What's new in version 5.3

SoilFluxPro® Software version 5.3 brings several new features and improvements. These updates make the software easier to use, but most importantly, they make it easier optimize computations.

Guidance

Chamber-based closed-transient measurements of gas flux from soils involve covering an area of soil with a chamber and recording a time-series of gas concentrations inside the chamber headspace. Fluxes are computed by fitting some portion of this time series with a mathematical model. Version 5.3 presents guidance for identifying the portion of the time series for the fit over which the observed data best fit the model assumptions.

Why guidance?

The time series of gas concentrations (referred to as an accumulation curve) during a chamber-based measurement can be divided into three phases. We call them turbulence development, simple diffusion, and lateral diffusion (Figure 1). In the first phase, which begins immediately after the chamber closes, the time series is dominated by the development of turbulence in the chamber. The second phase is dominated by vertical diffusion from the soil to the chamber. In the third phase, lateral diffusion begins to dominate observed changes in concentration.

The ideal window of the time series for the flux calculation includes as much data as possible from the second phase and excludes data from the first and third phases. Traditionally, this window of the time series was selected based on expert knowledge and examination of the data (often for only a subset of measurements) and applying the selection window to a broader dataset.

The Guidance tool removes the need for expert knowledge or subjective selection by automatically finding the window in the time series where the flux is least sensitive to the selection window. Guidance may be applied to a single flux observation, but is much more robust when applied to aggregate datasets. The benefit of this method is the ability to use an algorithm to identify the optimum window and to apply that algorithm to data from multiple observations.

Figure 1. Three phases of a soil gas flux observation are characterized by turbulence development, simple diffusion, and lateral diffusion. Observations are optimized when they include as much data as possible from the simple diffusion phase and exclude data from the turbulence development and lateral diffusion phases.

Turbulence development

The first phase, referred to as turbulence development, starts at the moment of chamber closure and extends into the accumulation curve to the point at which steady state mixing has developed in the chamber and system. The length of this period is impacted by volume and geometry of the chamber, transport delays in the system, physical response times for the gas analyzers, and surface characteristics of the soil being sampled. While some of these factors are constant and well known for a given instrument system like the LI-8250, others are not and are deployment specific. They are likely to differ between each chamber placement and may change throughout the course of an experiment. Where each gas is measured by a different analyzer in the system, they are likely to differ for each gas.

Simple diffusion

The second phase is characterized by simple diffusion. After the first phase, the behavior of the accumulation curve is well described by assuming that the diffusive flux is occurring between the chamber and the soil and the change in the chamber headspace concentration accounts for most of the change in the diffusion gradient. The length of this phase is affected by the diffusive transport characteristics of the soil being measured and the absolute magnitude of the flux being measured. The starting point, duration, and end point of the second phase will likely be different for each gas measured and each chamber deployment, and will almost certainly change over the course of an experiment.

Lateral diffusion

At some point, the conditions in the chamber deviate far enough from the assumptions that the model no longer describes the behavior of the curve. At this stage, the concentration through the soil profile is being altered and some significant portion of the soil gas may be moving laterally out from under the chamber. There is a gradual transition to this phase, occurring well after the chamber closes. Both the diffusive transport characteristics of the soil being measured and the absolute magnitude of the flux being measured affect the timing of this transition. It is likely to be different for each chamber deployment and each gas measured, and will almost certainly change over the course of an experiment.

Fitting the model

SoilFluxPro fits two mathematical models to derive the flux: a linear model and an exponential model. For any model, the window of the accumulation curve that is selected for fitting has a significant impact on the reported flux. For all fits, phase 1 must be excluded. This exclusion period at the start of the observation is called the deadband. Where the linear model provides an adequate representation of the accumulation curve, the flux is less sensitive to where the deadband is set in the start of phase 2. The exponential model, however, can be quite sensitive to this and for it, the deadband needs to be set at the transition from phase 1 to phase 2. For all fitting methods, a reasonable end point in the observation must be chosen during the transition from phase 2 to phase 3. This end point is referred to as the stop time.

The exponential model provides the best test case for evaluating the sensitivities of the deadband and stop time selections. By fitting it using a deadband starting from time zero and progressively increasing the deadband, you create a time series of flux over deadband. Or conversely, the fit may be done from an arbitrary deadband to the final point in the observation moving the stop time closer to the fixed deadband to generate a curve of flux versus stop time. These two curves represent the original analysis tools in SoilFluxPro that have been available for many years.

These curves are expected to exhibit some characteristic behaviors based on the contributions of the three phases. In the case of flux versus deadband, the phase 1 region will lead to a progressively changing or erratically noisy flux from the start. As phase 2 is approached, the curve should begin to approach a constant value and as it moves past the initial part of phase 2, the flux will begin to change or become noisy.

The same general behavior should be seen for the stop time analysis as it approaches the point where phase 3 completely dominates. This is expected from theory, but rarely does the analysis for a single observation conform to this expectation. Traditionally, stop-time analysis has been done by visual inspection, relying on expert knowledge, and has typically been limited to some subset of the total dataset.

