goFlux: A user-friendly tool for calculating GHG fluxes regardless of experience

Non-steady state chambers are widely used for measuring soil and ecosystem greenhouse gas (GHG) fluxes, but causes non-steady state diffusion between the soil air and headspace, leading to non-linear behavior of gas accumulation over time during enclosure. In turn, linear regression (LM), commonly used to estimate GHG fluxes, may underestimate the pre-deployment flux (f0). Many alternatives to LM have been developed to provide a more accurate estimation of f0, for instance the method of Hutchinson and Mosier (HM), which accounts for non-linearity in gas accumulation during enclosure. However, non-linear models may overestimate f0, due to exaggerated curvature at time zero. Users therefore need to make subjective choices between LM and HM, often based on visual inspection or unsuited statistical metrics, such as R2, which can have profound impacts on end results. High-precision greenhouse gas analyzers, often combined with automatic chamber systems, promise to broaden our understanding of soil-atmosphere feedback in time and space. On the other hand, such systems produce enormous amounts of data that need to be processed automatically and consistently for reliable outputs: for example, by automatically selecting the best flux estimate based on either linear or non-linear models. At the same time, the number of researchers measuring soil-atmosphere fluxes from chambers is increasing and there is a need to develop GHG flux analysis tools that transcends prior user experience and produce the highest-quality data.We here present the goFlux R package designed as an all-inclusive flux calculation tool to calculate chamber GHG fluxes. goFlux allows for an easy import of raw data directly into R from a variety of instruments (LI-COR, LGR, GAIA2TECH, Gasmet, Picarro, Aeris and PP-Systems); simplifies identification of start and end times of individual flux measurements; quality checks the results based on objective criteria that goes beyond simply using R2; and provides the user with a recommendation for the best flux estimate. In summary, goFlux is meant to be “student proof”, meaning that no extensive knowledge or experience is needed for data import and pre-processing in R, and selecting the best flux estimate (LM or HM).In goFlux, a central element is to constrain the maximal curvature allowed due to non-linearity, by using the parameter of kappa-max (k), first introduced in Hüppi et al. (2018). The advantage of the k parameter is that it is based on objective metrics of instrument precision and chamber specific dimensions and applying k essentially avoids excessive flux overestimation, especially for noisy or small fluxes which often appears in chamber-based applications. Furthermore, following flux calculation, the best flux estimate is selected based on a user selection of multiple statistical criteria, such as the g-factor (ratio between LM and HM flux estimates) and indices of model fit (e.g. MAE, RMSE, AICc). In addition, poor quality measurements may be flagged based on minimal detectable flux (MDF), an intercept out of bounds, or due to insufficient number of observations.For more information, visit our webpage: https://qepanna.quarto.pub/goflux/