Hiking with Reviewer 2
This is a deep dive into my own research — the backstory behind a single line in a recently published paper and the data-driven trip down memory lane that was spurred by an innocent question from Reviewer 2.
This research took place on Wabanaki land. I want to respectfully acknowledge the Maliseet, Micmac, Penobscot, and Passamaquoddy tribes, who have stewarded this land throughout the generations. I am certainly not the first person to devote time and energy to tracking seasonal changes on Mount Desert Island.
This week one of my dissertation chapters, ‘Trails-as-transects: phenology monitoring across heterogeneous microclimates in Acadia National Park, Maine’ was published in the journal Ecosphere. In this project, I pulled the space-for-time trick and hiked three mountains repeatedly to collect a lot of phenology observations across diverse microclimates. The mountains in Acadia are not huge — these granite ridges roll up from the Gulf of Maine and top out at 466 m — but my transect hikes were between 4.8 km and 9.7 km each, and I wore out a pair of trail runners each season. I took to heart Richard Nelson’s advice: “There may be more to learn by climbing the same mountain a hundred times than by climbing a hundred different mountains.”
A couple months ago, in our second round of reviews, Reviewer 2 noted, “I think that it would be useful for those wanting to replicate your transect-as-trails approach (especially land managers) to know approximately how many person hours it took to complete a transect observation, here in the main text or in the appendix.”
I had a magnet (which is apparently also available as a coaster) hanging next to my desk in grad school: over a silhouette of a golden retriever with three tennis balls in its mouth, it reads: “If it’s worth doing…it’s worth overdoing.” This magnet perfectly describes my response to Reviewer 2. I sent a back-of-the-envelope estimate to my coauthors, but I couldn’t shake the feeling that the precise person hours per transect was a knowable statistic. In addition to my field notes scribbled into weatherproof notebooks, I had collected my data via fulcrum, a smartphone app that automatically recorded the time of each observation. From my cache of fulcrum csv and xlsx files, I should be able to automatically pull the time of the first and last observation of each transect. The 10.7 MB of data in my fulcrum files represented four years of field work, hours and hours on the trails, slogging through rain, snow, and sun, training field assistants, combing through patches of lowbush blueberry and mountain cranberry for the first, hidden open flower. I became obsessed with the idea of seriously calculating person hours per transect, but I was increasingly convinced that a single number would be meaningless. I also realized that I lacked the coding chops to deal with my messy raw data: 171 files, each with 77 columns, usually containing data from a single transect, but occasionally comprising half a transect (when we had to bail due to weather) or more than one transect (when I ran ambitious double-days, or my field assistants and I split up). I turned to Porzana Solutions, and Auriel Fournier expertly helped me unlock my person hours data.
Over 177 the hikes in my fulcrum files, the mean time between first and last observation is 3.51 hours.
Three and a half hours does not even begin to tell the story. This blog post is my second supplemental appendix. Here is the story of person hours per transect — the lead time, the pregnant field season, and the phenology of phenology monitoring.
Before the first observation and after the last
There is a lead time in every transect hike. After rolling out of bed, pulling on the same old running shorts, race tshirt, and powder blue sunglasses, after packing the same handful of granola bars, dried papaya, and sharp cheddar, zipping my phone into its waterproof case, and slinging my backpack into the passenger seat, after driving to the trailhead and placing my research permit on my dashboard, there’s still a gap between the start of the fieldwork and the first official observation of the day. Especially as the summer crowds began arriving in June, I had to get out early to grab a spot at the limited parking by the north or the south end of Pemetic, or else add some extra miles from a spillover lot*. Even at the best parking spot, the approach to the Sargent South Ridge trailhead requires navigating 0.7 miles of carriage roads between the car and the trail on every hike. When I started the project in 2013, the Sequester kept Park Loop Road closed late into the spring season. For the first six weeks of fieldwork, I walked along the empty road to access Cadillac North Ridge, and Pemetic North and South Ridge.
The transect hikes were 4.8 km (Pemetic), 9.2 km (Cadillac), and 9.7 km (Sargent) up the North Ridge and down the South Ridge or vice versa (all of the mountains had uncreatively named north and south ridge trails). So at the end of a transect, I was 4.8, 9.2, or 9.7 km away from my car. I could run the carriage roads to connect the trailheads after Sargent or Pemetic (a 6.6 km run post-Sargent, and 7.2 km run post-Pemetic). From Cadillac South Ridge, a run up Route 3 to park loop road got me back to the north ridge trailhead in 10 km. Sometimes I arranged rides with friends to skip the run, and when I had funding for field assistants in 2015 and 2016 we often carpooled to drop a car at the finish line for each other. (There were some benefits to this running routine — in 2014 I won free ice cream after placing third in my age group in the Acadia Half Marathon.)
The person hours per transect statistic is limited because not every transect was a straight shot. Sometimes we had to bail 3km into a hike due to bad weather and finish the transect another day. Once, one of my field assistants took a wrong turn and recorded phenology observations on the wrong trail down Pemetic, and so I went back, retraced her steps, and picked up the right trail the next day. Once, I did a wild two-a-day and in the middle of Cadillac, I ran down the Canon Brook Trail, looped through the Pemetic transect, and then ran back up the Cadillac West Face Trail to finish Cadillac. Once, I had a friend in town and we caught a ride to the summit of Cadillac and then enjoyed the leisurely hike down the south ridge with my eight-month-old in the baby backpack.
