At a global level, the retrieval of the above ground biomass (AGB) of woody vegetation has focused on using data from one or a few global observing satellite sensors, with consideration given (in some cases) to the integration of environmental information. The European Space Agency’s (ESA) Biomass Climate Change Initiative (CCI) is aiming to contribute to the global effort of improved AGB estimation by working towards the production of global AGB maps across several epochs (2007-2010, 2017/18 and 2018/19) and change maps to support climate science and modelling. The approach will focus primarily on the use of C- and L-band Synthetic Aperture Radar (SAR) data but will also integrate spaceborne optical and lidar data and a range of relevant environmental variables retrieved primarily from Earth Observation data (with these linking to the Essential Climate Variable (ECV) concept). These variables relate to a) the vertical and horizontal structure of plant components (i.e., foliage, branches, trunks and above ground roots; describing vegetation structure), floristics, seasonality and plant functional types (PFTs), b) topographic position (e.g., altitude, slope and aspect), c) at-ground surface states and structures (e.g., urban areas, bare surface materials, inundation (including tidal) state) and d) at surface and within volume environmental conditions (e.g., snow cover, vegetation moisture, freeze/thaw state). The latter two consider the conditions prevailing at or close to the time of the acquisition of the satellite sensor data as well as over longer time frames (e.g., water or snow hydroperiod). These EVs are retrieved using consistent and standardized units (e.g., m, percent) and are being integrated within the framework of the UN Food and Agriculture Organisation (FAO) Land Cover Classification System (LCCS). Validation and uncertainty assessment are also being established with these linking to other international initiatives (e.g., the ESA-NASA MAAP, CEOS) and datasets. This publication provides a review of the approach and considers other international projects that focus on improved global biomass and biomass change mapping. We also convey how CCI Biomass products can contribute to global climate science, including through linking with other CCI projects.

The terrestrial carbon fluxes attributed to sources and sinks are due a combination of land use and climate induced disturbance and recovery processes and the response of the ecosystems to climate. However, the general location of sinks and sources, the long-term trends and the interannual variability have been a source of debate among both modeling and observation communities. By focusing on the dynamics of aboveground vegetation carbon stocks as one of the main sources of uncertainty in the global carbon cycle, we developed annual estimates of the carbon density of aboveground live woody vegetation at about 10 km spatial resolution over the past 18 years (2000-2017). Using systematic sampling of vegetation structure measured by spaceborne LiDAR waveforms (2003-2008) to train a machine learning algorithm, we produced global maps of vegetation carbon density using a set of more than 50 spatial and temporal metrics of vegetation dynamics from a suite of optical and microwave satellite observations. The results suggest that the tropical regions have a general trend of increasing carbon sources with the largest contribution from South America and Southeast Asia. Tropical Africa, across the northern forest-savanna ecotones have been increasing in carbon sinks due to a reduction of seasonal fires. Temperate and boreal forests contribute significantly to the global sinks in the past decade due a combination of warming of the northern circumpolar region, reduction of fires and regeneration from forest management. The results show large uncertainty in quantifying the annual magnitude and long-term trends in vegetation carbon sinks, particularly in high biomass forests due to lack of sensitivity of the existing satellite data to forest structure and aboveground biomass. With the launch of future space missions such as NASA’s NISAR (2021) and ESA’s BIOMASS (2012) the uncertainty in estimating biomass and quantifying carbon sinks and sources of global vegetation will reduce significantly.

The current debate on the spatial distribution of carbon pools, their uncertainties and the quantification of sinks and sources has been recently backed by a global dataset of forest aboveground biomass (AGB) estimated from imagery acquired by a number of spaceborne remote sensing sensors. Envisat ASAR and ALOS PALSAR images together with Landsat 7 images and ICESAT GLAS observations from the year 2010 have been combined with an algorithm that exploits the signal component related to biomass so to overcome issues with a retrieval based on a single observation (e.g., errors due to specific environmental conditions). Assuming that SAR observables are mostly related to forest structure, the algorithm was designed to generate a global dataset of forest growing stock volume (GSV). AGB was then obtained from GSV by applying biomass conversion and expansion factors (BCEF) derived from a worldwide dataset of in situ observations of wood density and forest biomass components. The GSV and AGB datasets, referred to as GlobBiomass global biomass datasets, have been validated with data from forest field inventory plots and provincial estimates published by forest inventories. The representation of the biomass spatial distribution appears to be correct, with the largest pools of biomass being in the wet tropics, followed by temperate forests in Chile, the Pacific Northwest, southern Australia and Far East Russia. Nonetheless, the estimates present a slight tendency to overestimation in low biomass forests and a systematic underestimation in high biomass forest (> 300 Mg/ha). Also small-scale patterns of biomass distribution in the wet tropics appear to be smoothened and the biomass is partly affected by the geometric effects of terrain topography on SAR observations. For each pixel, an estimate of the standard error is provided, this being on average 50% at a spatial resolution of 100 m, mostly due to the limited sensitivity of the SAR backscatter to biomass. Characterization of the error covariances are currently ongoing.

