fluxx: Estimate (ghg) flux rates from online dynamic closed chamber measurements in through-flow mode.

Description

(Bulk) estimates of (ghg) flux rates from medium frequency (mf) online concentration measurements with dynamic closed chambers in through-flow mode. The function tries to find stable linear conditions in concentration change and automatically detects and excludes concentration fluctations.

Usage

fluxx(x, var.par, subset, asterisks = FALSE, loop = "auto", ...)
mf.flux(x, var.par, method = "r2", time.unit = "S", all.through = TRUE, iv = 1, 
wndw = 0.1, pdk = 0.5, min.dp = 20, nrmse.lim = 0.1, r2.qual = 0.9, 
range.lim = 5, out.unit = "auto", elementar = FALSE, 
hardflag = list(range = TRUE))

Arguments

A list with data tables as returned by chop or just one of such data tables (for mf.flux which will rarely be called directly by the user). Each table contains data for one chamber measurement. Usually there will be identifying c

var.par

A named list specifying the variables and parameters that are used in the estimation process and variables that should be handed through the function so that they are easily available for further analysis. Some of the names are obligatory, others are opti

subset

An optional vector specifying a subset of concentration measurements to be used in the estimation process.

asterisks

Logical. If TRUE p-values are given as asterisks.

loop

Logical or character. Set the mode how bulk estimations are run. mf.flux tries hard to find the linear part of the concentration development during chamber deployment. It does so by running many regressions. Therefore, when used in bulk mode

...

Further arguments passed to mf.flux.

method

Character string specifying the statistic used for finding the linear part. Partial match to "r2", "rmse", "AIC". Defaults is "r2". This tries to find the linear part whilst maximimising the number of me

time.unit

Single character giving the time unit. Will typically be seconds, thus default is "S". Other options are "M" for minutes and "H" for hours. Be aware that fluxes will be incorrect when this setting does not reflect th

all.through

Logical. When TRUE, all data columns in x other than the ones needed for flux calculation are also handed through the function so that they can be used in later steps of analysis easily. You may also specify one or a few columns

Numeric. Sometimes there is no time information at all but the rows in x are just numbered consecutively. The correct temporal spread is calculated inside the function when the measurement interval is specified here. Defaults to 1 which expec

wndw

Numeric between 0 and 1. Relative width of a moving window in which the standard deviation of the concentrations is calculated to identify high frequency fluctuations. See details and next.

pdk

Numeric between 0 and 1. Minimum proportion of data points to be kept. See details. In case one single concentration value exists more than pdk * n times in the data, all other data is assumed to represent high frequency fluctuations and flux

min.dp

Numeric. The minimum number of data points. Defaults to 20. If there are less rows the estimation is run anyway but a warning is issued and min.dp is automatically adjustet to n-1.

nrmse.lim

The maximum acceptable normalized root mean square error. Numeric value between 0 and 1. Defaults to 0.1. If the final best solution has a higher nrmse it is flagged accordingly.

r2.qual

Numeric between 0 and 1. Quality parameter for the model fit. The minimum acceptable r2 of the best fitted model. Defaults to 0.8. A measurement run is not defined invalid when the value is below the quality setting. Instead, a quality flag is reported.

range.lim

Numeric. The minimum range of the concentration measurements during one chamber placement. The acceptable range limit depends on the accuracy of the concentration measurements. When the range of the concentration measurements is smaller than the repeatabi

out.unit

Output unit of the flux rate mass part. Character string. Defaults to "auto". In this case, the function tries to find a unit that ranges the output value between 0.01 and 10 which is fine for figures. Can be "ng", "mug", "mg", or "g". "mug" stands for "$

elementar

When the fluxes are wanted as element values set elementar = TRUE. Defaults to FALSE.

hardflag

Named list that controls which of the quality flags are to be hard flagged (the value is changed according to the quality flag). Default is that the range.lim is hard flagged. So when the range of the actual concentration values for a chamber

