This report examines the application of the Cross Sectional Flux method (CSF) to emission estimates from the Hail Creek mine, as presented in Sadavarte, et al., 2021 . To support and contextualise this analysis, a software package replicating the CSF algorithm GitHub TROPOMI_CSF was developed and released on GitHub.
The CSF method relies on ERA5 meteorological fields and TROPOMI observations; consequently, its performance is partly determined by the quality of these input datasets. Both TROPOMI mission data and ERA5 reanalysis products are publicly available through the Copernicus programme. Assessing the quality and validity of these datasets is beyond the scope of this report, and the software developed for this study treats the TROPOMI and ERA5 inputs as accurate. At the time of writing, early TROPOMI products have been reprocessed multiple times using different processor versions. The original analysis by Sadavarte, et al., 2021 likely used data processed with versions 010202, 010300, 010301, and 010302, which were available prior to 2022, although the authors appear to have evaluated multiple product versions.
The output of the GitHub TROPOMI_CSF implementation was compared using two publicly released TROPOMI processor versions, 010302 and 020400, for orbit 09956. The comparison demonstrates that the choice of image processor has a substantial impact on the derived emission estimates, even when the same CSF methodology is applied. It should also be noted that striping artefacts in TROPOMI imagery significantly affect the determination of both plume shape and background concentration, and these issues influence results obtained with both processor versions.
Sadavarte, et al., 2021 applied additional preprocessing to the TROPOMI data, including adjustments related to striping artefacts and pixel height. In contrast, the GitHub TROPOMI_CSF implementation deliberately omits such modifications in order to avoid introducing arbitrary alterations to the distributed datasets. Responsibility for applying, justifying, and documenting any additional data adjustments is therefore left to the user.
Using the GitHub TROPOMI_CSF software, we investigated the factors contributing to the high implied emission factor for Source 1 (the Hail Creek open cut mine). The elevated values are likely attributable to a combination of the following factors:
The determination of plume extent and background concentration is the most critical step in all emission estimates based on box model approximations, including CSF, TM, IME, and related methods.
Minor inconsistencies were identified in the activity data used in the original study, as these values do not appear to align with reported coal production figures for 2019 (see “Coal Production”).
The CSF implementation in GitHub TROPOMI_CSF exhibits low numerical stability. Emission estimates are highly sensitive to the choice of pixel geometry and masking strategy, which are selected primarily for computational convenience rather than being constrained by the physical characteristics of plume transport (see “Computational stability”).
Due to the mine's location, in a valley bounded by elevated terrain to the west and northeast, within a semi-arid region and close to the Sub Tropical Ridge (STR) during winter, Stable Nocturnal Boundary Layer (SNBL) play a significant role in emission dispersion. These conditions violate the steady wind assumption required by the CSF method. This effect is examined in detail for orbit “09956” . Analysis of Moranbah AWS observations indicates that similar conditions occur for many other orbits, as meteorological regimes favourable for clear sky TROPOMI observations are also conducive to SNBL formation. Such conditions are evident in approximately 30 of the 68 orbits for which the CSF algorithm reported successful completion.
To evaluate the performance of the CSF method, we conducted detailed analyses of three representative cases: orbits 09956, 11332, and 09445.
Using the GitHub TROPOMI_CSF software, we generated emission estimates with multiple methods for orbit 09956 across a range of background concentrations from 1800 () to 1820 (). Comparison with the TM approach indicates that GitHub TROPOMI_CSF overestimates emissions by a factor of approximately 3-4, primarily due to box models applied to an inhomogeneous plume. For other orbits, the magnitude of this overestimation may differ, depending on the duration of SNBL conditions and the time lag between SNBL breakdown and the satellite overpass, (for example GHGSat-D which has local time at descending node 09:30AM McLinden, et al., 2024 ).
We also identified cases in which SNBL did not form under easterly wind conditions, as illustrated by orbit 11332. In this scenario, the TM and IME methods are not applicable. We have also identified some occasions where has not formed in easterly winds. This pattern has been illustrated for orbit 11332. In this case and methods are not applicable.
We also identified a case (orbit 09445) in which a westerly wind component allowed partial insight into background concentration values. However, this wind regime also resulted in extensive data gaps over the mountainous terrain east of the mine, substantially limiting the number of pixels available for emission rate estimation. Despite these limitations, this case demonstrates that background concentrations can be significantly higher than those inferred using the CSF approach and highlights the need for a properly configured Chemical Transport Model (CTM) to support robust background determination.
It is important to emphasise that the results presented in this report are highly location specific. They apply to a large, deep open cut mine situated in a semi-arid, subtropical valley environment. Under such conditions, the steady-wind assumption underpinning the CSF box model framework is frequently violated, limiting the method's applicability. However, the same algorithm may perform well under different settings, for example, for elevated point sources with tall stacks that penetrate above the nocturnal boundary layer, or in higher latitude regions where persistent light wind conditions associated with the subtropical ridge are less common.
Finally, it should be noted that the use of a rectangular downwind box to define plume shape is not well aligned with Gaussian dispersion theory and appears to have been adopted primarily for computational convenience rather than physical realism.