Using the CCI Land Cover map of 2010 as reference for stratifying vegetation into different classes, we estimate the total aboveground biomass pool in forests to be 522 Pg (average: 107 Mg/ha) whereas forests and other wooded or naturally vegetated land stores 595 Pg of dry organic matter (average: 54 Mg/ha). The estimated aboveground biomass pool is 11% larger than the value reported by the FAO Forest Resources Assessment for 2010. Compared with pan-tropical estimates of forest AGB published in literature, we observe similar biomass levels in rain forest but almost half of the biomass in a latitudinal belt between 10° and 30° N, including moist tropical, dry tropical, sub-tropical forests with significant degradation. It appears that our findings are more consistent than estimates by other remote sensing datasets when compared against inventory data.

In this presentation, we shall present the methodology developed to obtain the global datasets of GSV and AGB, present and discuss strengths and weaknesses of the GlobBiomass datasets, and interpret the GlobBiomass datasets in a comparison with existing estimates of biomass based on inventory data and/or remote sensing. We shall also provide indications on the way forward towards the characterization of carbon sinks and sources, currently under investigation, for example, in the CCI Biomass project.

NASA’s Global Ecosystem Dynamics Investigation (GEDI) is scheduled for launch in late 2018 and will mount a multi-track, full-waveform lidar on the International Space Station (ISS) that is optimized for the measurement of forest canopy height and structure. GEDI will employ four high power beams and four coverage beams to produce near-global mapping of vertical canopy structure and aboveground biomass (AGB) across the worlds tropical and temperate ecosystems. The development of calibration equations that relate observed forest structure to biomass at a variety of spatial scales underpins this strategy. GEDI creates these calibrations by combining a large calibration and validation (cal/val) database of in situ measurements with coincident airborne laser scanning (ALS) observations that are used to simulate GEDI lidar waveforms. Assessment of GEDI performance is undertaken using an end-to-end simulator that links the GEDI cal/val database with simulations of GEDI waveforms and International Space Station (ISS) orbital tracks. This simulator allows us to evaluate the impact of instrument performance, changes in ISS orbits and processing algorithms, and losses of data that may occur due to clouds, snow, leaf-off conditions, and low signal-to-noise. By evaluating the consequences of operational decisions on GEDI science data products, this tool provides a quantitative framework for decision-making and mission planning.
In this presentation we present an assessment of the representivity of the GEDI cal/val database for global aboveground biomass mapping, and demonstrate use of the end-to-end simulator in the validation of GEDI aboveground biomass product algorithms and the global assessment of mission requirements. We will present first examples of on orbit waveforms and Level 1 and 2 data products from GEDI, and the status of GEDI post-launch biomass cal/val activities that target further validation of the science algorithms and assumptions, calibrating or updating the calibration of any necessary algorithm parameters, and evaluation of the science data products. Central to GEDI’s cal/val activities are an expanding network of in situ field observations, coordination with other mission campaigns and pre-launch cal/val collaborators, and participation in new ground and airborne campaigns including contemporaneous acquisition of GEDI orbital tracks using NASA’s Land, Vegetation, and Ice Sensor (LVIS) over the course of GEDI’s nominal two year mission.

