Data Package Metadata   View Summary

NSF Grant numbers DEB-1636476, DEB-1026415, DEB-0620579, DEB-0423442, DEB-0080609, DEB-9810217, DEB-9211769, DEB-8702629

USDA Forest Service, Pacific Northwest Research Station (Agreement # RJVA-PNW-01-JV-11261952-231)

corrections

Light responses are modeled monthly, therefore if too much data is lost, or has to be filtered out, the model parameters cannot be determined, for plots where this was the case plot-specific light response parameters from the previous year were used. In 2015 several plots had insufficient data to accurately evaluate model parameters for hyperbolic light responses. Prior to substituting 2014 parameters to 2015 data, we evaluated model fits using both paratmeters for plots with sufficient data. 2014 parameter estimates had good correspondence with 2015 estimates (mean R2=0.97 for 2014 modeled vs 2015 modeled parameters, ranging from R2=0.66 to R2=1). 2014 parameters were substituted for fence 6, plot 11 in May and September. This plots frequently floods, preventing the chamber from being placed on the plot. In May and September this is particularly problematic. In May the data did not fit a normal light response, so 2014 parameters were substituted as a best estimate, and in absence of better models that capture the response of CO2 flux to water table changes. In September no data was collected on this plot, again 2014 parameters were used to estimate fluxes. In contrast o light responses, the temperature responses for estimating ecosystem respiration are modeled for the entire growing season, therefore missing data has a much smaller effect on the parameter evaluation. None of the Reco parameters had to be replaced. In 2017 Fence 5, plot 10 had unreasonably high CO2 losses that could not be explained or disqualified (similar to CiPEHR 4-3 in 2015). The solutions in 2017 was: Fence 5, plot 10 in all other months had typical uptake and respiration patterns so 1 June - 7 July 2:30am data were removed, models reparameterised without those points, and NEE fluxes modeled with June 2016 parameters. The high flux data are included in a seperate column (that are excluded from the cumulative calculations NEE_g_high) because they represent extreme outliers and do not capture the average behaviour of this tundra site. Because this is a recurring behaviour the data are kept here for completeness and for others to puzzle over. In 2018 Fence 6, plot D was modeled with normal procedures PAR<5 = Reco with exponential temperature response and PAR>5 = NEE with light response curve. Fence 6, plot C in May had low CO2 uptake at PAR<15 and the light response function had a very low y-intercept and predicted flux values that were well out of range and reasonable biological values so NEE was modeled using the Reco function for PAR<15 in May only.

additional

Sampling: Continuous, with a measurement every 1.5 hours at each plot location. One control plot and one plot of each experimental treatment (total 24 plots/CO2 flux chambers, 4 chambers per fence) are measured at each fence at any given time. Chambers are rotated every 3, 5, and 7 days to obtain regular measurements from all 66 experimental plots. From 2013 onward the experimental site has experienced very dramatic subsidence, particularly on the soil warmed side of the fences. As a result many areas are periodically, or constantly, flooded depending on summer rainfall patterns. By the end of 2015 subsidence was preventing the placement of flux chambers on flooded plots (2-3 plots total). In 2016 CO2 flux chambers were rebuilt to allow CO2 flux measurements under flooded conditions, and we were able to collect data all season. After 2016 flooded plots often showed little seasonal or diurnal variation in CO2 flux and were thus decoupled from light and temperature dynamics, which resulted in poor predictability. In order to maintain a consistent gap-filling approach between plots and years we continued to use these models and report model fit (pseudo R2 = 1 - total sum of squares/residual sum of squares) and significance of parameters in the supplemental file so that users can make their own determination. Poorly fitting models estimated a mean CO2 flux and therefore represent neither an over or under estimation of cumulative CO2 flux. If R2 or parameter p-values are NA it is because this plot had insifficient measured data to fit a model and substituded parameters were used to predict CO2 flux. See Natali et al 2011 and 2014 for model details. Briefly, the light response is a hyperbolic function where a is the initial linear slope of the light response curve (quantum yield), GPmax is maximum photosynthesis at light saturation, R is dark ecosystem respiration. The temperature response is an exponential function where A is the basal respiration rate and B is the temperature sensitvity. In 2018 some plots had no light response all season and exponential temperature response parameters were estimated from the full NEE dataset to estimate daytime NEE with a temperature response function. Where applicable hese parameters are indicated by a/b_all. Instumentation: Campbell CR1000 data logger, Licor Li-820 CO2 analyser. Statistics/Algorithms: Exponential temperature response functions, hyperbolic light response curves.

Status

Ongoing