02/28/01 Martin et al. Do not duplicate or distribute - 2 -

A Calibration Method for Infrared Gas Analyzer (IRGA) - Based Soil Respiration Systems: An Artificial Flux Generator.

Jonathan G. Martin1,3, Paul V. Bolstad1, and John M. Norman2

Submitted to Global Change Biology as a note.

1 Department of Forest Resources, University of Minnesota. , 115 Green Hall, 1530 Cleveland Av. N., St. Paul, MN 55108 USA.

2 Department of Soil Science, University of Wisconsin-Madison, 1525 Observatory Dr, Madison, WI 53706 USA

3 Author to whom all correspondence should be addressed.


Abstract

An artificial flux generation system was developed to test the accuracy of a soil respiration system. The device consisted of an enclosed reservoir with a porous top and an Infrared Gas Analyzer (IRGA) that records the internal CO2 concentration of the reservoir. When the internal CO2 concentration within the reservoir is elevated, diffusion rates through the porous medium can be measured. This diffusion mimics natural soil respiration and can allow an independent verification of the accuracy of soil respiration measurement systems. By placing the soil respiration system to be tested on top of the porous layer of the flux generation system, deviations from a 1:1 relationship can be measured. A LI-COR 6400 portable photosynthesis system fitted with a 6400-09 soil CO2 flux chamber overestimated the generated flux rates by 3-4% when tested over five independent trials.

Introduction

Soil carbon fluxes are of great interest because of their impacts on global carbon budgets and carbon cycling. Approximately 2,100 Gt of carbon are stored in global soils (Bouwman, 1990; Gorham, 1991), and these stores could seep 50-75 Gt of carbon into the atmosphere annually through autotrophic and heterotrophic respiration (Raich and Schlesinger, 1992). Soil carbon cycles are complex and difficult to quantify; however, they are being better understood through advances in monitoring equipment. One such advance has been the portable infrared gas analyzer (IRGA) equipped with a chamber suited to measuring carbon dioxide fluxes at the soil surface. These devices have simplified soil respiration measurements, increased measurement accuracy, and allowed increased spatial and temporal sampling frequency when compared to older methods. However, these IRGA based systems do have their problems.

Estimates of CO2 flux from an IRGA are based on a measured increase in CO2 concentration within a measurement chamber, which is placed over a fixed area. This change results in CO2 evolution per unit area per unit time, commonly given in mg CO2 m-2 hr-1 or mmol CO2 m-2 sec-1. Unfortunately, improper use such as poor IRGA calibration, changes in pressure within the measurement chamber (Lund et al., 1999), uneven concentration gradients, and disturbances of the soil medium can affect diffusion and thereby give misleading flux rates. IRGAs may be calibrated with known gasses to ensure that the concentrations measured are reported accurately. Unfortunately, the physics of pressure and concentration gradients that are involved in quantifying a soil CO2 flux make testing soil respiration systems very difficult. While an IRGA may give the correct concentration, these readings may not be translated into a representative surface flux due to misuse or mechanical error.

Previous work has compared different measurement systems and techniques in attempts to validate estimates of soil CO2 flux. (de Jong et al., 1979; Norman et al., 1997). These comparisons were among measurement methods with unknown “true” values. The accuracy and repeatability of soil flux measurements have not been well documented, yet they should be if soil CO2 flux estimates are to be trusted. This paper describes a testing device that produces a known flux of CO2, which can be used to estimate the accuracy of IRGA-based soil respiration measurements. This calibration process should allow field measurements of fluxes to be used more confidently.


Materials and Methods

Flux Generation System Overview

The flux generation system is based on modifications of a design proposed by Nay et al. (1994; see Fig. 1). We constructed a device that works by diffusion and not mass flow. Earlier attempts that involved the active pumping of CO2 may have led to pressure differences and possibly caused underestimations of surface flux. Our flux generation system consisted of a large reservoir filled with carbon dioxide of a known concentration. Elevated CO2 within the reservoir creates a concentration gradient that drives CO2 diffusion through a porous medium at the top of the reservoir. This diffusion rate can then be calculated by measuring the decreasing internal CO2 concentration of the flux generation reservoir (Cireservoir) with an IRGA set in a closed cycle (reference IRGA).

