Interactive comment on “ Constraint of soil moisture on CO 2 efflux from tundra lichen , moss , and tussock in Council , Alaska using a hierarchical Bayesian model ”

General comments: Kim et al. and colleagues used a Hierarchical Bayes (HB) model to identify dominating factors controlling CO2 efflux from an arctic tundra ecosystem based on two years growing season field measurements. They used a well-established statistical method to fit the data and reproduced observed CO2 efflux. The layout of the manuscript is straightforward and easy to understand. However, the data presented were relatively limited (only one plot) and the findings were not quite new and novel. At its core, HB is a device using sub-models to account for uncertainty. This study does not really fully explored the advantages of HB model in the uncertainty analysis of the parameters and the subsequent knowledge we obtained from constrained parameters and model,

model.The authors misunderstood model descriptions in Biogeosciences discussion and have revised to describe them correctly.
The correct model description is as follows, 2.3 Description of Hierarchical Bayesian (HB) model In a HB model, in order to evaluate the relationship between CO2 efflux and environmental variables, we modeled observed CO2 efflux using a HB model with four explanatory variables: soil temperature (ST), soil moisture (SM), vegetation types (Vege), and thaw depth (THAW).First, CO2 efflux (FCO2) was assumed as normally distributed with mean parameter (µflux) and variance parameter (ð İIJ Ő): F_(ã Ȃ ŰCOã Ȃ Ů_2 ) â Ĺijnormal (µ_flux,σˆ2).(4) The scale parameter (µflux) was determined from the following equation: µ_flux=f_P f_ST f_SM f_THAW, (5) where fP represents the function of CO2 efflux potential, and fT and fSM are limiting response functions ranging from 0 to 1. fP was defined as follows: f_P=β_0+ ã Ȃ ŰVegeã Ȃ Ů_[k] + ã Ȃ ŰYearã Ȃ Ů_[l] + ã Ȃ ŰPosiã Ȃ Ů_ [ij] .( 6) fP is a linear predictor with intercept ('β0') and three random effects (Vege, Year, and Posi).The Posi term represents a spatial random effect from a conditional autoregressive model (CAR) proposed by Besag et al. (1991).Temperature (fT) is a modified van't Hoff equation as follows: f_ST=eˆ((STã Ȃ ŰSTã Ȃ Ů_ref)/10 logâ Ą ą(Q_10)), (7) where fST is the temperature response function, which varies from 0 to 1.The explanatory variable of this function, represented by ST and STref, is a constant, set at 25 • C in this study.The temperature sensitivity parameter is Q10.The soil moisture liming function (fSM) is defined as follows: f_SM= where the soil moisture response function is fSM, ranging from 0 to 1, and is the same as the temperature response function (Hashimoto et al., 2010).SM is the explanatory variable of this function, and a, b, c, and d are parameters for determining the shape of the soil moisture function.The function has a convex shape, and values range from 0 to 1. Parameters a and c are the minimum and maximum values of SM, respectively (i.e., g(a) = g(c) = 0).Parameter b, which ranges between a and c, is the optimum parameter (i.e., g(b) = 1).Parameter d controls the curvature of the function, though C4294 the three other parameters also affect the shape.This function was adopted from the DAYCENT model (Parton et al., 1996;Del Grosso et al., 2000).
P5906,L28: vegetation type was not really an explanatory variable in this study.Like variable "year", it was introducing uncertainty into model prediction resulting from vegetation type variability (in other words, it was formulated as random effect in the prediction model).Is variability from vegetation type separable from interannual variability?Are those two parameters correlated?
In this plot, the vegetation is perennial.Change in the vegetation within the plot is mostly not observed in this study period.Therefore, in theoretically, these two parameters are not correlated with each other.Actually, there were very low correlation (R = 0.137) between τ veg and τ year in our result.
»> We added to P5918 L3 explanation at the suggestion of Reviewer #2.
Because changes in vegetation within the plot were not observed in this study period, these two parameters are not correlated with one another.Actually, there was very low correlation (R2 = 0.019) between ï Ąt'veg and ï Ąt'year in our results.
P5906, L29: "under assumption of lognormal distribution" In the methods section all probability distributions are either normal or uniform, where did you use lognormal distribution?
P5907, L2-3: As I mentioned earlier, I don't think that under current model formulation it C4297 is possible to evaluate the characteristics of dominant plants on CO2 efflux (unless you account for variation of other environmental variables).However it would be accurate to say that you evaluated random effects on CO2 efflux introduced by vegetation types, assuming they are separable from the random effect of "year".
P5910, L19: variables beta1 and beta2 are not shown in the equation, and they are not shown in Table 3, where do they come into play?
»> We are sorry for the confusion regarding section 2.3 on the HB model.The authors misunderstood model descriptions in Biogeosciences discussion and have revised to describe them correctly, as previously described.
P5911, L7: please, include units and definition of variable WHPS (and THAW as well) »> The WFPS does not exist in this manuscript and has been corrected.
P5911, L8: "a, b, c, and d are the parameters" »> Explanation for parameters a, b, c, and d was added in P5911, L8: The function has a convex shape, and values range from 0 to 1. Parameters a and c are the minimum and maximum values of SM, respectively (i.e., g(a) = g(c) = 0).Parameter b, which ranges between a and c, is the optimum parameter (i.e., g(b) = 1).Parameter d controls the curvature of the function, though the three other parameters also affect the shape.This function was adopted from the DAYCENT model (Parton et al., 1996;Del Grosso et al., 2000).
P5912, L10: again, beta1 and beta2 are not shown in the equations, and they are not

