Abstract. This paper presents global land carbon fluxes for the period 1982–2010 (gross primary production, GPP, and net ecosystem exchange, NEE) estimated with the Max Planck Institute – Carbon Cycle Data Assimilation System (MPI-CCDAS v1). The primary aim of this work is to analyze the performance of the MPI-CCDAS when it is confronted with three different time periods for data assimilation (DA), and thereby to assess its prognostic capability. To this extend we assimilated nearly three decades (1982–2010) of space borne measurements of the fraction of absorbed photosynthetic active radiation (FAPAR) and atmospheric CO2 concentrations from the global network of flask and in situ measurements. Both data sets were incorporated with different assimilation windows covering the periods 1982–1990, 1990–2000 and 1982–2010. The assimilation results show a considerable improvement in the long-term trend and seasonality of FAPAR in the Northern Hemisphere, as well as in the long term trend and seasonal amplitude of the atmospheric CO2 concentrations when compared to the observations in sites globally distributed. After the assimilation, the global net land-atmosphere CO2 exchange (NEE) was −1.2 PgC yr−1, in agreement with independent estimates, while gross primary production (GPP; 92.5 PgC yr−1) was somewhat below the magnitude of independent estimates. The NEE in boreal eastern regions (Northeast Asia) increased on average by −0.13 PgC yr−1, which translated into an intensification of the carbon uptake in those regions by nearly 30 % than the contribution to the global annual average in the model before the assimilation. Our results demonstrate that using information only over a decade already yielded a large fraction of the overall model improvement, in particular for the simulation of phenological seasonality, its interannual variability (IAV) and long-term trend. Adding longer than decadal data did only lead to very moderate improvements in the long-term trend of the FAPAR simulated by the model, which may be attributed to the small model-data mismatch at the long timescales compared to the significantly larger observational signal and model-data mismatch error at seasonal cycle time scale. Decadal data also significantly improved the seasonality, IAV and long-term simulated trend in atmospheric CO2. Importantly, when running the MPI-CCDAS v1 with 30 years of data, the results remained in line with observations throughout this period, suggesting that the model can represent land uptake to a sufficient degree to make it compatible with the atmospheric CO2 record. Using data from 1982 to 1990 in the assimilation yielded only a difference to the observations of 2 ± 1.3 ppm for the period 15 to 19 years after the end of the assimilation. This suggests that despite imperfections in the representation of IAV, model-data fusion can increase the prognostic capacity of land carbon cycle models at relevant time-scales.

We describe the Max Planck Institute Carbon Cycle Data Assimilation System (MPI-CCDAS) built around the tangent-linear version of the JSBACH land-surface scheme, which is part of the MPI-Earth System Model v1. The simulated phenology and net land carbon balance were constrained by globally distributed observations of the fraction of absorbed photosynthetically active radiation (FAPAR, using the TIP-FAPAR product) and atmospheric CO2 at a global set of monitoring stations for the years 2005 to 2009. When constrained by FAPAR observations alone, the system successfully, and computationally efficiently, improved simulated growing-season average FAPAR, as well as its seasonality in the northern extra-tropics. When constrained by atmospheric CO2 observations alone, global net and gross carbon fluxes were improved, despite a tendency of the system to underestimate tropical productivity. Assimilating both data streams jointly allowed the MPI-CCDAS to match both observations (TIP-FAPAR and atmospheric CO2) equally well as the single data stream assimilation cases, thereby increasing the overall appropriateness of the simulated biosphere dynamics and underlying parameter values. Our study thus demonstrates the value of multiple-data-stream assimilation for the simulation of terrestrial biosphere dynamics. It further highlights the potential role of remote sensing data, here the TIP-FAPAR product, in stabilising the strongly underdetermined atmospheric inversion problem posed by atmospheric transport and CO2 observations alone. Notwithstanding these advances, the constraint of the observations on regional gross and net CO2 flux patterns on the MPI-CCDAS is limited through the coarse-scale parametrisation of the biosphere model. We expect improvement through a refined initialisation strategy and inclusion of further biosphere observations as constraints.

13 Data assimilation methods provide a rigorous statistical framework for constraining 14 the parametric uncertainty of land surface models (LSMs), with the aim of improving our 15 predictive capability as well as identifying areas in which the models need improvement. The 16 increase in the number of available datasets in recent years allows us to address different 17 aspects of the model at a variety of spatial and temporal scales. However, combining data 18 streams in a DA system is not a trivial task. In this study we highlight some of the challenges 19 surrounding multiple data stream assimilation for the carbon cycle component of LSMs. We 20 give particular consideration to the assumptions associated to the type of inversion algorithm 21 typically used for optimising the parameters of global LSMs – namely, Gaussian error 22 distributions and linearity in the model dynamics. We explore the effect of biases and 23 inconsistencies between the observations and the model (resulting in non Gaussian error 24 distributions), and we examine the difference between performing a simultaneous assimilation 25 (in which all data streams are included in one optimisation) and a step-wise approach (in 26 which each data stream is assimilated sequentially) in the presence of non-linear model 27 dynamics. In addition, we perform a preliminary investigation into the impact of correlated 28 errors between two data streams for two cases, both when the correlated observation errors are 29

