Development and application of the WRFDA-Chem three-dimensional variational (3DVAR) system: aiming to improve air quality forecasting and diagnose model deficiencies

• Main Authors: W. Sun, Z. Liu, D. Chen, P. Zhao, M. Chen
• Format: Article
• Language: English
• Published: Copernicus Publications 2020-08-01
• Series: Atmospheric Chemistry and Physics
• Online Access: https://acp.copernicus.org/articles/20/9311/2020/acp-20-9311-2020.pdf

<p>To improve the operational air quality forecasting over China, a new aerosol or gas-phase pollutants assimilation capability is developed within the WRFDA system using the three-dimensional variational (3DVAR) algorithm. In this first application, the interface for the MOSAIC (Model for Simulating Aerosol Interactions and Chemistry) aerosol scheme is built with the potential for flexible extension. Based on the new WRFDA-Chem system, five experiments assimilating different surface observations, including PM<span class="inline-formula">_2.5</span>, PM<span class="inline-formula">₁₀</span>, <span class="inline-formula">SO₂</span>, <span class="inline-formula">NO₂</span>, <span class="inline-formula">O₃</span>, and CO, are conducted for January 2017 along with a control experiment without data assimilation (DA). Results show that the WRFDA-Chem system evidently improves the air quality forecasting. From the analysis aspect, the assimilation of surface observations reduces the bias and RMSE in the initial condition (IC) remarkably; from the forecast aspect, better forecast performances are acquired up to 24&thinsp;h, in which the experiment assimilating the six pollutants simultaneously displays the best forecast skill overall. With respect to the impact of the DA cycling frequency, the responses toward IC updating are found to be different among the pollutants. For PM<span class="inline-formula">_2.5</span>, PM<span class="inline-formula">₁₀</span>, <span class="inline-formula">SO₂</span>, and CO, the forecast skills increase with the DA frequency. For <span class="inline-formula">O₃</span>, although improvements are acquired at the 6&thinsp;h cycling frequency, the advantage of more frequent DA could be consumed by the disadvantages of the unbalanced photochemistry (due to inaccurate precursor <span class="inline-formula">NO_<i>x</i></span>&thinsp;<span class="inline-formula">∕</span>&thinsp;VOC (volatile organic compound) ratios) or the changed titration process (due to changed <span class="inline-formula">NO₂</span> concentrations but not NO) from assimilating the existing observations (only <span class="inline-formula">O₃</span> and <span class="inline-formula">NO₂</span>, but no VOC and NO). As yet the finding is based on the 00:00&thinsp;UTC forecast for this winter season only, and <span class="inline-formula">O₃</span> has strong diurnal and seasonal variations. More experiments should be conducted to draw further conclusions. In addition, considering one aspect (IC) in the model is corrected by DA, the deficiencies of other aspects (e.g., chemical reactions) could be more evident. This study explores the model deficiencies by investigating the effects of assimilating gaseous precursors on the forecast of related aerosols. Results show that the parameterization (uptake coefficients) in the newly added sulfate–nitrate–ammonium (SNA)-relevant heterogeneous reactions in the model is not fully appropriate although it best simulates observed SNA aerosols without DA; since the uptake coefficients were originally tuned under the inaccurate gaseous precursor scenarios without DA, the biases from the two aspects (SNA reactions and IC DA) were just compensated. In future chemistry development, parameterizations (such as uptake coefficients) for different gaseous precursor scenarios should be adjusted and verified with the help of the DA technique. According to these results, DA ameliorates certain aspects by using observations as constraints and thus provides an opportunity to identify and diagnose the model deficiencies; it is useful especially when the uncertainties of various aspects are mixed up and the reaction paths are not clearly revealed. In the future, besides being used to improve the forecast through updating IC, DA could be treated as another approach to explore necessary developments in the model.</p>