On the Difficulty of Measuring Turbulent Fluxes Over Ice
The eddy covariance method is, in principle, straightforward: measure the covariance between vertical wind speed and some scalar quantity — temperature, water vapor, carbon dioxide — at high frequency, and you get a direct estimate of the turbulent flux. In practice, over an ice sheet at seventy degrees north, things get complicated quickly.
The Stable Boundary Layer Problem
Standard Monin-Obukhov similarity theory assumes a stationary, horizontally homogeneous surface layer with well-developed turbulence. Over a melting ice sheet in summer, you might get conditions that approximate this for a few hours in the afternoon. The rest of the time, you’re dealing with strongly stable stratification, intermittent turbulence, gravity waves, and conditions where the fundamental assumptions of the measurement technique begin to break down.
The literature on stable boundary layers is extensive and somewhat discouraging. Webb (1970) noted the difficulty decades ago, and while our instruments have improved enormously, the underlying physics hasn’t become any more cooperative. You can sample at 20 Hz with a state-of-the-art sonic anemometer and still find yourself staring at flux estimates that depend sensitively on your choice of averaging period, detrending method, and quality control criteria.
Instrument Challenges
Then there are the practical matters. Rime ice accumulates on sonic transducer paths. Radiation errors affect temperature measurements during calm, clear periods — precisely the conditions you’re most interested in from a surface energy balance perspective. Power systems fail. Data loggers fill up. Animals investigate your equipment with varying degrees of destructiveness.
A single field season on the Greenland Ice Sheet might yield three months of raw data, from which perhaps sixty percent survives quality control screening. Of that remainder, the periods of greatest scientific interest — strong stability, low wind, clear sky — are often the most contaminated by instrument artifacts and violated assumptions. There is a certain dark humor in this.
The Gap-Filling Question
Once you’ve accepted the losses in your time series, you face the gap-filling problem. You need continuous flux estimates to close the surface energy balance, but your measurements have holes. The standard approaches — marginal distribution sampling, artificial neural networks, mean diurnal variation — were developed primarily for vegetated ecosystems at temperate latitudes. Their performance over ice and snow has received less attention than it deserves.
I’ve been working on adapted methods that account for the peculiarities of polar sites: the extended periods of darkness or continuous daylight, the sharp transitions between stable and unstable conditions, the dominant role of the radiation balance relative to turbulent fluxes. The results are promising but not yet definitive. This is the kind of incremental, unglamorous work that constitutes most of actual science.
Broader Reflections
What keeps me coming back to this problem — literally, to the ice sheet, year after year — is the sense that getting the surface energy balance right matters for questions much larger than my own research program. The Greenland Ice Sheet contains enough frozen water to raise global sea level by seven meters. Whether and how fast it melts depends on the energy exchanges at its surface, which we are still learning to measure properly.
There is something humbling about deploying a three-meter instrument tower on a feature that covers 1.7 million square kilometers and hoping that your point measurements tell you something useful about the whole. The representativeness problem is real, and it connects to fundamental questions in observational science about scales, sampling, and the relationship between local measurements and large-scale processes.
The map is not the territory, but sometimes the territory is so large that maps are all we have. The challenge is making honest maps.
What Comes Next
Current work is focused on multi-year synthesis of flux measurements from several Greenland sites, with particular attention to inter-annual variability and trends. The goal is to move beyond case studies of individual melt events toward a more systematic understanding of how the surface energy balance is changing as the climate warms. This requires not just more data but more careful attention to consistency across sites, instruments, and analytical methods.
I’ll write more about the technical details in future posts. For now, I’ll note that the combination of improving instruments, longer time series, and growing computational resources makes this a genuinely exciting time to be working on these problems — even if the problems themselves remain stubbornly difficult.