Microwave observations are sensitive to plant water content and could therefore provide essential information on biomass and plant water status in ecological and agricultural applications. The combined data record of the C-band scatterometers on the European Remote-Sensing Satellites (ERS)-
Microwave remote sensing observations are sensitive to plant water content, which depends on aboveground biomass and plant water status
The current study is motivated by the availability of consistent C-band data from 1991 to at least 2030, and its potential value as a long-term data record for vegetation monitoring. ASCAT is a real aperture radar operating at 5.255 GHz with co-polarization (VV). There are currently three ASCAT instruments in orbit on Metop (Meteorological Operational satellite)-A, Metop-B, and Metop-C, launched in October 2006, September 2012, and November 2018, respectively. ASCAT builds on the success of the European Scatterometer (ESCAT) which flew on the European Remote-Sensing Satellites (ERS)-
Many early studies demonstrated the sensitivity of ESCAT and ASCAT backscatter to vegetation and explored the potential value of these data for vegetation monitoring
The ASCAT Dynamic Vegetation Parameters refer to the parameters of the second-order Taylor polynomial used to describe the incidence angle (
Results from
Study area. The map is colored by biome, and nine ecoregions of interest are highlighted based on the dataset of
The goal of this study is to improve our understanding of the ASCAT backscatter–incidence angle relationship and how this relationship might be used to monitor vegetation water dynamics. The Amazon basin and its surroundings has been chosen as a study area, as it provides a wide range in terms of expected variability in ASCAT backscatter, slope, and curvature. Backscatter in the evergreen forest was considered to be so stable that this region has been used for satellite radar calibration
Figure The Napo moist forests (fNW), located in northwestern Amazonia, receive some of the highest amounts of annual precipitation in the biome, reaching up to 4000 mm in some parts. This highly biodiverse region has canopies reaching 40 m. The Guianan moist forests (fNE) are one of the largest continuous stretches of relatively pristine lowland tropical rainforest in the world. There are two distinct wet seasons, i.e., from December to January and from May to August. The floral diversity is rich, with multi-tiered vegetation of 40 m tall trees with herbaceous plants below. The dry season (September–November) can see a substantial reduction in leaves, although the forest is evergreen. The Southwest Amazon moist forests (fSW) have significant variations in topography and soil characteristics, leading to extremely high biodiversity. The size and orientation of the ecoregion means that climatic conditions vary markedly, with the north being wetter and having less seasonal variability compared to the south. The inaccessibility of the region has aided in its conservation. The Madeira–Tapajós moist forests (fSE) are transected by the Trans-Amazonian Highway and have high levels of urbanization and deforestation. There are characteristic liana (woody vine) forests with a lower ( The Juruá–Purus moist forests (fC) are largely intact forests in the low Amazon Basin. The canopy can reach up to 30 m, with some patches of open canopy. The seasonally flooded forest, Marajó várzea (ff), is located at the mouth of the Amazon River. The vegetation is dominated by palms and shorter than surrounding forests. It has areas with tidal flows from the Atlantic Ocean and seasonally and permanently inundated forests. The annual seasonal flooding occurs during the peak precipitation period between January and May
The following three savanna ecoregions are also considered in this study:
The Cerrado (sC) borders the Amazon biome to the southeast. It occupies an area of 2 million km The Guianan savanna (sG) consists of forest patches encircled by extensive grasslands and shrub formations. The area is more susceptible to vegetation fires than typical humid moist forest environments, and the dry season lasts from December to March. The Beni savanna (sB) is a wetland region with riverine gallery forests and small forest islands. The landscape is dominated by the palm species
There are three Köppen–Geiger climate classes (KGCC) that cover most of the study region (Fig.
Köppen–Geiger climate zones in the study area (source:
The ASCAT data were processed using the same procedure as
Downwelling shortwave radiation at the surface and specific humidity were obtained from the Princeton meteorological dataset
Seasonal cycles were determined for precipitation, radiation, humidity, and EWT by averaging data from the entire study duration. Anomalies in precipitation during the drought years were also calculated (as drought year values minus climatology) to provide an indicator of the water stress against which to compare the backscatter, slope, and curvature anomalies.