Value

fluxx returns a complex object of class fluxxes that is a 2 entry list. When the object is printed to the console only the second entry is displayed in a modified form that is meant to maximize information display with small footprint for easy inspection. A table is printed to the console with three columns per gas. The first contains the quality flags (e.g. 111.9). The order is: r2.f, range.f,nrmse.f, nomba.f. The first three are considered more important, and if they are '1' everything is fine. For the one behind the dot its a bit different: It just gives the number of measurements below ambient.
The data.frame with the estimated flux rates contains all data needed for further analysis. The columns represent the entries in fluss of the single chamber measurements (including quality flags, see below) plus naming information according to the settings in the nmes argument of chop. export provides a simple way to export the results.
The first entry is itself a list of lists and data tables. It is called flux.res. The only one first level entry in this list contains the information for one gas which is itself a list. In this list each first level entry contains the information for one chamber measurement. It is named according to the nmes-setting in chop and contains the elements fluss (which is itself a list with the elements given below), mod, out (a list with hand through data, list items according to columns in x that have been handed trough via all.through or var.par), and inn - a data.frame with the input data that were relevant for estimating the flux (the obligatory part of var.par).
The elements of fluss:
ghgCharacter. The gas species for which the flux has been estimated.
fluxNumeric. Calculated flux rate in mass unit per m2 and hour.
r2r2 of the best fitted model that has been used for flux caclulation.
nrmsenrmse of the best fitted model that has been used for flux caclulation.
r2.fLogical. r2 quality flag telling whether the r2 quality setting given in r2.qual is fulfilled.
range.fLogical. Range quality flag telling whether the range of the concentration measurements exceeded the quality range of the measurement device that has been specified in range.lim.
nrmse.fLogical. nrmse quality flag telling whether the nrmse quality setting given in nrmse.lim is fulfilled (i.e. if the nrmse of the best model <= nrmse.lim).
nomba.fInteger. Reports the number of measurements below ambient. When one observes concentrations below ambient that might make the measurements unstable, it is possible to filter the results later and allow only a maximum acceptable number of measurements below ambient. The ambient concentration is build into the function with data from Mace Head Ireland (N2O, CH4) and global average (CO2) obtained from http://cdiac.ornl.gov/pns/current_ghg.html as of August 1st, 2011.
unitThe mass unit assigned.
podpuProportion (expressed as a number between 0 and 1) of data points used for constructing the linear model for estimating the flux rate. The higher the less disturbed the measurements.

encoding

UTF-8

Details

The function is similar to flux but uses a different algorithm to identify the most linear part of the concentration development. First high frequency fluctations are omitted. Then all possible pdk * n : n consecutive concentration measurements are regressed against the corresponding times. The model with the highest r2 is chosen.

Probably the most important argument is var.par. It specifies the variables (by referring to the names of the data columns) from x and parameters (fixed values that are constant for all chamber placements) that are used for the flux estimations. The approach is similar to the one used in flux with the exception that there is no list item for the device quality flag. For obligatory var.par items see flux and examples.

In contrast to flux there is just one workhorse function doing the actual estimation (mf.flux) per data table. Especially when there are many data tables in x and/or many data points per data table it takes some time. Progress is shown in the console. Each dot represents one finalized data table.

References

Nakano T (2004) A comparison of regression methods for estimating soil-atmosphere diffusion gas fluxes by a closed-chamber technique. Soil Biology and Biochemistry 36: 107-113.

Forbrich I, Kutzbach L, Hormann A, Wilmking M (2010) A comparison of linear and exponential regression for estimating diffusive CH4 fluxes by closed-chambers in peatlands. Soil Biology and Biochemistry 42: 507-515.

Examples

Run this code

## load data
data(tt.nee)

## prepare flux estimation
# make parts with chop
tt.parts <- chop(tt, factors=c("session", "spot"), 
nmes=c("spot", "date", "session"), min.cm=40)
# prepare var.par list (like with flux)
vp <- list(CO2 = "NEE", time = "datetime", area = "area", 
volume = "volume", t.air = "t.cham", p.air = 101325)

## do the flux estimation
# run fluxx. with lots of data it may take a while 
# (approx. 10 sec per chamber)
tt.flux <- fluxx(tt.parts, subset=c(1:30), vp, pdk=0.5, 
range.lim=3, out.unit="mg")
# inspect results table
tt.flux
# plot diagnostic plots
plot(tt.flux, dims=c(4,4), subs="spot")
# run fluxx with alternative method
tt.fluxa <- fluxx(tt.parts, subset=c(1:30), vp, pdk=0.5, 
range.lim=3, out.unit="mg", method="rmse")
# inspect results
tt.fluxa

Run the code above in your browser using DataLab