Fig. 1.
Fig. 1.

Illustration of how a Type I error is manifested. An apparent positive flux (A) or a negative flux (Fig. 1B) is calculated when no actual change in headspace gas concentration is occurring.

 


Fig. 2.
Fig. 2.

Sample histograms and normal probability density functions constructed for N2O (A), CH4 (B), and CO2 (B) measurements in 35 air samples. Probability levels in each panel indicate the Type I error rate associated with rejecting the hypothesis that the data match the pattern expected if the data were drawn from a population with a normal distribution.

 


Fig. 3.
Fig. 3.

Apparent fluxes computed by different computation methods from three-point (A) or four-point data sets (B) selected from the population of 35 ambient air N2O samples shown in Fig. 2A.

 


Fig. 4.
Fig. 4.

Cumulative probability density functions of fluxes calculated with different flux calculation methods. The probability density curves were generated using Monte Carlo samplings of four-time-point data sets with a simulated 0.75-h chamber deployment time and at different analytical precisions. H/M, Hutchinson/Mosier; HMR, revised Hutchinson/Mosier; LR, linear regression; Quad, quadratic; rH/M, restricted Hutchinson/Mosier; rQuad, restricted quadratic.

 


Fig. 5.
Fig. 5.

Cumulative probability density functions of fluxes calculated with the six methods for an analytical precision of 0.04, chamber deployment time of 0.75 h, and sampling intensity of four points. Expanded views of the upper and lower tails of the distribution (boxes) are shown in Fig. 6. H/M, Hutchinson/Mosier; HMR, revised Hutchinson/Mosier; LR, linear regression; Quad, quadratic; rH/M, restricted Hutchinson/Mosier; rQuad, restricted quadratic.

 


Fig. 6.
Fig. 6.

Expanded views of the lower (A) and upper (B) tails of the cumulative probability distributions functions shown in Fig. 5. The dotted lines in the figures show how the fluxes of each method corresponding to the 0.05 (Fig. 6A) and 0.95 (Fig. 6B) probability levels were determined. H/M, Hutchinson/Mosier; HMR, revised Hutchinson/Mosier; LR, linear regression; Quad, quadratic; rH/M, restricted Hutchinson/Mosier; rQuad, restricted quadratic.

 


Fig. 7.
Fig. 7.

Relationships between analytical precision and the positive and negative flux detection limits of the different flux calculation methods when three-time-point (A) or four-time-point (B) data series are used. H/M, Hutchinson/Mosier; HMR, revised Hutchinson/Mosier; LR, linear regression; Quad, quadratic; rH/M, restricted Hutchinson/Mosier; rQuad, restricted quadratic.

 


Fig. 8.
Fig. 8.

Relationships between chamber deployment time and scaled slope factors used to compute flux minimum detection limits for data sets with three (A) or four (B) time points. H/M, Hutchinson/Mosier; HMR, revised Hutchinson/Mosier; LR, linear regression; Quad, quadratic; rH/M, restricted Hutchinson/Mosier; rQuad, restricted quadratic.