The most detailed spatial assessment of the terrestrial carbon cycle for the forests and woodlands in Kenya to date was generated using in situ forest inventory data, in combination with multispectral Landsat 8 OLI surface reflectance, and Synthetic Aperture Radar (SAR) ALOS-2 PALSAR-2 backscatter intensity imagery.
Non-destructive in situ measurements were collected during the period 2013-2015 in the main woody biomes of the country. Pantropical allometric models (Chave et al., 2014) were used to estimate the aboveground biomass (AGB) pools considered in this study (i.e. trees, bamboos and lianas).
A non-parametric machine learning algorithm, Random Forests (Breiman, 2001), was calibrated within a k-fold framework to estimate AGB at 30 m spatial resolution for the whole of Kenya for the year 2014-2015. The accuracy assessment found coefficients of determination (R²) ranging from 0.64 to 0.90, and root mean square errors (RMSE) from 35.9 t/ha and 65.2 t/ha.
The results showed that the highest AGB densities are found in the dense tropical forests with values up to 530 t/ha (average 92 t/ha). However, the large area of wooded grassland and savannah in Kenya potentially stores around 58% of the total AGB in the country. National land cover information developed by the System for Land-based Emissions for Kenya (SLEEK), Copernicus Global Land Service 10-day composites of Leaf Area Index (LAI) (Dierckx et al., 2014, Sterckx et al., 2014), and soil organic matter data from the Harmonized World Soil Database (FAO, 2012), were then used together with the AGB map to constrain a model-data fusion framework to assess the terrestrial carbon cycle in Kenyan forests for the period 2014-2017.
Our analyses indicate that the median net ecosystems exchange (NEE) for Kenyan forest was 10.8 Tg C/y, and that Kenyan forest are very likely (confidence > 90%) to act as a net sink of atmospheric carbon.

Since the late 1970s, spaceborne microwave sensors have been providing measurements of radiation emitted by the Earth surface. From these measurements it is possible to derive vegetation optical depth (VOD), a model-based indicator that is related to both biomass and relative water content. Because of its high temporal resolution, VOD can be used to map dynamics in biomass and vegetation productivity. Studying long-term VOD dynamics and trends is generally hampered by the relatively short time span covered by the individual sensors. This can potentially be overcome by merging multiple VOD products into a single climate data record. However, combining multiple sensors into a single product is challenging because systematic differences between input products, e.g. biases, different temporal and spatial resolutions and coverage, need to be overcome.

Here, we present a new series of long-term VOD products, which combine multiple VOD datasets obtained by the Land Parameter retrieval Model from SSMI, TMI, AMSRE, Windsat, and AMSR2. We produced separate products for Ku-band (period 1987-2018), X-band (1997-2018), and C-band observations (2002-2018), thus preserving the unique characteristics of each frequency with respect to the structural elements of the canopy. The approach to merge the single-sensor VOD products closely follows that of the ESA CCI Soil Moisture products [1,2]: First, the datasets are co-calibrated via cumulative distribution function (CDF) matching using AMSR-E as scaling reference. Since we use only temporally overlapping observations, trends in the datasets are preserved. Using the Kolmogorow-Smirnow test we show that the bias correction performs well and is globally consistent. Second, we merge the datasets by taking the arithmetic mean between temporally overlapping observations of the scaled data. A first evaluation shows an increased spatio-temporal coverage of the datasets. Besides, using an autocorrelation analysis, we are able to show that the merged datasets have reduced noise compared to the input datasets.

Our presentation will give an overview of the methodology, including a robust assessment of the CDF-matching methodology, and the effect of averaging multiple sensors. To assess its potential for biomass studies, we will compare the VOD products to existing high-resolution, but static biomass maps. We will present a basic trend analysis on the resulting time series, compare them to trends in satellite-based leaf area index. We will also cover current problems and challenges, and possible solutions.



[1] Liu, Y.Y., Dorigo, W.A., Parinussa, R.M., de Jeu, R.A.M. , Wagner, W., McCabe, M.F., Evans, J.P., van Dijk, A.I.J.M. (2012). Trend-preserving blending of passive and active microwave soil moisture retrievals, Remote Sensing of Environment, 123, 280-297, doi: 10.1016/j.rse.2012.03.014.
[2] Dorigo, W., Wagner, W., Albergel, C., Albrecht, F., Balsamo, G., Brocca, L., Chung, D., Ertl, M., Forkel, M., Gruber, A., Haas, E., Hamer, P.D., Hirschi, M., Ikonen, J., de Jeu, R., Kidd, R., Lahoz, W., Liu, Y.Y., Miralles, D., Mistelbauer, T., Nicolai-Shaw, N., Parinussa, R., Pratola, C., Reimer, C., van der Schalie, R., Seneviratne, S.I., Smolander, T., & Lecomte, P. (2017). ESA CCI Soil Moisture for improved Earth system understanding: State-of-the art and future directions. Remote Sensing of Environment, 203, 185-215, doi: 10.1016/j.rse.2017.07.001