I saw part of a CBC show last night where they were discussing innovations in car technology – specifically safety sensors that sense how far you are from the car ahead, or how soon you have to hit the brake to avoid hitting a car coming from the side. They had multiple examples, including Google’s driverless car, and magnetic sensors to keep cars a set distance from each other. If you can’t wait to try some automation for yourself, Ford has come out with a Lincoln that will parallel park for you – you don’t even need to have your hands on the wheel!
The rationale for all these gadgets is billed as consumer safety, preventing collisions and reducing fatalities. But the one thing that seemed to come up repeatedly was everything else you could do while the car was driving for you. Check your email, work on your iPad, converse with clients or your boss. Something about that just didn’t sit right with me.
These technologies seem to be one more step in a long line designed to distance us from the here and now. We made a small step toward reducing attention required to drive when we went from manual to automatic transmissions. No longer did the driver have to pay as much attention to corners, hills, streetlights etc and shift down to accommodate them, as the car would do it for you. Now we’re taking another leap – to get people to trust technology enough to let the car drive for them. It’s a miracle: instead of focusing on driving you can do more important things like Facebooking your ‘friends’ that you’re driving hands-free, or checking your Twitter feed.
These changes aren’t specific to automotive engineering, but they’re also popping up in science. We’re increasingly distancing ourselves from the here and now (field data collection) in favour of more remote approaches (modelling, remote sensing, etc.).
During the second year of my PhD, one of my supervisor’s colleagues submitted a paper on our research. It was just when glaciologists were beginning to realize the importance of supraglacial streams in driving ice dynamics, and we’d measured and observed some interesting processes on an Arctic glacier that could help determine just how these systems worked. Part of the paper consisted of a description of the study site – specifically the configuration and morphology of the stream channels themselves, which were actually a series of long ponds connected by englacial channels that could be crossed on ice bridges. One of the manuscript reviewers was adamant that we had provided no proof of this channel configuration, thus (s)he didn’t believe it. Our field observations were inadmissible because the reviewer hadn’t seen them for his/herself, and (s)he couldn’t possibly believe our explanation.
I’ve often thought of what might have appeased this reviewer. Perhaps a map of the stream system, surveyed with a robotic total station or a differential GPS. Or perhaps they would have preferred a satellite image of sufficiently high resolution to show these streams. What both these solutions have in common is that they provide a ‘filter’ through which reality is observed. It’s no longer enough to have seen a physical process (or in this case a physical entity). We’re now required to step away from physical observation, and provide a proxy observation that’s considered more valid. As with the car technology, which advocated disengagement while driving from observing and interacting with the world around us, we cede the observational role to technology on the unspoken assumption that it can do it better.
More recently, a colleague and I received reviews on a paper about temporal changes in snow processes in forests infested by pine beetle. The gist of the comments from two of the reviewers was that the paper was largely based on field observations, and there should have been some modelling or satellite image interpretation to improve the science. Hmmm, ‘improve the science’? Since when did field observations become persona non grata in hydrology? Even during a job interview 6 years ago, I was told that my field program and research goals were fairly ‘conventional and basic’. The Chair of the department specifically asked what technologies I was using that were new and innovative.
It seems that field data on their own are too bland for science these days. They need to be spiced up with some numerical modelling, statistical tests (preferably Bayesian analysis), or remote sensing (whether that’s collection of LiDar data, interpretation of remote sensed imagery, or flying hi-res aerial photography/thermal imaging, etc.).
It’s not that I’m against these approaches – in fact our research group uses many of them in ongoing research programs. They provide valuable insights into hydrologic processes that we may never have derived from field data alone, and have significantly advanced our understanding of hydrologic linkages across scales. What I’m more concerned with is the denigration of direct field data itself, and the insistence that other approaches – that use more advanced technology – are somehow better.
Recent opinion pieces have documented the decline of field-based hydrology and the rise of modelling and statistical hydrology, and discussed the implications for our science. Stu Hamilton, in a 2007 paper in Hydrological Processes, argued that field data were critical for subsequent scientific endeavours, writing that “Many modelers view the world through the lens of their model algorithms, and sometimes this view of reality is as if seen through a kaleidoscope. In the absence of independent observations of reality how is any modeler to know whether their view of reality is realistic?” Roy Sidle, also writing in HP in 2006, observed that “the hydrological community is becoming more and more removed from process understanding based on direct field investigations”.
In the past, research papers were dense with qualitative field observation. For example, the first page of this 1932 paper on the Fiords of East Greenland (in The Geographical Review) contains an eloquent description of the geology, biology and glaciology of Franz Josef Fiord, linking it to locations such as the Grand Canyon or Spitsbergen. Obviously we’ve refined our field programs to collect more robust quantitative data that can give us a good understanding of physical processes. But it’s still observation- and measurement-based – and that seems to make some people uncomfortable.
Remember – modelling and remote sensing still require field data. The PUB initiative is a case in point, trying to understand basins that are ungauged and have no data. One solution is to use data from other catchments, which in some instances can be an excellent approach. Others require modelling or statistical approaches – usually predicated on existing data from other locations. Consider citizen science initiatives such as the Snowtweets project out of the University of Waterloo, or the Nature’s Notebook project, part of the USA National Phenology Network. These projects are based on field observations from a network of citizens, which are then used by researchers to quantify a range of other processes and interactions. But ultimately it’s the observations that are key – we’re not letting our science ‘car’ drive away without us being engaged with the physical world around us.
I’m interested to hear from others whether you’ve noticed changes in the relative merit of field vs. modelling/stats/remote sensing research. Maybe you have some good arguments why (or why not) field research is critical. Insights into this shift in research approach are also welcomed.