Data assimilation methods provide a rigorous statistical framework for constraining parametric uncertainty in land surface models (LSMs), which in turn helps to improve their predictive capability and to identify areas in which the representation of physical processes is inadequate. The increase in the number of available datasets in recent years allows us to address different aspects of the model at a variety of spatial and temporal scales. However, combining data streams in a DA system is not a trivial task. In this study we highlight some of the challenges surrounding multiple data stream assimilation for the carbon cycle component of LSMs. We give particular consideration to the assumptions associated with the type of inversion algorithm that are typically used when optimising global LSMs – namely, Gaussian error distributions and linearity in the model dynamics. We explore the effect of biases and inconsistencies between the observations and the model (resulting in non-Gaussian error distributions), and we examine the difference between a simultaneous assimilation (in which all data streams are included in one optimisation) and a step-wise approach (in which each data stream is assimilated sequentially) in the presence of non-linear model dynamics. In addition, we perform a preliminary investigation into the impact of correlated errors between two data streams for two cases, both when the correlated observation errors are included in the prior observation error covariance matrix, and when the correlated errors are ignored. We demonstrate these challenges by assimilating synthetic observations into two simple models: the first a simplified version of the carbon cycle processes represented in many LSMs and the second a non-linear toy model. Finally, we provide some perspectives and advice to other land surface modellers wishing to use multiple data streams to constrain their model parameters.

The capacity of earth system models (ESMs) to make reliable projections of future atmospheric CO2 and climate is strongly dependent on the ability of the land surface model to adequately simulate the land carbon (C) cycle. Defining "adequate'' performance of the land model requires an understanding of the contributions of climate model and land model errors to the land C cycle. Here, a benchmarking framework is applied based on significant, observed characteristics of the land C cycle for the contemporary period, for which sufficient evaluation data are available, to test the ability of the JSBACH land surface component of the Max Planck Institute Earth System Model (MPI-ESM) to simulate land C trends. Particular attention is given to the role of potential effects caused by climate biases, and therefore investigation is made of the results of model configurations in which JSBACH is interactively "coupled'' to atmosphere and ocean components and of an "uncoupled'' configuration, where JSBACH is driven by reconstructed meteorology.The ability of JSBACH to simulate the observed phase of phenology and seasonal C fluxes is not strongly affected by climate biases. Contrarily, noticeable differences in the simulated gross primary productivity and land C stocks emerge between coupled and uncoupled configurations, leading to significant differences in the decadal terrestrial C balance and its sensitivity to climate. These differences are strongly controlled by climate biases of the MPI-ESM, in particular those affecting soil moisture. To effectively characterize model performance, the potential effects of climate biases on the land C dynamics need to be considered during the development and calibration of land surface models.

(1) Vienna University of Technology, Department of Geodesy and Geoinformation, Vienna, Austria (wouter.dorigo@tuwien.ac.at), (2) Geoville GmbH, Innsbruck, Austria, (3) Laboratoire des Sciences du Climat et de l’Environnement, CEA CNRS, France, (4) Climate Research Unit, University of East Anglia, UK, (5) FastOpt GmbH, Hamburg, Germany, (6) National Reasearch Council, Perugia, Italy, (7) Ghent University, Ghent, Belgium, (8) VU University, Amstredam, Netherlands, (9) Transmissivity BV, Netherlands, (10) Lund University, Sweden, (11) MPI for Biogeochemistry, Jena, Germany

Ozone concentrations in the Mediterranean area regularly exceed the maximum levels set by the EU Air Quality Directive, 2008/50/CE, a maximum 8-h mean of 120 μg·m-3, in the summer, with consequences for both human health and agriculture. There are a number of reasons for this: the particular geographical and meteorological conditions in the Mediterranean play a part, as do anthropogenic ozone precursor emissions from around the Mediterranean and continental Europe. Ozone concentrations measured on-board the Italian Research Council’s R. V. Urania during summer oceanographic campaigns between 2000 and 2010 regularly exceeded 60 ppb, even at night. The WRF/Chem (Weather Research and Forecasting (WRF) model coupled with Chemistry) model has been used to simulate tropospheric chemistry during the periods of the measurement campaigns, and then, the same simulations were repeated, excluding the contribution of maritime traffic in the Mediterranean to the anthropogenic emissions inventory. The differences in the model output suggest that, in large parts of the coastal zone of the Mediterranean, ship emissions contribute to 3 and 12 ppb to ground level daily average ozone concentrations. Near busy shipping lanes, up to 40 ppb differences in the hourly average ozone concentrations were found. It seems that ship emissions could be a significant factor in the exceedance of the EU directive on air quality in large areas of the Mediterranean Basin.