So, you’ve got a set of chambers and you want to measure greenhouse gas emissions such as CO2 flux from a site. Now what? You want to be able to accurately comment on C budget by quantifying the soil CO2 flux at a site, but there is an impressive (and daunting?) level of spatial and temporal variability in soil CO2 flux.
Important questions arise: How many chambers do I need? Where do I place them? How often should I sample? First, it is important to consider the factors that influence soil CO2 emissions.
What are the drivers of soil respiration?
There is a large variation in soil physical properties, biological conditions, and nutrient availability over small distances (< cm3), which results in high spatial variability. Even at the scale of centimeters, high and low levels of soil respiration can be measured next to each other, highlighting the existence of “hot spots”.
What controls soil CO2 emissions? The drivers of soil respiration at a site-level scale are: soil microclimate (is it warm and wet enough?), litter thickness and nutrient availability, and fine root biomass. Soil respiration rates can also be reduced or enhanced for short time periods (hours to days) when soil diffusivity changes, due to altered transport of gases to the surface.
At the daily (diurnal) scale, three main factors are drivers for soil respiration: dependence on 1 – temperature: rise and fall with Q10 near two within typical limits and optimum moisture; 2 – moisture (humidity, synoptic/weather events); and 3 – substrate supply correlated with photosynthetic lags up to 7-12 hours (Tang and Baldocchi, 2005).
Attempts to characterize the site variability in order to obtain accurate estimates of soil gas fluxes can be biased or can be altered by systematic error in your estimates and the practical considerations of cost/labour/access.
How many chambers do I need?
A large number of sampling points are ultimately required in order to capture the spatial variability, and can be estimated from the coefficient of variance (CV), a measure of the relative error compared to the mean. This has been shown to vary between 10 to 150% in forests, grassland, and agricultural systems (Luo and Zhou, 2006).
The number of sampling points is also limited by labour (portable systems) and cost (automated systems). To capture the variability of a site, estimates for the number of randomly sampled points is high (> 30 samples), other sampling design approaches are commonly used: random and stratified (or 2-stage) sampling. Random sampling within a site would seem to be an unbiased approach to describe the variability, where each position within a site could have an equal chance of being selected. This requires a high number of samples. However, stratified sampling, where the same level of accuracy can be achieved by random sampling, with fewer samples: in fact, independent sampling of subpopulations was found to outperform random sampling (Rayment and Jarvis 2000; Rodeghiero and Cescatti 2009; He et al. 2016).
Where do I place them?
As an example, Xu and Qi (2001) stratify the site into gaps and covered areas, because this is an important control on soil respiration, since fine root biomass is higher under canopies (stable conditions), but identifying the strata (categories) is site specific. One approach identified by Rodeghiero and Cescatti (2009) is to carry out an intensive preliminary characterization on a high number of points across the site, thereby enabling identification of the relevant site factors. This retains the statistics by optimizing the selection of sub-samples. A transect (systematic sampling) is also a form of stratified sampling where sampling sites are spaced along an environmental gradient. For instance, at Howland Forest, soil respiration in the footprint of the eddy covariance tower is measured using automated chambers in several sites (replicated for > 3 sites within several meters) in areas that have contrasting moisture characteristics.
How often should I sample?
The other major query is how often to sample – i.e. at what temporal scale are you able to measure with your equipment and what scale are you concerned with? One can sample at 5 minute resolution, but in a system with stable microclimate, this level of detail would not be necessary to describe the soil respiration and its relationship to temperature. However, measuring only once per day misses or biases diel and synoptic events and soil’s response (e.g. the large reduction in soil fluxes right after a rainfall) (Savage et al., 2008).
However, if labour and cost constraints are relevant, then a particular time of day is important to identify an optimal representation of the daily mean: mid-morning (between 0900 and 1100) fluxes approximate the 24 hour flux (Larionova et al., 1989 and Davidson et al., 1998; Xu and Qi, 2001) at midmorning, it is 0.9 to 1.5% of sampling error. Refer to an excellent 2015 white paper by Eosense’s VP R&D Chance Creelman that deals this with this issue and also examines errors associated with the sparse temporal data.
Since soil respiration often lags behind temperature, a relationship between soil respiration and temperature is often not straightforward, and this varies with moisture content (Gabriel and Kellman, 2014; Phillips et al., 2011). It is thus important to also be able to measure temperature and moisture at the same time, and at the appropriate depth (appx. 5 cm) and frequency to suit your study and research questions.
The main take-away is that decisions about sampling (how many, where and when) are often site- and study-dependent, but efforts in the early stages of design are well-worth the effort in order to minimize labour and to still confidently report carbon and other greenhouse gas exchange estimates.
Chamber-based systems that are robust and low cost allow for a larger number of sample sites to be measured than other, more cost prohibitive and spatially-limited systems and allow for a deployment of chambers measuring at high temporal frequencies (5 minutes). Using a self-contained autonomous system like the eosFD may result in better accuracy and precision for soil respiration data, especially when a stratified sampling design is adopted.
Creelman, C. (2015) Biases of sparse data and measurement scheduling on flux estimates. White paper: Eosense, Inc. Dartmouth, NS, Canada.
Davidson., E. Belk, and R. Boone (1998) Soil water content and temperature as independent or confounding factors controlling soil respiration in a temperate mixed hardwood forest, Global Change Biology, 4: 217-227.
Gabriel, C. and L. Kellman (2014) Investigating the role of moisture as a constraint Soil Biology and Biochemistry, 68: 373-384.
He, Y., J Gubbons, and M. Rayment (2016) A two-stage sampling strategy improves chamber-based estimates of greenhouse gas fluxes, Agricultural and Forest Meteorology, 228: 52-59
Larionova, A., L. Rozanova , and T. Samoilov (1989) Dynamics of gas exchange in the profile of a gray forest soil, Soviet Soil Science, 3: 104-110.
Luo, Y. and X. Zhou (2006) Soil Respiration and the environment. Elsevier Academic Press: Massachusetts.
Phillips, C., N. Nickerson, D. Risk, B. Bond (2011) Interpreting diel hysteresis between soil respiration and temperature. Global Change Biology 17 (1), 515-527.
Rayment, M. and P. Jarvis (2000) Temporal and spatial variation of soil CO2 efflux in a Canadian boreal forest, Soil Biology and Biochemistry, 32(1): 35-45.
Rodeghiero, M. and A. Cescatti (2009) Spatial variability an optimal sampling strategy of soil respiration, Forest Ecology and Management, 255(1):106-112.
Savage, K. E. Davidson, A. Richardson and D. Hollinger (2008) Three scales of temporal resolution from automated soil respiration measurements, Agricultural and Forest Meteorology, 149: 2012-2021.
Tang, J. and D. Baldocchi (2005) Spatial-temporal variation in an oak-grass savanna ecosystem in California and its partitioning into Autotrophic and heterotrophic components, Biogeochemistry 73(1): 183-207.
Xu, M. and Y. Qi (2001) Soil-surface CO2 efflux and its spatial and temporal variations in a young ponderosa pine plantation in northern California, Global Change Biology 7: 667-677.