CO2 efflux rates at the surface of the diffusion medium follows the equation (From LI-COR, 1997):

Where Rsgenerated is generated soil respiration (mmol m-2 sec-1), k is a units conversion = 10/8.314 = 1.2028, P is pressure (kPa), V is the reservoir volume of the flux generation unit (cm3) including all hoses and the approximate air space in the diffusion layer, S is the area of the diffusion opening (cm2), Cireservoir is the internal CO2 concentrations (mmol mol-1), t is the time (sec), and W is the concentration of water vapor (mmol mol-1).

The measurement chamber for the soil respiration system to be tested (test IRGA) is placed upon the diffusion medium and is used as it would be in the field . Care must be taken to note the time at which each “test” measurement is recorded so that it can be compared with the calculated fluxes from the flux generation system. If done cautiously, a wide range of generated fluxes can be produced. We were able to produce efflux rates from approximately 11.0 mmol m-2 sec-1 down to 0.5 mmol m-2 sec-1. This range of generated rates spans the range reported for soils from various biomes (Winston et al., 1997; Davidson et al., 1998).

System Construction

The flux generation reservoir consisted of two, 19-liter cylinders (two 5-US gallon utility buckets) affixed top to top with the upper most bucket having the bottom removed (Fig. 1). A metal screen (5 mm mesh size) was glued to a layer of silk, and then inserted inside the upper bucket approximately 10 cm from top. We placed 4 cm of glass beads approximately 0.65 mm in diameter (Potters Industries Inc., Valley Forge, PA) on top of the screen to act as a constant diffusion medium (Fig. 1a). This layer was carefully leveled to reduce the variation in flux rates across the surface caused by differing diffusion column lengths. We then measured the surface area of the diffusion layer (557.8 cm2) and the entire reservoir volume (38,594.1 cm3) including the hoses, reference IRGA volume, and the approximate pore space of the bead layer.

Input gas for the reference IRGA was drawn from directly below the glass beads through a series of perforated hoses distributed in a radial pattern (Fig. 1b). This inlet consisted of five perforated hoses with 1/8 inch (0.32 cm) inside diameter (Bev-A-Line IV tubing, Thermoplastic Processes, Millington, NJ). The perforations were spaced approximately 1.5 cm apart and were 1.5 mm in diameter. These hoses were joined to a sixth hose that led to the inlet of the reference IRGA. Exhaust gas from the reference IRGA was fed back into the system at the bottom of the reservoir (Fig. 1c) and was mixed by a slowly turning, 3 inch fan (Fig. 1d). The fan was powered by a variable-speed DC power source (BK Precision, Placentia, CA). A valve was fixed on the chamber so CO2-enriched air could be fed directly into the system prior to measurement (Fig. 1e). The easiest method of introducing CO2 into the system is to simply exhale though the hose until the desired Cireservoir is reached. The valve was closed prior to measurement.

A second fan was placed at an angle 20 cm above the bead surface to mix the boundary layer between the beads and the atmosphere (Fig. 1f). This was necessary due to the sensitivity of the generated fluxes to changes in the diffusion gradient. As the CO2 concentration within the flux generation reservoir decreased, the ambient CO2 concentration at the surface of the beads increased unless the boundary layer was mixed. A build up of CO2 at the surface of the diffusion layer greater than that set inside the test IRGA chamber will result in over-estimations of CO2 flux. This is due to a stronger diffusion gradient inside the test IRGA chamber and a relatively weaker gradient outside. Some soil respiration systems use a specified ambient CO2 concentration within the measurement chamber. This concentration is set close to ambient at the soil surface and then the chamber volume is partially scrubbed of CO2. Measurements are then recorded as concentration within the measurement chamber increases from below ambient to above it. For systems that use a preset ambient CO2 concentration (LI-COR 6400) or where the ambient CO2 concentration is adjusted manually (LI-COR 6200), care must be taken to ensure that the ambient CO2 concentration used in the test chamber is similar to that at the surface of the diffusion layer.

Leak Test

We conducted a simple leak test prior to removing the bottom of the upper bucket for the installation of the diffusion layer. The reference IRGA was attached to the reservoir and the internal CO2 concentration of the flux generation reservoir was increased to roughly 2800 mmol mol-1. This closed volume was monitored for 1.3 hours. The Cireservoir dropped 9.6 mmol mol-1, which resulted in a flux error of 0.06 mmol m-2 sec-1 or 0.6% of the corresponding generated flux (with Cireservoir ≈ 2800, Rsgenerated ≈ 10.5 mmol m-2 sec-1). This was well below the sensitivity of the test IRGA so this error was ignored.

