During photosynthesis, plants open the stomata and let water escape through transpiration. Globally, transpiration is a crucial hydrological variable that can regulate precipitation and temperature. While transpiration is elusive to measure over large scales, its synchronisation to photosynthesis leaves room for hope, as the current generation of satellites can already observe a by-product of photosynthesis, i.e. solar-induced chlorophyll fluorescence (SIF). Here we explore the potential of SIF observations to infer transpiration at the global scale and without complex parameterisations.

We first evaluate globally-distributed eddy-covariance measurements of latent heat flux (λE) and satellite SIF retrievals. We show that, at monthly time scales, correlations between SIF and λE are higher than between SIF and in situ gross primary production (GPP). Through the use of an advanced radiative transfer model, we demonstrate a clear connection between SIF and λE. Whereas the relationship SIF–GPP is mostly controlled by plant physiology, the SIF–λE relationship is largely determined by atmospheric conditions, i.e. net radiation, air temperature and vapour pressure deficit.

These insights are used to develop a novel diagnostic of evaporative fraction, i.e. the ratio of λE to available energy. This diagnostic is based on observations of SIF and satellite-based photosynthetically active radiation (PAR). Estimates are validated and compared against the simulated evaporative fraction by the eartH2Observe land surface models. Despite its relatively coarse resolution, the SIF-based diagnostic is highly correlated against in situ-measured evaporative fractions, particularly during the growing season, outperforming the majority of land surface models. This diagnostic successfully demonstrates the ability to capture the effect of environmental stress on ecosystem transpiration, and adequately reflect the timing of this stress.

Overall, our results show the potential of SIF to constrain the hydrology in global climate models, and suggest that future advances in satellite SIF retrievals will allow for a more direct estimation of transpiration over large scales.

The importance of improving the coupling of water and carbon cycles in land surface models has been emphasised in recent studies (Humphrey et al., 2018; Li et al., 2018). Accurate quantification of carbon and water balances requires reliable methods to estimate gross primary production (GPP) and evapotranspiration (ET) across space and time (Wagle, Gowda, Xiao, & KC, 2016). Light use efficiency (LUE) and water use efficiency (WUE) based models are widely recognised as effective methods for estimating GPP and ET (Joiner et al., 2018).

The P model (for ‘production’) is grounded in plant functional ecology and links the carbon and water cycles via a theory of stomatal optimization and photosynthetic acclimation. It has the mathematical form of a LUE model while being traceable to first principles, including the standard model of photosynthesis, for the prediction of GPP (H Wang, Prentice, & Davis, 2014; Han Wang et al., 2017). The model is being applied in the TerrA-P project to develop a prototype of a new 300 m resolution satellite product for the monitoring of 10-daily GPP and annual above-ground biomass production (ABP) at a global scale (Prentice & Thomas, 2018; Swinnen et al., 2018). The model as implemented in TerrA-P has only three free parameters, of which two are estimated from independent observations on leaf stable carbon isotopes and leaf-level electron transport capacity. As input the model requires elevation, CO2 concentration, solar radiation, vapour pressure deficit (VPD) and temperature. ECMWF meteorological forcing data and AASTR (Advanced Along-Track Scanning Radiometer) land surface temperatures have been used as alternative sources of temperature data, both to drive photosynthesis and for the calculation of VPD. A merged data set from SeaWiFS (Sea-Viewing Wide Field-of-View Sensor) and MERIS (MEdium-spectral Resolution, Imaging Spectrometer) Global Vegetation Index (GVI) was employed to calculate the fraction of absorbed photosynthetically active radiation (fAPAR) (Swinnen et al., 2018). Validation of TerrA-P results with FLUXNET observations at a selection of high-quality sites showed that seasonal cycles and interannual variability of GPP could be reproduced well in nearly all ecosystems, despite the absence of any biome-specific parameters in the model (Balzarolo et al., 2018; Swinnen et al., 2018).

We will present a demonstration application of the P model using a novel approach that extends the algorithm to create a prototype of a universal transpiration (T) product. Both GPP and T outputs will be evaluated against FLUXNET observations. Stomatal conductance will be calculated based on the model’s predictions of GPP and the ratio of internal to external CO2 partial pressure, allowing transpiration to be calculated from VPD.

High levels of uncertainty continue to afflict the partitioning of latent heat fluxes into evaporation, transpiration and interception components (Talsma et al., 2018). In the initial phase of this research, transpiration will be calculated using a constant partitioning between stomatal and non-stomatal fluxes. The results will be benchmarked against existing products for both T and GPP.

The P model has many advantages, including its strong theoretical and empirical basis, extremely parameter-sparse nature, and the fact that it does not require the prior specification of plant functional types or land cover types. The research presented here will extend its application from primary production monitoring to include carbon-water cycle coupling and water resources assessment.




References

Balzarolo, M., Janssens, I., Vicca, S., Prentice, I. C., Thomas, R., Swinnen, E., & Van Hoolst, R. (2018). TerrA-P: Development and validation of a global GPP/NPP model using MERIS and Sentinel-3 data Validation report.

