Large-scale monitoring of vegetation properties is crucial to understand water, carbon and energy cycles. The Normalized Difference Vegetation Index NDVI, computed from space-borne observations at visible and infrared wavelengths has been widely used since the 1980s to study vegetation changes and their implications on animal ecology , global fire emissions , deforestation and urban development , global patterns of land–atmosphere carbon fluxes and the vegetation response to climate and extreme events such as droughts . NDVI is sensitive to the abundance of chlorophyll and therefore to the photosynthetically active biomass (which includes herbaceous vegetation and the leaves of trees) but insensitive to wood mass. NDVI is thus not considered as an accurate proxy of total above-ground biomass (AGB), except in areas of low vegetation density . In contrast, being sensitive to both green and non-green vegetation components, passive microwave observations can provide important complementary information on the state and temporal changes of the vegetation features, in particular regarding the AGB dynamics .

The thermal emission arising from the Earth surface at microwave frequencies depends on the soil characteristics such as soil temperature, soil roughness and soil moisture content, which controls the soil emissivity . In the presence of vegetation, part of the soil emission is absorbed and scattered. These effects can be parameterized using radiative transfer models such as the τ-ω model , where τ is the optical depth and ω is the single-scattering albedo. τ was shown to be linked to the vegetation water content (VWC, kg m-2) and to other vegetation properties such as the leaf area index . Therefore, τ is commonly known as vegetation optical depth (VOD). VOD is also a function of the vegetation structure, which determines its dependence on the incidence angle and on the polarization of the radiation .

Passive microwave radiometry is therefore a promising tool for monitoring vegetation on a global scale. VOD samples the vegetation canopy, including woody vegetation, which uses root zone soil moisture . VOD was used to study deforestation in South America and Africa . Using VOD, it has been possible to reveal teleconnections linking the state of the vegetation in Australia and El Niño Southern Oscillation . In addition, showed the high potential of microwave VOD to monitor the AGB dynamics on a large scale. Using both VOD and NDVI contributes to a more robust assessment of the vegetation characteristics . The VOD has also been used to study the VWC and variations in ecosystem-scale isohydricity .

The above-mentioned studies used VOD derived from different radiometers operating at different frequencies : SSM/I at 19 GHz (K-band), TRMM-TMI at 10.7 GHz (X-band) and the Advanced Microwave Scanning Radiometer – Earth Observing System (AMSR-E) at 10.7 and 6.9 GHz (C-band). It is worth noting that VOD is intrinsically dependent on the frequency of the electromagnetic radiation and VODs retrieved at different frequencies provide complementary information. Therefore, in the following, a specific VOD data set will be noted as B-VOD, where B stands for the microwave band (X-VOD, C-VOD, etc.). The lower the frequency, the lower the VOD for a given level of VWC . Consequently, L-band (1.4 GHz, 21 cm) observations, which are less attenuated through the vegetation canopy, are capable of sampling the vegetation layer up to higher biomass values compared to higher-frequency observations.

Currently, two missions are performing systematic L-band passive microwave observations: the Soil Moisture and Ocean Salinity (SMOS) satellite , launched by ESA in November 2009, and the Soil Moisture Active Passive (SMAP) satellite , launched by NASA in January 2015. SMAP measures the brightness temperature for a single incidence angle in two polarizations. A single-angle dual polarization retrieval algorithm decreases the quality of the soil moisture retrievals but using a multi-orbit approach, assuming that the L-VOD does not vary significantly in a few days window, it is possible to estimate soil moisture and L-VOD . The full-polarization and multi-angular capabilities of SMOS allow the simultaneous retrieval of the soil moisture content and L-VOD. and compared SMOS L-VOD to X-VOD and C-VOD measured by AMSR-E and to visible/infrared vegetation indices. In crop zones, such as the MODIS vegetation indices, L-VOD increases during the growing season and decreases during senescence . On a global scale, L-VOD is less correlated to optical/visible vegetation indices than X/C-VOD, suggesting that L-VOD can add more complementary information with respect to optical/infrared indices than X/C-VOD . For instance, found a significant correlation between L-VOD and tree height estimates. also discussed this relationship and compared it to the one estimated with X/C-VOD, which shows higher values for low tree-height than SMOS L-VOD, as expected. also showed a close to linear relationship between L-VOD and AGB at 20 selected points over Peru, Columbia and Panama. L-VOD has been recently used to study the evolution of carbon stocks in African drylands by .

