Supplemental Electronic Material
Appendix A. Freshwater marsh monthly predictive equations for methane flux
For the freshwater marsh, we ran the analysis on the measured mean monthly methane flux as the dependent variable and the year was not differentiated. For selecting variables for the final regression (PROC GLMSELECT in SAS) we used alpha =0.15. In this model we added the square and natural log of the class variables of water temperature, salinity,water elevation and diversion discharge. Next we ran the final regression model based on these variables at alpha=0.05. Mean monthly methane flux was predicted according to the following equation, which applies to each month:
FCH4_pred = 0.4497 - (Tw*0.01715) + (S*5.6909) + (Ew*0.1177) + (Tw2*0.000568)–(Ln S+1)*7.1815)-
(Ln Ew+1)*0.1774)–(Ln D+1)*0.01404) + residual
Where,
FCH4_pred= predicted mean monthly methane flux (µmol/m2/s)
Tw = water temperature (°C)
S = salinity (psu)
Ew= water elevation (ft NAVD88)
D = diversion discharge (ft3/s)
residual (t) = 0.33772*residual(t-1) + 0.13797*residual(t-2) + error(t)
The residuals of the final model met the requirements for normality and homogeneity. The R-square was 61.96% for the final mode with Root MSE=.050573. The residuals of the regression model followed and Autoregressive(2) process with the errors being normally distributed with constant variance(alpha=0.01).
Appendix B. Brackish marsh monthly predictive equations for methane flux
For the Brackish data we ran the analysis on measured mean monthly methane flux, where wesubsequently took the square root of the measured value and a constant (FCH4_measured +0.013302217) to accommodate negative fluxes. Months were not differentiated among years. For selecting variables for the final regression(PROC GLMSELECT in SAS) we used alpha =0.15. In this model we added the square and natural log of the class variables of water temperature and salinity. Next we ran the final regression model based on these variables at alpha= 0.05. The overall equation is the samefor eachmonth except the given coefficient in Table B1:
FCH4_pred = (0.8444 - 0.02084*S + 0.000238*Tw+ 0.001026*S2- a*(Ln Tw+1))2-0.013302217+ residual
Table B1
Month / Monthly Coefficient (a)Jan / 0.23762
Feb / 0.24045
Mar / 0.23589
Apr / 0.229818
May / 0.2235
Jun / 0.2143
Jul / 0.21398
Aug / 0.19375
Sep / 0.20272
Oct / 0.21743
Nov / 0.230527
Dec / 0.235
Where,
FCH4_pred= predicted mean monthly methane flux (µmol/m2/s)
Tw = water temperature (°C)
S = salinity (psu)
a = monthly coefficient
Note that each of the above equations by month has an overall residual, where residual(t) = 0.40312*residual(t-1) + 0.18904*residual(t-2) + error(t)
The residuals of the final model met the requirements for normality and homogeneity. R-square was 67.00% for the final mode with Root MSE=.037581. The residuals of the regression model followed an Autoregressive (2) process with the errors being normally distributed with constant variance (alpha = 0.01).
1