Humphrey, V., Zscheischler, J., Ciais, P., Gudmundsson, L., Sitch, S., & Seneviratne, S. I. (2018). Sensitivity of atmospheric CO2 growth rate to observed changes in terrestrial water storage. Nature, 560(7720), 628–631. https://doi.org/10.1038/s41586-018-0424-4

Joiner, J., Yoshida, Y., Zhang, Y., Duveiller, G., Jung, M., Lyapustin, A., … Tucker, C. J. (2018). Estimation of Terrestrial Global Gross Primary Production (GPP) with Satellite Data-Driven Models and Eddy Covariance Flux Data. Remote Sensing, 10(9), 1346. https://doi.org/10.3390/rs10091346

Li, L., Wang, Y., Arora, V. K., Eamus, D., Shi, H., Li, J., … Yu, Q. (2018). Evaluating Global Land Surface Models in CMIP5: Analysis of Ecosystem Water- and Light-Use Efficiencies and Rainfall Partitioning. Journal of Climate, 31(8), 2995–3008. https://doi.org/10.1175/JCLI-D-16-0177.1

Prentice, I. C., & Thomas, R. (2018). Development and validation of a global GPP/NPP model using MERIS and Sentinel-3 data (TerrA-P) ATBD v2. Mol, Belgium.

Swinnen, E., Van Hoolst, R., Prentice, I. C., Janssens, I., Vicca, S., & Balzarolo, M. (2018). TerrA-P: Development and validation of a global GPP/NPP model using MERIS and Sentinel-3 data Requirement Baseline Document. Mol, Belgium.
Talsma, C. J., Good, S. P., Jimenez, C., Martens, B., Fisher, J. B.,

Miralles, D. G., … Purdy, A. J. (2018). Partitioning of evapotranspiration in remote sensing-based models. Agricultural and Forest Meteorology, 260–261, 131–143. https://doi.org/10.1016/j.agrformet.2018.05.010

Wagle, P., Gowda, P. H., Xiao, X., & KC, A. (2016). Parameterizing ecosystem light use efficiency and water use efficiency to estimate maize gross primary production and evapotranspiration using MODIS EVI. Agricultural and Forest Meteorology, 222, 87–97. https://doi.org/10.1016/j.agrformet.2016.03.009

Wang, H., Prentice, I. C., & Davis, T. W. (2014). Biophysical constraints on gross primary production by the terrestrial biosphere. Biogeosciences, 11, 5987–6001. https://doi.org/10.5194/bg-11-5987-2014

Wang, H., Prentice, I. C., Keenan, T. F., Davis, T. W., Wright, I. J., Cornwell, W. K., … Peng, C. (2017). Towards a universal model for carbon dioxide uptake by plants. Nature Plants. https://doi.org/10.1038/s41477-017-0006-8

In times of climate change it is important to deeply understand how the biosphere responds to climate drivers. However, these interactions are (i) not happening instantaneously, and (ii) may differ depending on the time scale under scrutiny. For instance, at the seasonal scale we may expect a different sensitivity of vegetation to temperature than on monthly or inter-annual scale. However, so far there is no systematic assessment of this phenomenon reported at the global scale. Here we explore long-term (>30 year) remote sensing records and climate data to characterize atmosphere-biosphere interactions explicitly on different time scales. We use time series decomposition to explore sub-monthly oscillations to decadal trends.

Our results show that interactions between a vegetation index (NDVI), air temperature and precipitation strongly differ across time scales. Importantly, they often strongly diverge from the relationship of the integrated signal. This suggests that processes need to be characterized specifically on the temporal scale under investigation. Furthermore, we correlate variables on a global map for each time scale and observe complex spatial patterns in ecosystem variability not resolved with conventional methods. For example, our results indicate that the tropics constitute a highly heterogeneous system in terms of atmosphere-biosphere interactions on different temporal scales.

Both instantaneous and delayed interactions between atmosphere and biosphere are important in understanding and predicting ecosystem trajectories under changing climatic conditions. Our results suggest that process parameters and sensitivities can vary strongly across time scales. This could help explain heterogeneity of ecosystem responses on different scales. These dynamics should be taken into account for accurate long- and short-term climate modeling.

