| 29 Jun 2026
Spatial resolution versus precision in CO2 point-source emission retrievals from the DQ-1 spaceborne lidar
Abstract. This study examines the trade-off between spatial resolution and measurement precision in satellite-based quantification of CO₂ point-source emissions using observations from DQ-1, the first spaceborne active CO₂ lidar mission. Allan deviation analysis is used to characterize scale-dependent random errors in XCO₂ retrievals over homogeneous surfaces, and the resulting error estimates are incorporated into Gaussian plume simulations to evaluate how spatial averaging affects emission retrieval under different emission strengths, wind speeds, and observation distances. The results show a nonlinear response: moderate averaging reduces random noise and improves retrieval stability, whereas excessive averaging degrades plume representation through loss of spatial resolution. The preferred averaging scale depends mainly on emission strength, transport distance, and local plume geometry. For strong idealized sources (2000 kg s⁻¹), 50–100 averaging points (3.5–7 km) generally provide the best compromise in favorable controlled simulations, with R² values up to 0.68. For weak sources (500 kg s⁻¹), single-overpass estimates remain close to the detection limit even after averaging (R² < 0.10). Application to seven DQ-1 overpasses of power plants shows that retrieved emissions are often more consistent with reference inventories at intermediate-to-coarse averaging scales, especially 75–150 points (5.25–10.5 km), but this range should not be interpreted as a universal optimum. The agreement should instead be treated as an inventory-based consistency check rather than an independent validation of instantaneous emissions. These findings provide a quantitative basis for choosing spatial averaging scales in DQ-1 point-source applications and identify the main conditions under which single-overpass lidar retrievals are likely to be informative.
General comments:
Active lidar remote sensing can complement the current GHG satellite observing system by contributing measurements at high latitudes, at night, and in between cloud gaps.
This technical paper analyzes the relationship between along-track averaging length of DQ-1 IPDA lidar observations and the CO2 emission retrieval skill for strong overflown point sources.
The authors conclude that the optimum averaging length depends on case-to-case methodical and geophysical conditions and therefore cannot be determined automatically.
The manuscript has a fair presentation quality yet lacks substance. Only few DQ-1 measurements are described, and many simulation results are trivial, such as small averaging length - too much noise - wrong CO2 emission retrieval, or, large averaging length - CO2 emission plume too coarse - retrieval fails. Similar but passive remote sensing studies can be found in the literature. I would still recommend its publication in AMT, after consideration of my comments below, because the present study is a first necessary step towards a future more comprehensive assessment of the highly innovative DQ-1 lidar mission.
line 55 (and line 493): the reference Amediek et al., 2017 is missing
Eq. 1: after Eq. A4 in Bovensmann et al, 2010, the Gaussian plume standard deviation is defined as a(x+x0)^0.894, instead of a(x/x0)^0.894, with unit m. Please correct.
line 116: ...and surface wind speed u (m/s).
line 118: ... is converted (instead of 130 converted)
line 144: better "time interval" instead of "averaging factors"
Figure 2: the different colors cannot be distinguished; include the tau^-1/2 relationship for pure white noise; add a km-axis for the averaging length.
section 3.1: explain which data you use to calculate your "coefficient of determination", R^2
Figure 11: following Eq. 1, increasing windspeed u decreases the plume enhancement. Therefore, the skill to detect the plume, expressed by your "coefficient of determination", R^2, green lines, should decrease as well. This is observed in Figure 11 at 2, 6 and 8 m/s, but not at 4 m/s. The R^2 values at 4 m/s are lower than at 6 m/s. Please explain. I think this is an error.
Figure 12b shows the same anomaly at 4 m/s. Please explain. I think this is an error.
Figure 13: was the wind direction parallel to the orbit track? To show the wind variability, please add the ERA5 wind vectors of the pressure level you used to estimate the emissions.
Table 2: after which criteria were these cases selected: wind direction parallel to the orbit track, other criteria?
Table 3: explain how you determined the averaging points.
line 410: can you comment when the DQ-1 data will become publicly available?