Figure
Mean and range of ASCAT normalized backscatter, slope, and curvature in the study period (2007–2016). Note that there are no data gaps, so white indicates that the quantity has a value equal to or less than the minimum value indicated on the color bar.
Mean backscatter is highest, with the least variability, in the evergreen forest regions (Fig. S1 in the Supplement). Mean backscatter is 2–2.5 dB lower in the savanna areas, but the range is up to 3 dB, compared to just 0.5 to 1 dB in the forest. The stability of the forest is also apparent in the maps of the slope and curvature. Though there is some variability among the forest ecoregions, the most striking differences in slope and curvature are between the forest and savanna areas. Limited structural and water content changes in the forest canopy result in a limited range of the slope and curvature values in the forest ecoregions. The range of both slope and curvature are highest in the Cerrado areas (Fig. S1). One interesting feature is the difference in mean slope between the Guianan savanna (sG) in the north and the Cerrado (sC) region in the south. The Guianan savanna, with sparse vegetation, has low mean slope values. The Cerrado, on the other hand, shows mean values higher than the evergreen forests. This is unexpected since slope is generally considered to be a measure of vegetation density, and the evergreen forests are much denser than savannas. This will be discussed in detail in Sect.
The mean seasonal cycles in backscatter for all ecoregions of interest are compared in Fig.
Climatologies of backscatter for all ecoregions, with five evergreen forests (dark green), flooded forests (cyan), and three savannas (light green)
Figure
Climatologies of the slope for all ecoregions, with five evergreen forests (dark green), flooded forests (cyan), and three savannas (light green)
In Fig.
In the Cerrado (Fig.
Figure
Climatologies of curvature for all ecoregions, with five evergreen forests (dark green), flooded forests (cyan), and three savannas (light green)
Dominant land cover type (left panel) and fraction (right panel) derived from the Copernicus Global Land Service Land Cover (2015) for the Cerrado region
As described in Sect.
Mean, maximum, and DOY of maximum for backscatter, slope, and curvature over the Cerrado.
Time series averaged per land cover class and box plots of mean, maximum, and DOY of maximum for backscatter, slope, and curvature over the Cerrado.
Seasonal cycle of the slope and radiation per land cover class in the Cerrado region. Only ASCAT pixels in which the fraction of the dominant land cover type exceeds 80 % are included.
Figure
Averaged backscatter as a function of incidence angle for several dates in the Marajó várzea
Recall, from Fig.
Figure
Maps of monthly mean diurnal differences in
For most of the domain, especially the evergreen forests, high values in EWT coincide with negative diurnal differences in backscatter and vice versa. During periods of maximum EWT, the backscatter is higher in the evening than in the morning. This is consistent with the finding that precipitation in tropical South America (since it is generally produced by convective systems) predominantly occurs in the late afternoons and evenings
During the drier periods – e.g., September (h, k) and November (i, l) in the south of the study area – backscatter is higher at 10:00 LT than at 22:00 LT, consistent with the loss of moisture through transpiration during the day. In a light-limited evergreen forest such as the Amazon (rather than a water-limited forest), the canopy photosynthetic capacity seasonality is driven by radiation
In Fig.
Seasonal cycle of diurnal difference in backscatter (black line), radiation (red line), EWT (blue line), and precipitation (bars) for different cover types. Green (yellow) fill indicates days on which backscatter is higher (lower) in the morning than in the evening.
During the study period (2007–2016), two major droughts occurred in Amazonia in 2010 and 2015. Figure
Negative anomalies are observed in
Spatial patterns in anomalies in backscatter, slope, and curvature in response to the 2010 and 2015 droughts.
Figure
Time series of anomalies in backscatter, slope, and curvature for moist forest (fsW) and Cerrado region. The shaded areas indicate the 5th and 95th percentile. The peak drought intervals (June–September 2010 and September–December 2015) are shown within dash–dotted blue lines.
