查看更多>>摘要：Tropical dry forests harbor major carbon stocks but are disappearing rapidly across the globe as agriculture expands into them. Unfortunately, carbon emissions from deforestation in dry forests remain poorly understood as high spatial-temporal and vertical heterogeneity complicate biomass mapping. Here, we use a novel Gradient Boosted Regression framework to test the relative gains of combining optical (MODIS) and radar (Sentinel 1) time series, as well as lidar-based (GEDI) canopy-height information, to map biomass in tropical dry forests. We apply our approach across the entire Dry Chaco ecoregion (about 800,000 km2), using an extensive ground dataset of forest inventory plots for training and validation, to map above-ground biomass (AGB) for the year 2019. Our best AGB model had an r2 of 0.89 (RMSE = 15.1 t/ha) with an estimated AGB in remaining natural vegetation of 4.65 Gt (+/- 0.9 Gt). Seasonal metrics from EVI time-series, combined with seasonal Sentinel 1 metrics, had the highest predictive power, while adding GEDI-based canopy height did not improve models. Our resulting AGB maps had a much higher level of agreement with independent ground-data than global AGB products (agreements between r2 = 0.07-0.41), which all suffer from a huge, up to 14-fold, underestimation of AGB in the Chaco. Most of the remaining AGB stored in Chaco woodlands is found in Argentina (2.4 Gt AGB), followed by Paraguay (1.13 Gt AGB) and Bolivia (1.11 Gt AGB). Our results also highlight that 71% of the remaining AGB is located outside protected areas, and around half of the remaining AGB occurs on land utilized by traditional communities. Together, our analyses reveal substantial risk of continued high carbon emissions should agricultural expansion progress. Considerable co-benefits appear to exist between protecting traditional livelihoods and carbon stocks. Our map, the most accurate and fine-scale AGB map for this global deforestation hotspot, can serve as a basis for land-use and conservation planning aimed at leveraging such co-benefits. More broadly, our analyses reveal the considerable potential of combining time series of optical and radar data for a more reliable mapping of above-ground biomass in tropical dry forests and savannas.
Morris, AshleyMoholdt, GeirGray, LaurenceSchuler, Thomas Vikhamar...
16页
查看更多>>摘要：Satellite radar altimetry is widely used to measure glacier and ice sheet elevation changes, but can suffer from uncertainties related to geolocation and signal penetration. The unique capabilities of ESA's CryoSat-2 allow for accurate geolocation but impacts from signal penetration persist. This study uses surface elevations from Global Navigation Satellite System and airborne laser transects over the Austfonna ice cap, Svalbard, to measure the elevation bias of CryoSat-2 Point-of-closest-approach (POCA) and swath points, and to provide validation for dhdt estimates derived through the application of a least-squares plane-fit algorithm to these data. The mean elevation bias of swath points varies between 1 and 1.5 m of penetration, which is close to observed winter snow depths. Histograms of POCA elevation bias for the applied leading-edge retracker peak near the surface, with a distribution skewed towards the sub-surface. At the onset of surface melt, surface scattering dominates backscatter, and penetration reduces. This results in spurious peaks in derived elevation and mass change time series. In spite of this seasonal variability in elevation bias, the validation dhdt dataset demonstrates that the CryoSat-2 dhdt estimates are robust on multi-year timescales. The transition from volume to surface scattering suggests the potential to estimate yearly snowpack thickness.
查看更多>>摘要：We report a new version and an empirical evaluation of a forest reflectance model based on photon recollision probability (p). For the first time, a p-based approach to modeling forest reflectance was tested in a wide range of differently structured forests from different biomes. To parameterize the model, we measured forest canopy structure and spectral characteristics for 50 forest plots in four study sites spanning from boreal to temperate biomes in Europe (48 degrees -62 degrees N). We compared modeled forest reflectance spectra against airborne hyperspectral data at wavelengths of 450-2200 nm. Large overestimation occurred, especially in the near-infrared region, when the model was parameterized considering only leaves or needles as plant elements and assuming a Lambertian canopy. The model root mean square error (RMSE) was on average 80%, 80%, 54% for coniferous, broadleaved, and mixed forests, respectively. We suggest a new parameterization that takes into account the nadir to hemispherical reflectance ratio of the canopy and contribution of woody elements to the forest reflectance. We evaluated the new parameterization based on inversion of the model, which resulted in average RMSE of 20%, 15%, and 11% for coniferous, broadleaved, and mixed forests. The model requires only few structural parameters and the spectra of foliage, woody elements, and forest floor as input. It can be used in interpretation of multi- and hyperspectral remote sensing data, as well as in land surface and climate modeling. In general, our results also indicate that even though the foliage spectra are not dramatically different between coniferous and broadleaved forests, they can still explain a large part of reflectance differences between these forest types in the near-infrared, where sensitivity of the reflectance of dense forests to changes in the scattering properties of the foliage is high.