C4298
shown in the joint posterior probability and P5913, L14-15: environmental variables among the plots with different species differ.Can the differences in CO2 efflux be attributed to environmental variables rather than species cover?
»> Strictly speaking, we agree with these comments, regarding different species indicating differences in CO2 effluxes under different environmental variables.However, much higher CO2 efflux in tussock tundra was observed than in other species, as previously reported (Oechel et al., 1997;Fahnestock et al., 1998).Fahnestock, J. T., Jones, M. H., Brooks, P. D., Walker, D. A., and Welker, J. M.: Winter and early spring CO2 efflux from tundra communities of northern Alaska, J. Geophys., Res., 103, D22, 29023-29027, 1998.As I mentioned earlier the Results and Discussion section should be carefully revised.Please, make sure that your conclusions are supported by clearly stated evidence.For instance, the conclusion from P5913, L21-23 states that "suggesting that CO2 efflux in tussock is a significant atmospheric CO2 source, ten times greater than in wet sedge", however it is not supported by evidence the way it is given earlier in the sentence.C4299 »> We rewrote the sentence from P5913, L21-23, as suggested by R#2.
CO2 efflux in tussock and wet sedge was 0.23 and 0.022 mgCO2 m-2 min-1, respectively (Oechel et al., 1997), suggesting that CO2 efflux in tussock is a more significant atmospheric CO2 source than in wet sedge.This may be due to the difference in the size of the tussock covered by the chamber.
P5913, L23-24: what does this sentence suggest?The conclusion I should draw from this sentence does not seem very clear.Paragraph on pages 5913-5914 needs to be broken down into 2 or 3 paragraphs.
P5914, L16-29: I think the results will have better flow if changes in the environmental variables are described first, followed by description of changes in the CO2 flux.
P5915, L6: "significant" instead of "significantly"; where is the result showing one-way ANOVA for thaw depth?
P5915, L7-8: the statement that thaw depth was not related to CO2 flux and soil temperature contradicted results in Figure 5.
The distribution of thaw depth (not shown) seems similar to the pattern of soil moisture, which is inversely related to those of CO2 efflux and soil temperature.3, and are often outside of the 97.5% confidence interval.It would be very interesting to see C4300 the explanation for the differences in the values.Where the differences caused by variation in soil moisture, thaw depth, and/or other factors?
»> We derived Q10 values suggested in Table 2 from the relationship between CO2 efflux and soil temperature alone; however, the Q10 value reported in Table 3 is from the HB model.According to soil temperature as well as soil moisture/thaw depth from the HB model, the Q10 values from Table 3 may be much lower than those from Table 2, due to the inverse relationship between CO2 efflux and two parameters for the entire growing seasons.
Table 3: where in equations was the term "deviance" estimated?»> Deviance is the index of fitting the model to observed data, and not parameter.It may be confusing here.The fitness of the model is described in the following Figure (RMSE and ME), rather than the deviance.

Figures 2
Figures 2 and 3: I don't think figures 2 and 3 are critical to show in this study »> We deleted Figure 3, as suggested by both R#2 and R#1.

Figure 6 :
Figure 6: this figure repeats what is already shown in figure 1 and figure 5 »> We deleted Figure 6 and provided description in the text, and added to Figure 1 the explanation of accumulative rainfall in 2011 from Figure 6, as suggested by R#2: This seems to be the effect of heavy rainfall since August 20, 2012, as shown in Figure 1, which represents daily and accumulative precipitation in 2011 and 2012.Interestingly, accumulative rainfall began to surpass 2011 accumulative precipitation in August 20, 2012 (not shown).