In this study, ASCAT backscatter, slope, and curvature were analyzed in conjunction with meteorological data and terrestrial water storage from GRACE in the Amazon region. Previous results, limited to grasslands, had suggested that the slope and curvature contained useful information for monitoring vegetation water dynamics. However, the current study is the first to attempt to explain the spatial and temporal variations in slope and curvature in terms of seasonal variations in moisture availability and demand. Furthermore, it confirms that the conclusions of
Results show that the unique viewing geometry of ASCAT provides valuable insight into vegetation water dynamics across a diverse range of ecoregions. The timing of the seasonal cycle of normalized backscatter was consistent with that of GRACE EWT, with the maximum (minimum) normalized backscatter coinciding with the maximum (minimum) EWT in all ecoregions. Spatial patterns in the mean and range of the slope reflect the ecoregions within the study area. The seasonal cycle in the slope was found to follow the moisture availability and demand indicated by meteorological data and their influence on phenology. A detailed analysis per land cover type over the Cerrado demonstrated this. Slope dynamics were concurrent with precipitation in croplands and herbaceous cover, although herbaceous cover showed a second peak coinciding with the maximum in radiation. Slope dynamics in shrubs and forest corresponded with radiation, although the onset in increasing slope preceded the onset of increasing radiation. This may be due to leaf flushing, but it is difficult to draw a firmer conclusion given the limited availability of ground data
Diurnal variations (i.e., the difference between morning and evening overpasses) were generally small, particularly in the evergreen forests. Nonetheless, their relation to the timing of precipitation highlights the importance of overpass time in using microwave observations for vegetation monitoring. Diurnal differences in backscatter during the dry season are dominated by transpiration losses. Long-term monitoring of these diurnal differences could provide insight into moisture availability and its influence on transpiration and vegetation functioning
For regions with non-closed-canopy conditions and significant soil contribution, the water sensitivity of the slope and curvature may be influenced, or even dominated, by soil moisture dynamics
The improved understanding of the slope and curvature gained in this study is valuable in terms of our ability to use ASCAT for vegetation monitoring and specifically for vegetation water dynamics. Slope and curvature may be influenced by the number and distribution of the scatterers, and their dielectric properties, all of which influence the optical depth, i.e., the attenuation of the signal by the vegetation. Our improved understanding of the slope and curvature and how they are affected by vegetation structure and water content and interactions between the soil and vegetation is essential to improve our ability to interpret and optimally use VOD derived from ASCAT. Therefore, this research contributes directly to the continued development of the ASCAT VOD products. For example, it provides further insights in the VOD calculated from ASCAT by
The ASCAT data were processed using the WAter Retrieval Package
As input for the slope and curvature calculation with the WAter Retrieval Package developed by TU Wien, the ASCAT Level 1 Sigma0 resampled at 12.5 km Swath Grid – Metop – Global calibrated backscatter data are used. These data are available from the EUMETSAT data centre.
The supplement related to this article is available online at:
AP, SCSD, and MV were responsible for the conceptualization, methodology, formal analysis, investigation, visualization, and writing (original draft preparation). SH provided resources (ASCAT data). RO contributed to the investigation. SCSD and MV provided supervision. All authors contributed to writing (review and editing) the paper.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “Microwave remote sensing for improved understanding of vegetation–water interactions (BG/HESS inter-journal SI)”. It is a result of the EGU General Assembly 2020, 3–8 May 2020.
Susan Steele-Dunne has been supported by the Netherlands Organization for Scientific Research (NWO) User Support Programme Space Research (project no. ALWGO.2018.036 – “A new perspective on global vegetation water dynamics from radar satellite data”) and NWO Vidi Grant 14126. Mariette Vreugdenhil has been supported by ESA's Living Planet Fellowship SHRED (contract no. 4000125441/18/I-NS).
This paper was edited by Julia K. Green and reviewed by two anonymous referees.