While the time between first and last observation averaged just over 4 hours for Cadillac, 2.5 hours for Pemetic, and 3 hours and 40 minutes for Sargent, those times discount the bookends of the hikes. As much as I’m railing against the answer to my query here, the process of working with Porzana Solutions to calculate these times has been incredibly rewarding. I feel like I’m getting to know my both raw data and the tidyverse in a weirdly intimate way that goes way beyond a standard tutorial.
The pregnant field season
In 2015 I was 17 weeks pregnant at the start of my field season. In addition to my daughter, I was also joined in the field by two field assistants. According to the Porzana analysis, I hiked less than half as many transects in 2015 (15) compared to each of the two previous years (2013: 35** hikes, 2014: 37 hikes). I actually hiked 20 transects that year — my assistants were entering the data (and getting credit for the hike in fulcrum) while we hiked together in the beginning of the season***. On my solo transects in 2015, I felt sloooooow. I averaged thirty minutes slower than 2013 and 2014 on Cadillac, 50 minutes slower on Pemetic, and 22 minutes slower on Sargent. On top of this, I was covering less ground — in 2013 and 2014 I had monitored phenology in off-trail Northeast Temperate Network plots near my transects in an effort to compare trail-side phenology with forested sites that was ultimately cut from my dissertation. In 2015, I stuck to the trails.
I remember feeling pretty terrible at the beginning of most hikes that year. I had one favorite spruce tree on the south ridge of Sargent, and I can picture myself looking up through the needles on more than one occasion from my lie-down-spot while I tried to decide if a bite of granola bar would make me feel more or less nauseous. As I climbed above treeline and into the breeze the fog of morning sickness would lift, and as I hiked downhill, my daughter would do this funny little fetus-roll and kick in a way that I interpreted to be happy.
Hiking while pregnant was hard, but it felt easier than grappling with the looming challenges of becoming a parent. I liked the hard of fieldwork, it was the kind of hard that I felt capable of conquering. I also loved being pregnant in Bar Harbor. It was my fifth field season in Acadia and I had this wonderful community of supportive colleagues and mentors at the park service and in town. I had a favorite yoga class, a favorite milkshake, a favorite iced chai and blueberry muffin spot. I also had two field assistants — my pregnancy fortuitously aligned with NSF funding! — and working with Ella and Natasha that season was great. The person hours per transect figure obscures my field assistants, folding us into each other and masking the time we spent training together on the ridges. It also hides my pregnancy in the averages. I want to recognize those extra 22-50 minutes: they were some of the best worst minutes of my PhD.
The phenology of phenology monitoring
The person hours per transect average doesn’t show the sprint finishes of June. I monitored thirty species (the paper highlights the 9 most common taxa) of spring-flowering plants. On the transect hikes, I recorded leaf out and flowering phenology. In April, this was a bit of a scavenger hunt, and I’d pour over thickets of shrub stems for the first sign of bud break, then in May I’d peek into each curled Canada mayflower leaf for flower buds. By early June, my plants had leafed out, and the flowering season was winding down. I knew the trails by heart, and the location of each focal taxa along the ridge was bright in my mental map; each transect became a point-to-point trail run between the last phenological hold outs. Did the rhodora finish flowering on Cadillac? Had the last sheep’s laurel buds opened on Pemetic? Were the blueberries beginning to ripen below Sargent’s summit?
As I followed the spring phenology, I grew faster, my calf muscles more defined, my appetite more voracious. Acadia’s steep climbs will whip you into shape. I remember in 2013 arriving in the field a month after passing my comps and feeling so sluggish after a winter of studying instead of running. In comparison, I ran hard in the winter of 2013-2014, set a personal best half marathon time in a trail race in March, and just cruised through the early season field work in 2014. Even in 2015, as I grew rounder each week, I also grew more comfortable with the trails. Hiking while pregnant became easier over the season, although I’m happy it ended when it did, because that trend was not sustainable into the third trimester.
I think about Reviewer #2 and I want to ask: do you mean the person hours per transect in April? Or at the end of June? What kind of mileage were you averaging before the start of the field season? Do you have any old hamstring injuries? Tell me about your field assistants. Do you like to stop for lunch at the summit or are you an on-the-go-snacker? Did you pack a couple bucks to buy a Harbor Bar at the Cadillac souvenir shop? Are you saving your energy for the 10k run at the end of the transect? Is the National Park Service well-funded in this year’s federal budget? How do you feel about stopping for a swim in Sargent Mountain Pond?
I love these questions because each one pulls on a thread winding through my Acadia memories. I hiked upwards of 125 transects between 2013 and 2016, and now that the paper is done, I’m a little sad to be shelving the fieldnotes for good. The trail runners that I wore are long gone, my field hat fell apart, most of my baggy race tshirts carried me through my second pregnancy and suffered for it. In the end, the idiosyncrasies of the hikes were smoothed and flattened into the sentence, “Each transect could be completed in under 6 person-hours.” This is both true and wildly circumscribed. Not unlike a well done chapter of a PhD dissertation.
*Acadia National Park actually closed the lot by the Pemetic North Ridge trailhead in 2017 and it’s now exclusively a bus stop for the island explorer, the free bus that begins running right as my season wraps up at the end of June.
**This doesn’t include hikes before I had figured out the fulcrum platform. There was “no” data on those hikes (nothing was leafing or blooming, no signs of budburst) and they only exist in my field note books.
***I hired three field assistants for this project and, concurrently, a common garden experiment. In 2014, Paul was my garden guy, but we also hiked two transects together and he hiked two solo. In 2015, Ella, Natasha, and I split the transect and garden work. Ella came back for most of the 2016 season and then I finished the two projects solo in June 2016.