The SPRUCE experiment is being designed to explore these science objectives:
- To determine ecological responses for trees, shrubs, bryophytes, microbial communities and whole-ecosystem processes across a broad range of above- and belowground temperatures increases, and
- To understand how those responses to increased temperature will be altered by elevated atmospheric CO2 concentration.
The treatment levels and their allocation to the available experimental units on the S1 bog are being configured to provide optimal data for characterizing temperature response curves that might develop for plant or ecosystem level phenomenon. Temperature treatments will range from ambient condition to a +9°C differential from ambient, for both air and deep soil (Hanson et al. 2011), and will be repeated in combination with ambient or elevated CO2 atmospheres approaching 800 to 900 ppm.
Original Design Considerations
The original SPRUCE plan proposed an incomplete factorial using 28 experimental units that included four replicates of each of the following 7 treatments: Ambient plot, control plot at +0°, warmed plots at +3, +6, and +9°C and warmed plots exposed to elevated CO2 atmospheres at +3°C and +9°C. Although this approach is a viable option, project participants realized that the incomplete factorial design was statistically weak, open for criticism, and not the best approach for addressing our priority science questions surrounding responses to warming.
This engineering site plan for the S 1-bog shows the proposed access roads, parking, parking areas, and temporary office buildings in upland areas to the west side of the bog, and the extensive network of boardwalks that must be added to the bog itself to allow repeated access to up to 28 experimental units.
Regression-based Design Decision
Through quantitative analysis of different possible experimental designs, we concluded that a more flexible regression based experimental design including a broad range of temperature levels would provide us with more statistical power and better long-term data to characterize response curves for application within ecosystem and earth system models.
A regression-based experimental design appropriate for uncovering the shape of the temperature or other environmental response curves, has been proposed as a better experimental design with more flexibility than traditional analysis of variance (ANOVA). This approach will allow us to evaluate and parameterize response curves for the shape of previously unmeasured phenomenon (Cottingham et al. 2005).
Quantitative analyses with simulated data sets with a realistic variance structure revealed that a regression approach with 10 temperature levels had greater statistical power to resolve temperature responses than the originally proposed 16 plots arranged as four levels of temperature in four blocks, and at substantially lower cost and infrastructure requirements.
One plausible experimental design would include the following treatments and be randomly assigned to the available
experimental units in the S1 bog:
Ambient [CO2] enclosures at +0. +1, +2, +3, +4, +5, +6, +7, +8, +9°C
Elevated [CO2] enclosures at +0, +1, +2, +3, +4, +5, +6, +7, +8, +9°C
We will also instrument and monitor chamberless control areas to assess the influence of the chamber infrastructure on the bog.
If necessary, this combination of treatment plots occupying 20 experimental units might still be justifiably binned into low, medium and high temperature treatments for ANOVA based assessments for some variables. A design with fewer levels of temperature and three levels of elevated CO2 is also possible. Such a modification of the experimental design provides the flexibility to modify the number of treatment levels within the footprint of 28 potential experimental units to accommodate the costs of constructing, operating, and adding instruments to the warming enclosures.
In a regression based design the number of experimental units does, however, need to exceed the number of parameters included within a given regression equation to be evaluated.
Experimental units deployed over this broad range of temperatures would provide enough redundancy to protect against infrastructure failure while still allowing the flexibility to evaluate a range of forms for response curves. An important assumption is that there are no strong gradients across the experimental area that would mandate a block design. Preliminary survey data from the site justify making this assumption.
Cottingham KL, Lennon JT, Brown BL (2005) Knowing when to draw the line: designing more informative ecological experiments. Frontiers in Ecology and the Environment 3:145-152.
Hanson PJ, Childs KW, Wullschleger SD, Riggs JS, Thomas WK, Todd DE, Warren JM (2011) A method for experimental heating of intact soil profiles for application to climate change experiments. Global Change Biology 17:1083-1096.