Diffusion Lag Time

After the diffusion layer was installed, we measured the lag time for gas movement across the diffusion medium. The reference IRGA logged Cireservoir every second while the test IRGA, resting on the bead surface, was configured to do the same. Pulses of CO2 were introduced into the reservoir and were observed within seconds in the test IRGA. Thus, a drop in Cireservoir corresponded to a similar diffusion rate at the bead surface, so no time correction for the test IRGA measurements was needed.

Test Measurements and Flux comparison

For the reference IRGA, we used a LI-COR 6400 that had been re-plumbed to form a closed system. The test IRGA was a second LI-COR 6400 fitted with a 6400-09 soil CO2 flux chamber. The internal CO2 of the reservoir was elevated to ~3000 mmol mol-1, and the reference IRGA was programmed to log Cireservoir every minute. The maximum, accurately detectable concentration listed by the specifications of the LI-COR 6400 is ~3000 mmol mol-1, so care was taken to ensure Cireservoir remained less than this. The test IRGA was then readied as specified by the LI-COR 6400-09 soil CO2 flux chamber instruction manual and placed on a special 10-cm diameter PVC collar inserted into the diffusion layer. This collar was similar to the ones provided by LI-COR for use in field soil respiration measurements, but was 6 cm tall rather than 4.5 cm to ensure the top was well above the beads of the diffusion layer. Also, holes were drilled in this collar, well below the surface of the beads, to aid in lateral gas movement within the diffusion layer. Test measurements were then taken and the time carefully noted.

Once the testing cycle was completed (i.e. the Cireservoir was near the ambient concentration), we used equation 1 to calculate fluxes from the diffusion layer for the one minute increments between consecutive Cireservoir measurements (Fig. 2a). Then, using nonlinear regression (Rsgenerated = a*e(-b*time), JMP 3.2.5), we predicted Rsgenerated for each measurement from the test IRGA. The test and generated fluxes were then compared (Fig. 2b). This cycle of measurements was done five times at different locations on the diffusion surface. The five sets of comparisons were pooled and a t test was used to determine if the slope of the line was significantly different from 1. To help visualize the magnitude of the differences between the predicted and the measured values, we used linear regression (JMP version 3.2.5) to predict the measured test fluxes for each of the 5 trials at generated fluxes of 0.5, 2.5, 4.5, 6.5, 8.5, and 10.5 mmol m-2 sec-1.


Results and Discussion

The slope of the measured values in figure 2b was significantly different than 1 (n = 1157, β1 = 1.0512, p < 0.0001); however, this difference was very small. The trend shows an over-estimation at high rates and an under estimation at low rates. This error is much smaller than the 15% reported by Nay (1994). He reported on a different type of soil respiration system, which could indicate that systems might differ. For the LI-COR 6400 used in this study, the generated rates of 0.5, 2.5, 4.5, 6.5, 8.5, and 10.5 mmol m-2 sec-1 resulted in mean measured respiration rates for the 5 trials of 0.3, 2.5, 4.6, 6.7, 8.8, and 10.9mmol m-2 sec-1 respectively. With the exception of the low flux rates (0.5 mmol m-2 sec-1), the LI-COR 6400 over-estimated soil surface CO2 flux by 3-4 %.

We feel that the small discrepancies between test and reference measurements may stem from a changing boundary layer CO2 concentration. Although the addition of the boundary layer mixing fan helped to alleviate this problem, at high internal reservoir concentrations (e.g. 2000 – 3000 mmol mol-1) the ambient set point may have been lower than boundary layer concentration, resulting in a stronger diffusion gradient inside the test chamber. The opposite is true at low Cireservoir (e.g. 600-700 mmol mol-1 ) where the ambient set point may be too high. Attempts to increase the mixing of the boundary layer through additional fans created pressure differentials and lead to over-estimations; an effect similar to the concentration differentials. Careful monitoring of the ambient CO2 concentration and continually adjusting the ambient set point of the IRGA could possibly reduce these slight measurement errors.