Harvard University Graduate student, Nate Edelman, was able to analyze the data that thousands of K-12 students have contributed to the Harvard Forest Schoolyard Ecology program over the past 12 years.
So what is all the data showing anyhow?
Nate has created a series of graphs from the complex and large Buds, Leaves and Global Warming project dataset. We are pleased to share these graphs with you now.
Until now, we have focussed most of our data analysis efforts on helping you, teachers, to understand our Project Ecologists' long term data as well as to represent patterns in your own site's data. Last year, we were challenged by The Long Term Ecological Research (LTER) network to do some synthesis of Schoolyard data over time and across sites to see if any patterns are evident.
How to interpret these graphs:
According to Nate Edelman, the blue lines on these graphs is a "smoothed Loess regression" calculated by R (Graphing program), which uses information from before and after each date to estimate the mean value for that date. The grey area is the 95% confidence interval for that estimation. The black dots are the calculated means using only information from that exact day.
According to Nate Edelman, the blue lines on these graphs is a "smoothed Loess regression" calculated by R (Graphing program), which uses information from before and after each date to estimate the mean value for that date. The grey area is the 95% confidence interval for that estimation. The black dots are the calculated means using only information from that exact day.
Notice that the line drops towards the end of the season in these 2 figures above. We would not expect the line to go down because more leaves would presumably fall at the end of the growing season. Please see Nate's comments followed by my comments below to understand why the curve took this unexpected turn.
Nate's thoughts about the drop in percent leaf fall of Oak and Beech at the end of the season:
I've plotted a histogram of the number of observations taken on each Julian data and superimposed it on the leaf fall graph.What they show is that the number of observations reported dropped significantly at the end of the year, and that's when we get the unexpectedly low values. I have an idea for how to get around this but didn't have time to test it out yet.
Schoolyard Coordinator notes: Please keep in mind that both species of Oak and Beech frequently retain leaves through much of the winter. The leaves turn brown but many do not fall in the autumn when leaves from other species typically fall. Some researchers stop monitoring leaf drop once most of the leaves have turned brown. Our protocol asks to record oak and beech leaves as fallen after they have turned completely brown, in order to show that those leaves have stopped photosynthesizing and therefore the growing season is officially over for those leaves. Sometimes, researchers forget to mark the leaves as "fallen" and this affects the ability to accurately graph the end of the growing season. I'm interested to see that even Red Maple leaves fallen go down which must be due to far fewer observations at that time per Nate's notes above.
I've plotted a histogram of the number of observations taken on each Julian data and superimposed it on the leaf fall graph.What they show is that the number of observations reported dropped significantly at the end of the year, and that's when we get the unexpectedly low values. I have an idea for how to get around this but didn't have time to test it out yet.
Schoolyard Coordinator notes: Please keep in mind that both species of Oak and Beech frequently retain leaves through much of the winter. The leaves turn brown but many do not fall in the autumn when leaves from other species typically fall. Some researchers stop monitoring leaf drop once most of the leaves have turned brown. Our protocol asks to record oak and beech leaves as fallen after they have turned completely brown, in order to show that those leaves have stopped photosynthesizing and therefore the growing season is officially over for those leaves. Sometimes, researchers forget to mark the leaves as "fallen" and this affects the ability to accurately graph the end of the growing season. I'm interested to see that even Red Maple leaves fallen go down which must be due to far fewer observations at that time per Nate's notes above.
Comparison of Schoolyard Data to Project Ecologist, John O'Keefe's Data
The next thing I did (above) was to take the two species we talked about, red oak and red maple, and compare them to John’s data. Qualitatively, the data for average 50 percent leaf fall and bud burst look similar, but when we look at the leaves on days, there are some fairly major discrepancies. In John’s data, there is a nice consistent separation of about 3 weeks of leaves on days between the species. However, the schoolyard data has some years in which the maple has a longer growing season than the oak, and in all cases the difference was much smaller.
Finally (Below), I used the same species and tried to get a more geographical perspective. I combined the data by town, and made two maps for each species. One is the average number of leaves on days across the whole experimental period. The size of the dot corresponds to the number of years of data we have, and its color corresponds to the number of leaves on days.
Below are map graphs of the most common 3-5
species with the larger circles showing where there are more trees of a
certain species. I looked at the number of individuals
of each species, and a natural cutoff seemed to be at 4 species (Red Oak, Red
Maple, Sugar Maple, and Beech). I grouped all the trees by town, and plotted
circles on those towns with sizes corresponding to how many trees of each
species were used. I made two graphs – the first one is of unique trees, and
the second is of all observations. So if a certain town had a single Red Maple,
it would show as a small circle on the first graph, but if they had followed it
for several years and had many records, it would be relatively larger on the
second graph. I also included the total numbers next to the key. These graphs
can unfortunately be a little hard to read because so many of the points
overlap one another.
Below, I
made the J.O.K.-style graphs and the leaf fall curves for each of the four species
mentioned above, as well as black birch and black cherry
-Nate Edelman
What is the spatial distribution of trees in the study?
Finally (Below), I used the same species and tried to get a more geographical perspective. I combined the data by town, and made two maps for each species. One is the average number of leaves on days across the whole experimental period. The size of the dot corresponds to the number of years of data we have, and its color corresponds to the number of leaves on days.
-Nate Edelman
-Nate Edelman
The other graphs below show the change in leaves on days across the
time period. I only included sites that had at least three years worth of data
for this, and the color corresponds to the slope of the regression
line. Positive slope (and red color) indicates an increase in growing
season length. Negative slope (and blue color) indicates a decrease in growing
season length. There is a big caveat here, which is that to get the growing
season each year, I averaged the 50p bud burst and the 50p leaf fall for that
town and took the difference. That would be fine if there was lots of data, but
sometimes there were very few (or only one) trees, and they might be different
trees for the bud burst date vs the leaf fall date! This is in contrast to
John’s data, where we are looking at the exact same tree each season and each
year.
-Nate Edelman