查看更多>>摘要：A fully physically-based algorithm is presented and used to retrieve nighttime sea-surface (skin) temperatures (SSTs) from spectra recorded by the Infrared Atmospheric Sounder Interferometer (IASI). The main advantage of the proposed approach, based on radiative transfer calculations, is to provide a dataset totally independent of any in-situ measurement or model. The SSTs are retrieved at the IASI spot resolution (clear sky) for the three Metop platforms, today leading to a continuous record over more than 13 years, which is planned to be extended for at least another decade. Five atmospheric transparency windows centered at 3.8, 4.0, 4.7, 9.0 and 11 mu m have been used in the retrieval. Results are compared with in-situ temperatures provided by the buoys network as well as with other satellite datasets. This first analysis enables to draw the following three main conclusions. First, we confirm the interest of using short wavelengths around 4 mu m which enables accurate estimates characterized by a reduced bias with respect to those around 9 and 11 mu m, thanks to significantly weaker sensitivities to errors in the modeling of the water vapor contribution to the atmospheric absorption and emission. Second, we show that the observed dependences of the skin-depth temperature difference on various observed-scene parameters (wind speed, air-sea temperature difference, humidity) are consistent with predictions of theoretical and empirical models, as well as with those from other SST products. Finally, the overall averaged bias between our retrieved values converted to depth temperatures and those provided by in-situ measurements is -0.04 K (median: -0.024 K), with a standard deviation (SD) of 0.38 K (Robust SD = 0.25 K). The bias is thus well below the 0.1 K required for a SST product usable for climate studies and extremely stable over more than a decade. Then, we investigate the consistency between the three generations of IASI on-board Metop-A, -B, and -C over their overlap periods (2013-2020 for Metop-A and Metop-B, and 2019-2020 for Metop-C) by comparing the associated retrieved monthly-averaged SSTs over 1 degrees x1 degrees grids. The mean differences obtained are lower than 0.02 K, with a SD of 0.3 K consistent with the natural variation of the SST within a month. This demonstrates the high stability of the SSTs retrieved using spectra recorded from the IASI suite and opens the possibility to generate long time series for climate studies. This is tested in a preliminary investigation of the SST anomalies between 2008 and 2020, with the successful detection of short-term (El Nin similar to o and La Nin similar to a) changes as well as of a global warming trend of about +0.3 K/decade quantitatively consistent with other observations.
查看更多>>摘要：A novel method is introduced to objectively and automatically identify persistent thermal fronts across the ocean from 60 degrees N to 60 degrees S: FROnt Delineation in the Ocean (FRODO), an algorithm developed to delineate sea surface temperature (SST) fronts, was applied to global satellite observations. By relying on long-term averaged frontal probability fields of SST, persistent SST fronts were located globally in an automated manner. The algorithm was applied to AVHRR Pathfinder and MODIS observations, spanning the period from 1982 through 2011, and resulted in new insights regarding frontal zones: new fronts were detected, the location of some fronts identified in the existing literature has been updated, the accuracy of the algorithm was tested through comparison with visible color and SST data, and the variability in frontal location over climatological seasons was explored. While significant spatial variability prevents the algorithm from accurately tracking some fronts, the results were found to agree very well with observations in the case of fronts associated with ocean bathymetry. The output of the algorithm consists of five digital global sets of persistent fronts: one based on the 30 year time series and four based on the climatological seasons derived from the same time series, which will facilitate the study of the links between SST fronts and other oceanic processes.
查看更多>>摘要：Remote sensing of nighttime light (NTL) offers a unique opportunity to monitor urban dynamics and human socioeconomic activities directly from space. However, angular observations lead to inconsistencies among observations over the same area on different days, introducing uncertainty into daily NTL time series. This study aims to investigate this angular effect and its drivers using the Visible Infrared Imaging Radiometer Suite/Suomi (VIIRS) Black Marble NTL dataset. First, we proposed a conceptual model of the angular effect and hypothesized the mechanism of how urban three-dimensional (3D) landscapes form the anisotropic characteristics of artificial light observations. Second, we quantified the spatial patterns of the angular effect within five representative cities, and identified three distinctive types of angular effects: negative, U-shaped, and positive. Subsequently, the contribution of landscape factors to the direction (i.e., the type) and magnitude (i.e., NTL change rate with angle) of the angular effect is quantified using multinomial logistic regression and mediation analysis, respectively. The results show that the direction of the angular effect is mainly controlled by building height which determines the blocked and visible parts of artificial light at different satellite viewing angles. The magnitude of the angular effect is determined by both NTL brightness and landscape factors. The mediation analysis shows that landscape factors can have a direct effect on the magnitude of the angular effect as well as an indirect effect on the magnitude by affecting NTL brightness. Among the landscape factors, both vegetation and buildings are indicated to be significantly influential factors with direct and indirect effects. The findings of this research deepen our understanding of the NTL angular effect, guide the development of technologies for reconstructing high-quality daily NTL time series by correcting the angular effect, and help us better monitor high-frequency socioeconomic activities.