Curves of dFlux/dDeadband and dFlux/dStoptime can be generated by taking the absolute value of the change in flux per deadband and change in flux per stop time. These are used to identify the transition points without the need for visual evaluation or expert knowledge. The point where the minimum value from either set is reached marks the effective transition between phases and as such, the appropriate deadband or stop time. On a single observation basis, this curve is subject to noise in the original measurement that may make selection of the real transition points difficult. By taking the average dFlux/dDeadband or dFlux/dStoptime of many observations, the selection becomes more robust.

The selection is made by finding time of all values within some percentage of the minimum dFlux/dDeadband or dFlux/dStoptime and averaging those times to select the deadband or stop time respectively. This minimizes the effects of outliers dominating the selection, and for aggregate data, ensures the selection is made on more than one point.

The procedure can be summarized in five steps: 1) compute a set of fluxes for all deadbands or stop times from time zero to the end of the observation for one or more observations comprising a logical aggregate of observations, 2) take the absolute value of the instantaneous derivative of that set for each observation, 3) average the derivative sets across observations, 4) find the deadband or stop time for all averaged values within some percentage of the minimum value, 5) take the average of the deadbands or stop times from the minimum set.

Using guidance

In SoilFluxPro 5.3, the Guidance feature can operate on a range of observations instead of on a single observation, as in previous versions. With multiplexed data, it is useful to apply guidance by port (chamber), for example. Guidance will give advice on two aspects of the data: Deadband and Stop Time.

Figure 2. After adding Port to the display list, you can sort by Port number. You can select all observations on a port, and this allows you to assess observations by port and thus, collar (and soil conditions for the observation).

Guidance with deadband analysis

Deadband analysis is designed to operate on multiple observations or single observations. Figure 3 shows how to start deadband analysis on observations from Port 8.

Figure 3. For guidance with deadband, select at least one observation and click Analyze.

The analysis is computationally intensive - it may take a moment. After the analysis is complete, you can choose a specific flux analysis to plot.

Figure 4. The plot shows change in flux as a function of deadband. The software finds the point where the flux is least sensitive to the start time and recommends the average value for that period as the deadband. This option is especially useful for gases that have a high concentration, such as CO2. Gas flux measurements of CH4 and N2O may not benefit from this procedure.
Figure 5. The more observations included in the analysis, the better the results. Here, the Deadband analysis includes observations grouped by chamber.

The plotted results show the current deadband, and the suggested deadband. If they are different, you recompute the results with the suggested deadband.

Guidance with stop time analysis

Stop time analysis optimizes the third phase of the observation. Optimizing the stop time ensures that the computations do not include the effects of lateral diffusion.

Figure 6. For guidance with stop time, select at least one observation, choose Stop Time Analysis, and click Analyze.
Figure 7. The more observations included in the analysis, the better the results. Here, the Stop Time analysis includes observations grouped by chamber.

The plotted results show the current stop time, and the suggested stop time. If they are different, you recompute the results with the suggested stop time.

Deadband flux report

The Deadband Flux Results Report presents a table for current and suggested deadbands, if you want the details as a table. You can also read values used for purple points in the plot by clicking Generate minimum dataset report.

Saving your favorite display setting

After loading a dataset, you can choose which parameters to display. If you have display settings that you like, you can save them as Favorites. Figure 8 shows how to add the Port number as a column and save the view as a favorite. Figure 9 shows how to delete favorites.

Figure 8. You can choose which columns are displayed and save the settings as a favorite. This steps above show how to add port to the display group and save it.
Figure 9. The previous display settings will load by default, but you can apply favorites to the current dataset. If a setting is no longer a favorite, you can delete it.

Although not a new feature, you can Export whatever is displayed in the view. The ability to save a favorite improves the utility of the export feature. If you want to export a subset of data or group of parameters, you can display them with a favorite and extract them with the Export feature.

Figure 10. You can save a subset of data with Export.

Repair

Previously, the Repair tool allowed you to replace a missing temperature measurement with a different one (called Replace Temperature by Port). Now it allows you to filter erroneous values and replace them with interpolation from adjacent values. Filtering is user-configurable, with settable minimum and maximum thresholds to remove outliers, and a graphical method.

Figure 11. The Repair tool allows you to replace erroneous values with interpolated ones.

Recompute

You can also filter when viewing charted data. For example, you can remove outliers graphically. And, if you remove something you wanted to keep, there is an undo button for that.

Figure 12. New Filter tools can be used to select outliers manually.

The graphical filter tools include the following:

  • Restore points: If you have selected points and removed them, click this to restore them to the chart and calculations.
  • Remove selected points: After selecting points with the rectangle or polygon selector, click this to remove them from the chart and calculations.
  • Select rectangle: Click this, then draw a rectangle around points to select them.
  • Select polygon: Click this, then draw a free-form shape around points to select them.
  • Keep previous selection: Undo a selection.
  • Clear: Remove a selection rectangle or polygon and clear the selection, leaving them intact.

Edit

Values in the Details view are editable.

Figure 13. Edit values directly by clicking the Edit button and then editing the parameter. Click the edit button to save the changes.

Pop-out windows

If you tire of scrolling up/down and left/right, you can pop out the window and expand it.

Figure 14. For larger windows, click the pop-out button.