In summary, L-VOD derived from the new SMOS L-band observations is a promising tool for monitoring global vegetation characteristics. There is, however, a lack of in-depth studies on how L-VOD relates to established vegetation characteristics. The goal of the current study is to get further insight into the sensitivity of L-VOD to vegetation properties (such as tree height and AGB) and precipitation, which can drive the vegetation dynamics for some biomes. Taking into account the novelty of these observations, three distinct SMOS L-VOD data sets were evaluated against several data sets independent of L-VOD: (i) optical/infrared indices (representing the greenness of vegetation, also often used as a proxy for primary productivity), (ii) AGB benchmark maps, (iii) lidar-derived tree height and (iv) a precipitation data set. The area selected for this study is Africa, as it is a continent with several climate regions and biomes and with a large variability in the vegetation biomass, from sparse shrubs to savannah and very dense rainforests. In addition, have recently discussed the first biomass map of African savannahs computed from L-band active microwave (synthetic aperture radar) observations.

In contrast to passive measurements, for which the goal is to study how the thermal emission arising from the Earth is affected by the vegetation layer, active measurements allow us to study how the radiation emitted by a human-made radiation source is backscattered by the vegetation, which depends mainly on the vegetation water content and the vegetation structure.

Since this study is mainly devoted to AGB, long-term averages (typically annual) will be used. Studying the evolution of VWC would require using much shorter timescales. The document is organized as follows. Section  presents the different SMOS L-VOD data sets as well as the data sets used for the evaluation (tree height, cumulated precipitation, NDVI, EVI and four AGB data sets). Section  deals with the evaluation methods. Section  presents the results, which are discussed in Sect. , in particular the potential of L-VOD to estimate AGB on a large scale. Finally, Sect.  summarizes the results and presents the conclusions of this study.

The region selected for this study was the African continent because the Bouvet–Mermoz data set, which is the only one that has been produced using SAR observations made in the same frequency band (L-band) as SMOS, is limited to Africa. The African continent contains arid, equatorial and temperate regions with deserts, shrublands, mediterranean woodlands, grasslands, savannah and rainforests . Therefore, this study covers a wide range of climate regions and biomes and allows the analysis of L-VOD data to be extended to monitor vegetation properties, in particular biomass, on larger scales than in previous studies .

Unlike SMOS-IC and SMOS L3 products, which are produced natively on the 25 km EASE-Grid 2.0, the SMOS L2 L-VOD products are provided on the ISEA4H9 grid. A spatial interpolation was required to align the SMOS L2 L-VOD to the 25 km EASE-Grid. In order to maintain the meaning of the opacity as much as possible, e.g. close to the coastline or transitions between the two grid systems, this interpolated level 2 (hereafter iL2) L-VOD is obtained using (i) a DeLaunay triangulation linear interpolation whenever possible (three valid L2 L-VOD), (ii) a linear interpolation (only two valid L2 L-VOD) or (iii) the nearest L2 L-VOD (only one valid L2 L-VOD) to the 25 km EASE-Grid grid point within a neighbour defined by the 25 km EASE-Grid square cell.

AGB, precipitation, tree height and MODIS NDVI/EVI data were aggregated and resampled to the EASE-Grid 2.0, which are common to the SMOS L3 and IC data sets using the Geospatial Data Abstraction Library (GDAL) routine gdalwarp in average mode. Regarding, the SMOS level 2 data, several SMOS level 2 retrievals are available for a given day for high northern and southern latitudes. At these latitudes, the best retrievals (corresponding to lower values of the cost function Chi2) were selected.