Figure
Figure7: it seems that temperature limitation function is well constrained unlike moisture limitation function or thaw function.Why do you think they are unconstrained?Can it be related to different vegetation types?It would be interesting to estimate parameters from table 3 for each vegetation type separately (except the standard deviation

Fig. 1 .
Figure 1 Davidson et al. (1998)reported CO2 efflux increased with soil moisture of 0.2 m3/m3" I think giving an interval would be more appropriate, e.g."with soil moisture from 0 to 0.2 m3/m3" »> I rewrote the following, as suggested by R#2.L7-10: such high Q10 value may not be a true temperature response value.The burst in CO2 efflux in spring may be due to release of CO2 trapped in soil over winter as described in Elberling and Brandt[2003]»> We appreciate your comments; Higher CO2 concentration in frozen soil came from a spring burst event during soil thawing, and also related to the trapping of produced CO2 during the winter.Also, there is a distinct difference in Q10 value above and below zero; Q10 value below zero was 430 when water content was 39 %(Elberling  and Brandt, 2003).On the other hand,Monson et al. (2006a; b), as noted, observed a much higher Q10 value of 1.25 × 106 in the beneath-snowpack soil of a subalpine forest in early spring.
diagnostic as an index.For the model, we ran the Gibbs sampler for 20,000 iterations, for three chains, with a thinning interval of 10 iterations.We discarded the first C4295 10,000 iterations as burn-in, and used the remaining iterations to calculate posterior estimates.R was used to call JAGS/WinBUGS and calculate the statistics in R. »> We cited the reference in the introduction of P5906 L7-10, as suggested by R#2.Monson, R. K., Lipson, D. L., Burns, S. P., Turnipseed, A. A., Delany, A. C., Williams, M. W., and Schmidt, S. K.: Winter forest soil respiration controlled by climate and microbial community composition, Nature, 439, doi:10.1038/nature04555,2006a.Monson, R. K., Burns, S. P., Williams, M. W., Delany, A. C., Weintraub, M., and Lipson, D. L.: The contribution of beneath-snow soil respiration to total ecosystem respiration in a high-elevation, subalpine forest, Global Biogeochem.Cycles, 20, GB3030, C4296 doi10.1029/2005GB002684,2006b.P5906,L11: soil temperature is an analogue of soil microbial activity only under certain assumptions, e.g.under an assumption that soil moisture and substrate availability are not limiting factors.»> We have added this comment to P5906 L11, as suggested by R#2.

Table 2 :
Q10 values in this table are different from the value reported in Table Vege parameter), and see whether parameter values were significantly different from each other.This way it would be possible to estimate the effect of vegetation on the environmental limitation function.»>Thankyouforyourcomments.We tried to model different sensitivity for each vegetation type.However, we failed to estimate some (divergent) sensitivity due to some vegetation types with small numbers of samples.Figure 8: not sure this figure is essential to present for this study »> We deleted Figure 8, as suggested by R#2.Figure 9: this figure is useful to illustrate how well your model represents the data used for calibration, however, model validation is an essential stage in model development.I suggest merging the data from 6 panels into one, and do some data mining from the literature to find co2 efflux, thaw depth, soil moisture etc to fit the model for validation.An example for model validation data could be data from Oberbauer et al.[1992], who also estimate model parameters to CO2 flux data.It would be also interesting to see whether the model in this study performs better than the model presented in Oberbauer et al.'s study.»>Iappreciateyoursuggestion.This reference is very important for our study.However, there were no data regarding soil moisture.On the other hand, our study also lacked observation regarding the soil water table.If we had observations for the water table, we could conduct the study you have noted here.The empirical model fromOberbauer et al. (1992)is very similar to our model.We have cited this model in the Introduction section of P5906, L16.We have aggregated six panels into one figure as follows.Reference: Oberbauer, S. F., C. T. Gillespie, W. Cheng, R. Gebauer, A. S. Serra, and J. D. Tenhunen (1992), Environmental effects on CO2 efflux from riparian tundra in the northern foothills of the Brooks Range, Alaska, USA, Oecologia, 92(4), 568- 7: it seems that temperature limitation function is well constrained unlike moisture limitation function or thaw function.Why do you think they are unconstrained?Can it be related to different vegetation types?It would be interesting to estimate parameters from table 3 for each vegetation type separately (except the standard deviation C4301 for theInteractive comment on Biogeosciences Discuss., 11, 5903, 2014.