Studying global spatio-temporal patterns and footprints of key climate essential variables (ECVs) is of paramount relevance to understand climate dynamics and variability. It is widely recognized that the inter-annual ocean-atmosphere variability is widely induced by the El Niño Southern Oscillation (ENSO), which is the dominant source of climate extremes in hydrological and ecological systems. During the positive phase of ENSO (El Niño), atmospherical drivers change at tropical latitudes, inducing an anomalous dry/wet pattern and lagged extra-tropical patterns. Water availability is generally considered a dominant variable in the terrestrial carbon cycle too, being one of the principal inter-annual drivers. Consequently, ENSO induces part of the inter-annual variability of water availability in the terrestrial carbon cycle.
In this work, we study the global teleconnection patterns between ENSO and the water content in soils and vegetation, as represented by SMOS-based retrievals of soil moisture (SM) and vegetation optical depth (VOD). A novel nonlinear dimensionality reduction method, the rotated complex principal component analysis (ROCK-PCA), is applied to extract the inter-annual variability of SM and VOD through a nonlinear spatio-temporal decomposition, and obtain independent spatial and temporal components representing the global long trend variability of these variables during the study period (2010 to 2017). We compare the extracted temporal patterns from SM/VOD with the ENSO4 index using cross-correlation (see Fig. 1), and then spatialize it obtaining the maps of ENSO influence on the two variables (Fig. 2a-b), as well as global maps of lagged impacts. The objective of this work is to revisit the well-known ENSO global teleconnections patterns and quantify its impact over soils and vegetation using remote sensing data. Deriving spatially explicit and temporally resolved links of interaction of key variables using remote sensing data is crucial to unravel the complex Earth system relations, while the characterization of the memory effects may have implications in the study of future extreme events as drought and carbon losses.
Fig. 1: ENSO4 index and the inter-annual components extracted from VOD and SM time series.
Fig. 2: Correlation maps for different temporal lags between ENSO4 index and SM inter-annual component. a) 5 days lagged correlation map, b) 18 months lagged correlation map.

As a result of global warming, major changes are taking place in the processes of our climate system, including the cycles of energy, water, and biogeochemicals. A large portion of the variability in the water cycle and, hence vegetation dynamics, is controlled by oceanic-atmospheric oscillation patterns. However, the connection between ocean-atmospheric oscillations and hydrology is still poorly understood, due to the uncertainties in the observations and the co-varying behavior of multiple oscillations. A better understanding of these connections is needed to provide better predictions about the evolution of climate and hydrology.
In our study we make use of a data-driven method, called LASSO (least absolute shrinkage and selection operator) regression, to quantify the impact of a multitude of climate oscillation indices, used to describe the state of the atmospheric-oceanic circulations, and hydrological response variables (e.g. precipitation) over land. This supervised learning approach uses automatic feature selection and regularization to quantify the impact of each individual oscillation on the hydrological variable. Additionally, it allows to account for cross-correlations in the drivers, i.e. the climate oscillation indices, which leads to an improvement in the quality of the model compared to simple correlations. Furthermore, time lags ranging between zero and five months are introduced in every feature for considering the not necessarily instantaneous impact of ocean-atmospheric oscillations on hydrology. Besides using the entire available timespan at once, we also perform the analysis for each season separately. Furthermore, a nested cross-validation for validating the predictive LASSO regression models and a significance test using Benjamini-Hochberg procedure is applied.
Our results show that in some regions, up to 40% of the variability in the hydrological variables can be explained by ocean-atmospheric oscillations. Especially for the winter season model, which is trained using only data of December, January, and February, we reveal a particularly high explained variance for the precipitation signal. Analyzing the influence of each climate oscillation index shows that the signal of the Northern Annular Mode, East Atlantic Oscillation, East Atlantic West Russia Pattern, and North Atlantic Oscillation have a significant impact in the western Mediterranean area. Through these results, we gain new insights and improve the general understanding of the connection between ocean-atmospheric oscillations and hydrology.

This study investigates the potential of LDAS-Monde offline land data assimilation system to monitor the impact of the summer 2018 heatwave over Europe on vegetation. LDAS-Monde forced by ECMWF’s (i) latest reanalysis, ERA-5, and by (ii) its operational high resolution analysis (IFS, Integrated Forecasting System, HRES) are used in conjunction with the assimilation of Copernicus Global Land Service (CGLS) satellite derived observations: Surface Soil Moisture (SSM) and Leaf Area Index (LAI). It leads to reanalyses of the Land Surface Variables (LSVs) at 0.25°x0.25° (over 2008-07/2018) and 0.10°x0.10° (over 04/2016-09/2018) spatial resolution, respectively. Analysis of long time series of satellite derived CGLS LAI (2000-09/2018) and SSM (2008-09/2018) permits to highlight strong negative anomalies for July 2018 affecting large areas of North Western Europe and reflecting the heatwave impact. Such large anomalies spreading over large part of the considered domain have never been observed in the LAI product in this 18-yr period. LDAS-Monde forced by either ERA5 or HRES leads to a good a representation of the vegetation state in general and for this specific event, with HRES configuration exhibiting better monitoring skills than ERA-5 configuration. The consistency of model runs forced by ERA5 and IFS HRES over a common period of time (April 2016 to July 2018) is explored leading to a spatial resolution vs system study. In order to understand if the improvement from the use of ERA-5 to HRES was due to the resolution only (e.g. better representation of the land cover, orography) or the forcing quality, we have evaluated two others LDAS experiments down-scaling(up-scaling) ERA-5(HRES) to HRES(ERA-5) spatial resolutions. Results suggest that resolution is the key but that system still adds some skills. If there is a surface physiography and modeling advantage of HRES configuration, there is added value in down-scaling ERA5 too. If the improvement from LDAS-Monde analysis on control variables (soil moisture from layers 2 to 8 of the model representing the first meter of soil and LAI) from the assimilation of SSM and LAI was expected, other model variables benefit from the assimilation through biophysical processes and feedbacks in the model. Finally, it is shown that a 10-day forecast of LAI initialized by LDAS-Monde analysis has more skills than the model alone.