Plume pattern - Hail Creek 10^th August 2019 - Orbit 09445

Mean Sea Level Pressure

Mean sea level pressure analysis indicates a steady south-westerly flow over the Hail Creek mine, driven by the passage of a low pressure trough through the region. This flow transports air masses originating from Victoria and western New South Wales toward Hail Creek. During this period, the STR is positioned to the north of the site.

Moranbah AWS

During the 24 hours preceding the satellite overpass, winds remained between 240° and 270°, with speeds around 25 ($\frac{\mathrm{km}}{\mathrm{hour}}$) on the evening of 9^th August, decreasing overnight to 7.4 ($\frac{\mathrm{km}}{\mathrm{hour}}$) before increasing again into the mid-20s. Low dew point temperatures and relative humidity below 35% were observed. Owing to the arrival of a cooler south-westerly air mass, there is no evidence of UNKNOWN development in the observations.

TROPOMI

A one degree box around Hail Creek shows elevated concentrations in the southwest corner, likely reflecting emissions transported from New South Wales and Victoria. Increased ${\mathrm{CH}}_4$ concentrations are also visible east of Hail Creek. Numerous pixels are missing over the Dividing Range, located downwind of the mine. This limits the number of available pixels in the downwind box for the CSF method and renders both the TM and IME approaches inapplicable. See Figure 46

HYSPLIT

HYSPLIT forward modelling indicates that emissions travel across the Dividing Range to the northeast of Hail Creek. When these trajectories are overlaid on the TROPOMI image, only about three hours of emissions, those released at 2019-09-08 00Z, 01Z, and 02Z, are located over land. Beyond this point, the plume extends over the ocean, where the TROPOMI${\mathrm{CH}}_4$ retrieval algorithm does not operate. See Figure 47.

HYSPLIT backward modelling, extended to the TROPOMI overpass from the previous day (orbit 09431), indicates that background concentrations may be estimated by tracking the advected air parcels. See Figure 48.

The locations of the four background air parcels, situated in the lower left portion of the domain between 142°E and 143.5°E and later reaching Hail Creek to mix with emission puffs released between 2019-08-09 23Z and 2019-08-10 02Z, are provided in Table 24.

Using these coordinates, we delineate a box from [141.75°E, -25.25°S] to [143.5°E, -24.7°S] and compute an average background concentration of 1809 ($\mathrm{ppb}$) (see Figure 49). While this is useful for illustration, the choice of box size is inherently arbitrary without information on the turbulent diffusion tensor, and a more robust estimate would normally require a CTM. This approach is likely analogous to the algorithm used by Palmer, et al., 2021 .

It should be noted that substantial changes can occur to an air parcel over a 23 hour transport period. The parcel will undergo turbulent diffusion (approximated here through spatial averaging), may accumulate emissions from additional sources along its path, and may be diluted by expansion or by vertical mixing into layers with different concentrations. These factors require considerable caution when interpreting the results. Nevertheless, the approach outlined above provides a background concentration estimate with at least some physical basis and may serve as a useful foundation for more rigorous concentration modelling.

HYSPLIT modelling indicates that, in this case, it is feasible to determine the origin of the air mass into which the emissions are entrained. In principle, the previous day's TROPOMI image could be used to initialise an atmospheric dispersion model. However, determining background concentrations from backward trajectories should normally be carried out using a properly initialised chemical transport model CTM. The manual analysis presented here is intended only to illustrate the concept and to provide basic tools for interpreting the model output.

CSF

The CSF algorithm, as implemented in our software, completed successfully and produced an emission estimate of 29.768 ($\frac{t}{\mathrm{hour}}$) (260.769 ($\frac{\mathrm{Gg}}{\mathrm{year}}$) ). The wind speed was 7.854 ($\frac{m}{s}$), and the algorithm used an upwind background value of 1796 ($\mathrm{ppb}$). This may be compared with the 1809 ($\mathrm{ppb}$) background suggested by the HYSPLIT backward trajectory analysis, and with the domain median of 1806 ($\mathrm{ppb}$). The upwind box contains 24 valid pixels, while only 13 valid pixels fall within the downwind box. See Figure 50.

It should be noted that the CSF derived downwind box direction diverges from that indicated by the HYSPLIT modelling, likely due to the large number of missing pixels, as the CSF algorithm relies solely on the available observations.

Analysis of the positive pixels (i.e., the inferred plume shape) shows that only nine pixels exceed the background value of 1796 ($\mathrm{ppb}$). See Figure 51.

Inspection of the transects shows estimated emission rates ranging from 28.5 to 30.5 ($\frac{t}{\mathrm{hour}}$), with only transects 4, 5, and 15 meeting the validity criteria; the remaining transects are rejected due to an insufficient number of pixels. See Figure 52.

TM

TM cannot be applied in this case due to the large number of missing pixels.

IME

IME cannot be applied in this case due to the large number of missing pixels.

Summary

This case study was selected because meteorological analysis indicates persistent south-westerly winds during the 24 hours preceding the TROPOMI overpass. Due to the mountain range east of the mine, this wind direction results in good pixel coverage upwind but poor pixel coverage downwind of the source.

HYSPLIT backward modelling enabled us to narrow down the background air mass location 24 hours earlier (orbit 9431). Averaging concentrations within a box surrounding the relevant air parcel locations yields a background value of 1809 ($\mathrm{ppb}$).

Although the CSF algorithm reports a successful calculation for this orbit, the results are highly uncertain due to the very limited number of pixels intersecting the transects. The upwind box has good pixel coverage, yielding a background concentration of 1796 ($\mathrm{ppb}$), about 13 ($\mathrm{ppb}$) lower than the estimate derived from the HYSPLIT based background analysis.

Both the TM and IME methods are not applicable in this case due to limited pixel coverage over the mountainous downwind region and the presence of steady wind crossing the box boundary.