In spite of observing in a protected band dedicated to research observations, some radio frequency interferences (RFI) from human-built equipment affect the quality of the SMOS observations. Several quality indicators are present in the SMOS L2 and L3 products. The DQX parameter uses the inverse linear tangent model (Jacobian) to translate the observation uncertainty (radiometric accuracy) into the parameter space uncertainty. The forward models are much more sensitive for lower values of the (SM, L-VOD) parameter space (leading to low DQX) than for higher values (leading to high DQX). Therefore, filtering to keep the lowest DQX implies a risk of biasing our results toward the lowest retrieved values, particularly for tropical forest where both SM and L-VOD are high. In addition, the DQX parameter does not give information about the correctness of the solution, which is based on a quality of a fit. Therefore, the Chi2 (goodness of the fit) was used to filter out the retrieved solutions. Several tests were done and a value of 3, corresponding approximately to the peak of the Chi2 probability distribution was found to be a good threshold. This is in agreement with the values used in other studies see for instance,.

In the case of SMOS-IC, data with a root mean squared difference between modelled and observed brightness temperatures larger than 10 K were filtered out. In addition, the L-VOD time series of the three products were analysed from grid point to grid point, and values with a deviation (in absolute value) larger than 2.5 with respect to the grid point average σ (where σ is the standard deviation) were considered as outliers and also filtered out.

The main evaluation strategy used in this study is to compare L-VOD data to the evaluation data sets presented in Sect. . These variables such as above-ground biomass, tree height or long-term averages of mean annual precipitation are not expected to change quickly over time. The biomass data sets discussed in Sect.  were produced with observations from years 1995 to 2010. The comparison of L-VOD with the other data sets was done using L-VOD data from 2011 and 2012, as 2011 is the first complete year after the SMOS commissioning phase, which ended in June 2010. The L-VOD data for 2011 and 2012 were averaged to avoid short-term variations due to changes in the vegetation water content over short time periods.

To get a quantitative assessment of the correlation and the dispersion of L-VOD versus the evaluation data sets, three correlation coefficients were computed. The Pearson correlation coefficient R is a measure of the linear correlation between two variables. If the relationship linking these variables is linear with no dispersion, R equals 1 (both variables increase together) or -1 (one variable increases when the other decreases). However, the relationships between L-VOD and the evaluation data are not expected to be linear in most of the cases. Therefore, the Spearman and Kendall rank correlations (which can range from -1 to 1) were also computed to quantify monotonic relationships, whether linear or not (the exact definition of the Spearman and Kendall rank correlations is given in the Supplement).

The AGB and L-VOD relationship was studied for different biomes using the IGBP land cover classes . Table S2 summarizes the IGBP classes, and Fig. S1 in the Supplement shows their spatial distribution using the Bouvet–Mermoz AGB map. For a single biome, a linear function gives a good fit to the AGB and L-VOD relationships (see Sect. ). In contrast, the global relationships linking the AGB data sets and L-VOD are significantly non-linear; therefore fits were computed following the approach used by . The L-VOD data were binned in 0.05-width bins. For each L-VOD bin, the 5th and 95th percentiles and the mean of the AGB distribution were computed, providing three AGB curves as a function of L-VOD. The three curves were fitted with the function used by : AGB=a×arctan⁡(b(vod-c))-arctan⁡(-bc)(arctan⁡(∞)-arctan⁡(-bc))+d, and with a logistic function, AGB=a1+e-b(VOD-c)+d. In Eqs. () and (), the parameters a,b,c and d are varied to get the best fit to the curves. The fitted curves give AGB in Mg h-1 units as a function of L-VOD, which is a dimensionless quantity. Therefore the units of a and d are Mg h-1 and b and c are dimensionless quantities.

