Abstract. Carbonyl sulfide (COS) has the potential to be used as a climate diagnostic due to its close coupling to the biospheric uptake of CO2 and its role in the formation of stratospheric aerosol. The current understanding of the COS budget, however, lacks COS sources, which have previously been allocated to the tropical ocean. This paper presents a first attempt at global inverse modelling of COS within the 4-dimensional variational data-assimilation system of the TM5 chemistry transport model (TM5-4DVAR) and a comparison of the results with various COS observations. We focus on the global COS budget, including COS production from its precursors carbon disulfide (CS2) and dimethyl sulfide (DMS). To this end, we implemented COS uptake by soil and vegetation from an updated biosphere model (Simple Biosphere Model – SiB4). In the calculation of these fluxes, a fixed atmospheric mole fraction of 500 pmol mol−1 was assumed. We also used new inventories for anthropogenic and biomass burning emissions. The model framework is capable of closing the COS budget by optimizing for missing emissions using NOAA observations in the period 2000–2012. The addition of 432 Gg a−1 (as S equivalents) of COS is required to obtain a good fit with NOAA observations. This missing source shows few year-to-year variations but considerable seasonal variations. We found that the missing sources are likely located in the tropical regions, and an overestimated biospheric sink in the tropics cannot be ruled out due to missing observations in the tropical continental boundary layer. Moreover, high latitudes in the Northern Hemisphere require extra COS uptake or reduced emissions. HIPPO (HIAPER Pole-to-Pole Observations) aircraft observations, NOAA airborne profiles from an ongoing monitoring programme and several satellite data sources are used to evaluate the optimized model results. This evaluation indicates that COS mole fractions in the free troposphere remain underestimated after optimization. Assimilation of HIPPO observations slightly improves this model bias, which implies that additional observations are urgently required to constrain sources and sinks of COS. We finally find that the biosphere flux dependency on the surface COS mole fraction (which was not accounted for in this study) may substantially lower the fluxes of the SiB4 biosphere model over strong-uptake regions. Using COS mole fractions from our inversion, the prior biosphere flux reduces from 1053 to 851 Gg a−1, which is closer to 738 Gg a−1 as was found by Berry et al. (2013). In planned further studies we will implement this biosphere dependency and additionally assimilate satellite data with the aim of better separating the role of the oceans and the biosphere in the global COS budget.

Abstract. The uptake of carbonyl sulfide (COS) by terrestrial plants is linked to photosynthetic uptake of CO2 as these gases partly share the same uptake pathway. Applying COS as a photosynthesis tracer in models requires an accurate representation of biosphere COS fluxes, but these models have not been extensively evaluated against field observations of COS fluxes. In this paper, the COS flux as simulated by the Simple Biosphere Model, version 4 (SiB4), is updated with the latest mechanistic insights and evaluated with site observations from different biomes: one evergreen needleleaf forest, two deciduous broadleaf forests, three grasslands, and two crop fields spread over Europe and North America. We improved SiB4 in several ways to improve its representation of COS. To account for the effect of atmospheric COS mole fractions on COS biosphere uptake, we replaced the fixed atmospheric COS mole fraction boundary condition originally used in SiB4 with spatially and temporally varying COS mole fraction fields. Seasonal amplitudes of COS mole fractions are ∼50–200 ppt at the investigated sites with a minimum mole fraction in the late growing season. Incorporating seasonal variability into the model reduces COS uptake rates in the late growing season, allowing better agreement with observations. We also replaced the empirical soil COS uptake model in SiB4 with a mechanistic model that represents both uptake and production of COS in soils, which improves the match with observations over agricultural fields and fertilized grassland soils. The improved version of SiB4 was capable of simulating the diurnal and seasonal variation in COS fluxes in the boreal, temperate, and Mediterranean region. Nonetheless, the daytime vegetation COS flux is underestimated on average by 8±27 %, albeit with large variability across sites. On a global scale, our model modifications decreased the modeled COS terrestrial biosphere sink from 922 Gg S yr−1 in the original SiB4 to 753 Gg S yr−1 in the updated version. The largest decrease in fluxes was driven by lower atmospheric COS mole fractions over regions with high productivity, which highlights the importance of accounting for variations in atmospheric COS mole fractions. The change to a different soil model, on the other hand, had a relatively small effect on the global biosphere COS sink. The secondary role of the modeled soil component in the global COS budget supports the use of COS as a global photosynthesis tracer. A more accurate representation of COS uptake in SiB4 should allow for improved application of atmospheric COS as a tracer of local- to global-scale terrestrial photosynthesis.

