey the entire Greenland ice sheet in 1993 and 1994, obtaining elevations to better than 10 cm accuracy (Krabill et al., 1999). Twenty-two flight lines were re-surveyed in 1998 and 1999 revealing widespread thinning of the ice sheet at elevations below 2000 m (Dahl-Jensen, 2000; Krabill et al., 2000). Utilizing existing satellite imagery, shape-from-shading algorithms attempt to recreate topography from variations in shading of a uniform surface caused by changes in relief. However, these studies are limited to broad, homogeneous, light-colored features such as the polar ice sheets and sandy deserts.

Melt detection

Another interesting application of remote sensing of the cryosphere involves the strong microwave signal differences between dry snow facies and wet snow facies. As soon as liquid water is present in snow, the microwave signal changes dramatically. Exploiting these differences, algorithms have been developed (Abdalati and Steffen, 1995) and recently utilized to derive a melt history of the Greenland ice sheet. This history reveals that the ice sheet has recently been exhibiting much greater areas of melt, and areas that were once labeled as dry snow, where the snow never melts, are being renamed.

Methods and potential pitfalls in remote sensing of the cryosphere

Aside from the many problems already mentioned, studies of ice extent change on alpine glaciers have several problems, which although they are not unique to these types of studies, they may be exacerbated. Here I will use the case of ice-extent change on the tropical Quelccaya ice cap as an example of the typical problems encountered.

The first step in any remote sensing study is data acquisition. In order to clearly delineate ice extent, the imagery should be free of any cloud cover and free of non-perennial snow cover. Clouds, while not impossible to delineate from ice, often obscure the edges of glaciers, precluding the proper mapping of the glacier extent. Non-perennial snow cover may also extend far beyond the margins of the ice cap, and is indistinguishable from perennial snow. Furthermore, illumination and view angles in all images should be similar, so obtaining imagery from the same time of year is ideal. With these constraints, it is often difficult to find imagery, especially from satellites with the high spatial resolution required for studies in alpine regions.

Once appropriate data are collected, the next step in a change detection study is the geometric correction and registration of the imagery. Proper geometric correction is essential if absolute areas are to be reported. Accurate geometric correction in montane areas requires the use of a high-resolution digital elevation model (DEM) to de-warp the image. Although DEMs of proper resolution for alpine research exist, they are seldom available to the public. If the satellite sensor is in a low enough orbit and the mountains are high enough, some degree of pixel distortion from elevation may also be present in the image.

For the image registration, very few ground control points are available in the mountains that are easily distinguishable on both maps and satellite imagery. For example, in the regions surrounding the Quelccaya ice cap, no roads or road intersections exist. The most easily distinguishable features in the imagery, both for the original registration, and eventually for the co-registration of the two images, are stream confluences. The problem with using streams (as opposed to roads) is that streams fluctuate and meander over time and with changes in the local hydrography. Once images are coregistered, there are still many issues with delineating ice extent, especially in areas of debris-covered moraines, shadows, steep slopes, and crevasses.

Albert, T. H. (2002), Evaluation of remote sensing techniques for ice-area classification applied to the tropical Quelccaya Ice Cap, Peru, Polar Geography, 26, 210-226.

Aniya, M., et al. (1996), The use of satellite and airborne imagery to inventory outlet glaciers of the Southern Patagonia Icefield, South America, Photogrammetric Engineering and Remote Sensing, 62, 1361-1369.

Barry, R. G., and G. N. Kiladis (1982), Climatic characteristics of Greenland, in Climatic and Physical Characteristics of the Greenland Ice Sheet, edited by U. Radok, pp. 7-33, CIRES, Boulder, CO, USA.

Bayr, K. J., et al. (1994), Observations on glaciers in the eastern Austrian Alps using satellite data, International Journal of Remote Sensing, 15, 1733-1742.

Bürgmann, R., et al. (2000), Synthetic aperture radar interferometry to measure Earth's surface topography and its deformation, Annual Review of Earth and Planetary Sciences, 28, 169-209.

Clery, D. (1997), RADAR IMAGING: Monitoring a killer volcano through clouds and ice, Science, 276, 1985.

Cuffey, K. M., and S. J. Marshall (2000), Substantial contribution to sea-level rise during the last interglacial from the Greenland ice sheet, Nature, 404, 591-594.

Dahl-Jensen, D. (2000), CLIMATE CHANGE: Enhanced: the Greenland Ice Sheet reacts, Science, 289, 404-405.