Figure  shows the average L-VOD computed over 2011 and 2012 using both ascending and descending orbits for the three SMOS L-VOD products. In addition, it also shows the standard deviation (SD) and the number of points of the local time series after applying the filters discussed in Sect. 3. The three SMOS L-VOD products show a similar spatial distribution but the SMOS-IC L-VOD shows a smoother spatial distribution than the iL2 and L3 data sets. The highest values are found in equatorial forest regions and L-VOD decreases monotonically with distance to the equatorial forest in the tropical area and beyond. The SD of the L-VOD time series also increases towards the equatorial forest, in particular for the iL2 and L3 data sets. The number of points in the time series is lower for the IC data set due to the lower revisit frequency arising from the requirement of having brightness temperature measurements spanning an incidence angle range of at least 20∘ .

Figure  shows the evaluation data after resampling to a 25 km EASE-Grid 2.0: the 2011–2012 average of the MODIS NDVI and EVI indices, tree height, mean annual precipitation and AGB data sets. EVI and NDVI also decrease with increasing distance to the equator but more slowly than L-VOD. The tree height map shows two main populations: the equatorial forest, with heights larger than 20 m, and the rest of the continent, where most of the vegetation is lower than ∼5 m. In contrast to the previous quantities, AGB can vary by 2 orders of magnitude; therefore AGB maps are shown in logarithmic units in Fig. . The Baccini, Saatchi and Bouvet–Mermoz maps show a similar AGB distribution. In contrast, the Avitabile map shows a much sharper decrease in AGB from the equatorial forest region to the rest of the continent.

Three different SMOS-based L-VOD data sets were evaluated and compared to precipitation, tree height, NDVI, EVI and AGB data. Lower dispersion and smoother relationships were obtained by using SMOS-IC L-VOD compared to the iL2 and L3 L-VOD data sets. Consistently, the rank correlation values obtained with SMOS-IC were significantly higher by 5 %–15 % than those obtained with level 2 and level 3 L-VOD data sets.

The relationships between AGB estimates and L-VOD were strong (R=0.85–0.94) but differed among the products. For low vegetation classes (grasslands to woody savannah), the best performance was achieved with the Bouvet–Mermoz, Baccini and Saatchi biomass data sets. The biomass data by Baccini and Saatchi showed the best agreement with L-VOD for dense forest (R=0.70–0.79). Avitabile's AGB data showed low correlation values with L-VOD for low vegetation classes and a similar performance to Bouvet–Mermoz for dense forest (R=0.64–0.67). The AGB and L-VOD relationships can be fitted over the entire range of both variables with a single law using a sigmoid logistic function. However, an analysis per land cover class showed that within the same land cover class, the L-VOD and AGB relationship is close to linear. Therefore, the global non-linear relationship, found when all the different land cover are considered together, arises from different slopes in the L-VOD/AGB relationship obtained for different land cover classes considered separately. For low vegetation classes, the annual mean of L-VOD spans a range from 0 to 0.7 and could be related to the mean annual precipitation.

The relationship between AGB versus L-VOD was compared to the ones between AGB versus NDVI and AGB versus K/X/C-VOD from . As expected, NDVI saturates strongly and it becomes weakly sensitive to AGB changes from ∼70 to ∼300 Mg h-1. With respect to K/X/C-VOD, the AGB also increases slowly as VOD increases for most (∼70 %) of the K/X/C-VOD dynamic range but it saturates for VOD >0.8. In contrast, AGB values show a steady increment as L-VOD increases over the whole L-VOD dynamic range.

The equations computed in this study can be used to estimate AGB from SMOS-IC L-VOD. Of course, these equations depend on the data set used as reference to fit the AGB and L-VOD relationship. Three of them (those determined with , and Bouvet–Mermoz) gave very similar values when the 5th and 95th percentiles of the distributions were taken into account.

The results obtained in this study showed that the L-VOD parameter estimated from the SMOS passive microwave observations is an interesting index with which to monitor AGB at coarse resolution (∼40 km). Despite its coarse spatial resolution, the advantage of using SMOS L-VOD is that it is possible to compute one AGB map per year, for instance, which allows temporal estimations of the changes in the global carbon stocks on large scales .