We present a comparison of atmospheric transport model simulations for carbonyl sulfide (COS), within the framework of the ongoing atmospheric tracer transport model intercomparison project "TransCom". Seven atmospheric transport models participated in the inter-comparison experiment and provided simulations of COS mixing ratios in the troposphere over a 9-year period (2010-2018), using prescribed state-of-the-art surface fluxes for various components of the atmospheric COS budget: biospheric sink, oceanic source, sources from fire and industry. Since the biosphere is the largest sink of COS, we tested sink estimates produced by two different biosphere models. The main goals of TransCom-COS are (a) to investigate the impact of the transport uncertainty and emission distribution in simulating the spatio-temporal variability of COS mixing ratios in the troposphere, and (b) to assess the sensitivity of simulated tropospheric COS mixing ratios to the seasonal and diurnal variability of the COS biosphere fluxes. To this end, a control case with state-of-the-art seasonal fluxes of COS was constructed. Models were run with the same fluxes and without chemistry to isolate transport differences. Further, two COS flux scenarios were compared: one using a biosphere flux with a monthly time resolution and the other using a biosphere flux with a three-hourly time resolution. In addition, we investigated the sensitivity of the simulated concentrations to different biosphere fluxes and to indirect oceanic emissions through dimethylsulfide (DMS) and carbon disulfide (CS 2 ). The modelled COS mixing ratios were assessed against in-situ observations from surface stations and aircraft.

Abstract. The uptake of carbonyl sulfide (COS) by terrestrial plants is linked to photosynthetic uptake of CO2 by a shared diffusion pathway. Applying COS as a photosynthesis tracer in models requires an accurate representation of biosphere COS fluxes, but these models have not been extensively evaluated against field observations of COS fluxes. In this paper, the COS flux as simulated by the Simple Biosphere Model, version 4 (SiB4) is updated with the latest mechanistic insights and evaluated with site observations from different biomes: one evergreen needleleaf forest, two deciduous broadleaf forests, three grasslands, and two crop fields spread over Europe and North America. To account for the effect of atmospheric COS mole fractions on COS biosphere uptake, we replaced the fixed COS mole fraction originally used in SiB4 with spatially and temporally varying COS mole fraction fields. The lower COS mole fractions in the late growing season reduces COS uptake rates in agreement with observations. We also replaced the empirical soil COS uptake model in SiB4 with a mechanistic model that represents both uptake and production of COS in soils, which improves the match with observations over agricultural fields and fertilized grassland soils. SiB4 was capable of simulating the diurnal and seasonal variation of COS fluxes in the boreal, temperate and Mediterranean region. The daytime vegetation COS flux is on average 8 ± 27 % underestimated, albeit with large variability across sites. On a global scale, our model modifications caused a drop in the COS biosphere sink from 922 Gg S yr−1 in the original SiB4 model to 753 Gg S yr−1 in the updated version. The largest drop in fluxes was driven by lower atmospheric COS mole fractions over regions with high productivity, which highlights the importance of accounting for variations in atmospheric COS mole fractions. The change to a different soil model, on the other hand, had a relatively small effect on the global biosphere COS sink. The small role of the modeled soil component in the COS budget supports the use of COS as a global photosynthesis tracer.

The prediction of summer precipitation patterns (PPs) over eastern China is an important and topical issue in China. Predictors that are selected based on historical information may not be suitable for the future due to non-stationary relationships between summer precipitations and corresponding predictors, and might induce the instability of prediction models, especially in cases with few predictors. This study aims to investigate how to learn as much information as possible from various and numerous predictors reflecting different climate conditions. An objective prediction method based on the multinomial logistic regression (MLR) model is proposed to facilitate the study. The predictors are objectively selected from a machine learning perspective. The effectiveness of the objective prediction model is assessed by considering the influence of collinearity and number of predictors. The prediction accuracy is found to be comparable to traditionally estimated predictability, ranging between 0.6 and 0.7. The objective prediction model is capable of learning the intrinsic structure of the predictors, and is significantly superior to the prediction model with randomly-selected predictors and the single best predictor. A robust prediction can be generally obtained by learning information from plenty of predictors, although the most effective model may be constructed with fewer predictors through proper methods of predictor selection. In addition, the effectiveness of objective prediction is found to generally improve as observation increases, highlighting its potential for improvement during application as time passes.