Experimental Design

The SPRUCE experiment was designed to explore the following science questions:

Will deep belowground warming in the future release 10,000 years of accumulated carbon from peatlands that store 1/3 of the earth’s terrestrial carbon? At what rate?
Will releases of C be in the form of CO₂ or CH₄ with 30 times the warming potential?
Are peatland ecosystems and organisms vulnerable to atmospheric and climatic change? What changes are likely?
Will ecosystem services (e.g., regional water balance) be compromised or enhanced by atmospheric and climatic change?

SPRUCE will determine ecological responses for trees, shrubs, bryophytes, microbial communities and whole-ecosystem processes across a broad range of above- and belowground temperatures increases, and understand how those responses to increased temperature will be altered by elevated atmospheric CO2 concentration.

Implemented Design

The treatment levels and their allocation to the available experimental units on the S1 bog were configured to provide optimal data for characterizing temperature response curves that might develop for plant or ecosystem level phenomenon. Temperature treatments range from ambient conditions to a +9°C differential from ambient, for both air and deep soil (Hanson et al. 2017). All warming treatment are repeated in combination with ambient or elevated CO₂ atmospheres approaching 800 to 900 ppm.

Aerial photograph of the SPRUCE experimental site on 5 August 2015. Plot numbers (1 to 21) and assigned temperature treatments are superimposed on the image. Dashed circles indicated established plot centers for plots that are monitored annually for tree growth. Plots 4, 10, 11, 16 and 19 receive elevated CO2. The middle boardwalk is 112 m long.

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 was considered 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 S1-bog shows the proposed access roads, parking, parking area, 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, was deployed as a better experimental design with more flexibility than traditional analysis of variance (ANOVA). Our approach allows 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. However, funding limitations did not allow this design to be fully constructed.

Instead, an experimental design with 5 warming levels was constructed and is operated in the S1 bog. Treatments are as follows:

Ambient [CO₂] enclosures at +0, +2.25, +4.5, +6.75, and +9°C

Elevated [CO₂] enclosures at +0, +2.25, +4.5, +6.75, and +9°C

We also instrumented and monitored chamberless control areas to assess the influence of the chamber infrastructure on the bog.

Full-scale prototype of the enclosure for testing evaluation and continued improvement constructed at ORNL

Practical Considerations

If necessary, the combination of treatment plots occupying 10 experimental units might be justifiably binned into low, medium and high temperature treatments for ANOVA based assessments for some variables.

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 a broad range of temperatures provides us with some 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.

The SPRUCE Experiment is a large-scale, largely automated whole ecosystem manipulation with 10 enclosures warmed in a regression design including 5 temperature treatments plus ambient plots.

References

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, Riggs JS, Nettles WR, Phillips JR, Krassovski MB, Hook LA, Gu L, Richardson AD, Aubrecht DM, Ricciuto DM, Warren JM, Barbier C (2017) Attaining whole-ecosystem warming using air and deep soil heating methods with an elevated CO2 atmosphere. Biogeosciences 14: 861–883, doi: 10.5194/bg-14-861-2017