Bias Correction of Long-Path CO2 Observations in a Complex Urban Environment for Carbon Cycle Model Inter-Comparison

The GreenLITE™ trace gas measurement system, jointly developed by Atmospheric and Environmental Research, Inc. and Spectral Sensor Solutions LLC, provides high-precision, long-path measurements of atmospheric trace gases CO₂ and CH₄ over extended (0.04 km² – 25 km²) areas of interest. In 2015, a prototype was deployed in Paris, France, to demonstrate its ability to provide continuous measurements of CO₂ and two-dimensional maps of time-varying concentrations over a complex urban environment. Subsequently, these data have been bias corrected to create a physically consistent set of mean concentration values for inter-comparisons with: 1) Highly accurate collocated in situ point measurements obtained within the Greater Paris area, and 2) Results from high-spatial and temporal resolution regional carbon cycle model data. To achieve these ends, the GreenLITE™ data were corrected using existing point measurements to reconcile constant systematic as well as slowly varying temporal differences that exist between in situ and GreenLITE™ measurements. These efforts provide a method for comparisons and the potential for long-term co-assimilation of both measurements into urban-scale emission models. While both the constant systematic biases and the slowly varying differences may have different impacts on the measurement accuracy and/or precision, they are in part due to several potential common terms that include limitation in the instrument design, uncertainties in spectroscopy, and imprecise knowledge of the atmospheric state. This work provides a brief overview of the system design and the current gas concentration retrieval approach, a description of the bias correction approach, and the results as applied to data collected in Paris, France.

This work provides an overview of the method for correcting systematic biases in new and novel long-path estimates of GHG concentrations over extended and complex regional domains by co-registering them with precise yet highly localized nearby in situ measurements of GHG concentrations. While this does not directly address the absolute accuracy of these long-path remote sensing measurements, it does provide a well-defined mechanism for minimizing biases between the long-path measurements and precise in situ measurements that vary slowly in time and are due to a number of sub-factors. This work also demonstrates that the defined approach may provide additional mechanisms for minimizing spectroscopic mismatches between observed and modeled long-path differential absorption spectrometer data. In the case illustrated above, retrieved values of integrated GHG concentration based on either HITRAN 2012 and 2016 produce similar to corrected results over a broad range of environmental conditions. While both implementations produce similar corrected results, the required correction factors were significantly different in magnitude, and the difference between the two correction factors remained nearly constant over the period of observation. This constant difference in conjunction with continuous measurements or locking of the on- and off-line wavelengths may provide a metric to assess different RT parameterizations and retrieval approaches.