Dozier, J., et al. (1987), Prospects and Concerns for Remote Sensing of Snow and Ice, DRAFT, 31 pp, Ad Hoc Panel on Remote Sensing of Snow and Ice, Committee on Glaciology, Polar Research Board, National Academy of Sciences.

Gratton, D. J., et al. (1990), Combining DEM parameters with Landsat MSS and TM imagery in a GIS for mountain glacier characterization, IEEE transactions on Geoscience and Remote Sensing, GE-28, 766-769.

Gurney, R. J., et al. (1993), Atlas of Satellite Observations Related to Global Change, , 470 pp., Cambridge University Press, Cambridge.

IPCC (2001), Climate Change 2001: The Scientific Basis, Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), 944 pp., Cambridge University Press, Cambridge, UK.

Isbell, D., and T. Jones (1998), Ball Aerospace to Provide ICESAT Spacecraft, News Release, NASA, Washington, D.C.

Joughin, I., et al. (2003), Timing of Recent Accelerations of Pine Island Glacier, Antarctica, Geophysical Research Letters, 30.

Kerr, R. A. (1996), Watching the Earth move, Science, 272, 1870.

Krabill, W., et al. (1999), Rapid thinning of parts of the Southern Greenland Ice Sheet, Science, 283, 1522-1524.

Krabill, W., et al. (2000), Greenland Ice Sheet: high-elevation balance and peripheral thinning, Science, 289, 428-430.

Martinson, D. G., et al. (1998), Decade-to-Century-Scale Climate Variability and Change: A Science Strategy, 142 pp, Panel on Climate Variability on Decade-to-Century Time Scales, Board on Atmospheric Sciences and Climate, Commission on Geosciences, Environment and Resources, National Research Council, National Academy Press, Washington, D.C.

Meade, C., and D. T. Sandwell (1996), Synthetic aperture radar for geodesy, Science, 273, 1181-1182.

Meier, M. F. (1984), Contribution of small glaciers to global sea level, Science, 226, 1418-1421.

Oerlemans, J., and J. P. F. Fortuin (1992), Sensitivity of glaciers and small ice caps to greenhouse warming, Science, 258, 115-117.

Paterson, W. S. B. (1994), The Physics of Glaciers, 3rd edition ed., 480 pp., Elsevier, Yarrytown, NY.

Paul, F. (2000), Evaluation of different methods for glacier mapping using Landsat TM, paper presented at EARSeL-SIG-Workshop Land Ice and Snow, Dresden/FRG, June 16-17, 2000.

Popovin, V. V. (1996), Modern evolution of the Djankuat Glacier in the Caucasus, Zeitschrift für Gletscherkunde und Glazialgeologie, 32, 15-23.

Reeh, N., et al. (2002), Glacier specific ablation rate derived by remote sensing measurements, Geophysical Research Letters, 29, 10.1-10.4.

Rignot, E. J., et al. (1997), North and Northeast Greenland ice discharge from satellite radar interferometry, Science, 276, 934-937.

Rott, H. (1994), Thematic studies in alpine areas by means of polarmetric SAR and optical imagery, Advances in Space Research, 14, 217-226.

Serandrei-Barbero, R., et al. (1999), Glacier retreat in the 1980s in the Breonie, Aurine and Pusteresi groups (eastern Alps, Italy) in Landsat TM images, Hydrological Sciences Journal, 44, 279-296.

Sidjak, R. W., and R. D. Wheate (1999), Glacier mapping of the Illecillewaet icefield, British Columbia, Canada, using Landsat TM and digital elevation data, International Journal of Remote Sensing, 20, 273-284.

Thomas, R. H. (1993), Ice Sheets, in Atlas of Satellite Observations Related to Global Change, edited by R. J. Gurney, et al., pp. 385-400, Cambridge University Press, London.

Thompson, L. G. (2000), Ice core evidence for climate change in the tropics: implications for our future, Quaternary Science Reviews, 19, 19-35.

Thompson, L. G., et al. (2000), Ice core paleoclimate records in tropical South America since the Last Glacial Maximum, Journal of Quaternary Science, 15, 377-394.

Williams, R. S., Jr., and D. K. Hall (1993), Glaciers, in Atlas of Satellite Observations Related to Global Change, edited by R. J. Gurney, et al., pp. 401-422, Cambridge University Press, Cambridge.

Williams, R. S. J., and J. G. Ferrigno (1997), Final Report of the Workshop on Long-term Monitoring of Glaciers of North America and Northwestern Europe, 144 pp, USGS, Woods Hole, MA.