查看更多>>摘要：Field spectroscopy is a powerful tool for monitoring leaf functional traits in situ, but it remains unclear whether universal statistical models can be developed to predict traits from spectral information, or whether recalibration is necessary as conditions vary. In particular, multiple leaf traits vary simultaneously across growing seasons, and it is an open question whether these temporal changes can be predicted successfully from hyperspectral data. To explore this question, monthly changes in 21 physiochemical leaf traits and plant spectra were measured for eight deciduous tree species from the UK. Partial least-squares regression (PLSR) was used to evaluate whether each trait could be predicted from a single PLSR model from reflectance spectra, or whether species- and month-level models were needed. Physiochemical traits and spectra varied greatly over the growing season, although there was less variation among mature leaves harvested between June and September. Importantly, leaf spectroscopy was able to predict seasonal variations of most leaf traits accurately, with accuracies of prediction generally higher for mature leaves. However, for several traits, the PLSR estimation models varied among species, and a single PLSR model could not be used to make accurate species-level predictions. Our findings demonstrate that leaf spectra can successfully predict multiple functional foliar traits through the growing season, establishing one of the fundamentals for monitoring and mapping plant functional diversity in temperate forests from air- and spaceborne imaging spectroscopy.
查看更多>>摘要：The top-of-atmosphere (TOA) albedo, a key component of the earth's energy balance, can be monitored regularly by satellite observations. Compared to the previous study Song et al. (2018), this paper estimates TOA albedo by directly linking Advanced Very High Resolution Radiometer (AVHRR) narrowband reflectance with TOA broadband albedo determined by NASA's Clouds and the Earth's Radiant Energy System (CERES) instead of Moderate Resolution Imaging Spectroradiometer (MODIS). The TOA albedo product developed in this study has an increased spatial resolution, from 1 degrees to 0.05 degrees, and its starting year has been extended from 2000 to 1981, compared to the CERES TOA albedo product. Models of lands and oceans are established separately under different atmospheric and surface conditions using gradient boosting regression tree (GBRT) method instead of the linear regression models in the previous study. The root mean square errors (RMSEs) of the cloudy-sky, clear sky and snow-cover models over land are 11.2%, 9.2% and 2.3%, respectively; over oceans they are 14.6%, 10.6% and 5.6%, respectively. Compared to Song et al. (2018), the improvements of the three models over land are 28.8%, 29.2% and 68.6%, respectively. Compared to the CERES product, the new product is much more accurate than that from our previous study. The global monthly mean differences of the TOA albedo obtained with the GBRT product and CERES from 2001 to 2014 are mostly less than 5%.
查看更多>>摘要：L-band passive microwave remote sensing is currently considered a robust technique for global monitoring of soil moisture. However, soil roughness complicates the relationship between brightness temperature and soil moisture, with current soil moisture retrieval algorithms typically assuming a constant roughness parameter globally, leading to a potential degradation in retrieval accuracy. This current investigation established a towerbased experiment site in Victoria, Australia. P-band (-40-cm wavelength/0.75 GHz) was compared with L-band (-21-cm wavelength/1.41 GHz) over random and periodic soil surfaces to determine if there is an improvement in brightness temperature simulation and soil moisture retrieval accuracy for bare soil conditions, due to reduced roughness impact when using a longer wavelength. The results showed that P-band was less impacted by random and periodic roughness than L-band, evidenced by more comparable statistics across different roughness conditions. The roughness effect from smooth surfaces (e.g., 0.8-cm root-mean-square height and 11.1-cm correlation length) could be potentially ignored at both P- and L-band with satisfactory simulation and retrieval performance. However, for rougher soil (e.g., 1.6-cm root-mean-square height and 6.8-cm correlation length), the roughness impact needed to be accounted for at both P- and L-band, with P-band observations showing less impact than L-band. Moreover, a sinusoidal soil surface with 10-cm amplitude and 80-cm period substantially impacted the brightness temperature simulation and soil moisture retrieval at both P- and L-band, which could not be fully accounted for using the SMOS and SMAP default roughness parameters. However, when retrieving roughness parameters along with soil moisture, the ubRMSE at P-band over periodic soil was improved to a similar level (0.01-0.02 m3/m3) as that of smooth flat soil (0.01 m3/m3), while L-band showed higher ubRMSE over the periodic soil (0.03-0.04 m3/m3) than over smooth flat soil (0.01 m3/m3). Accordingly, periodic roughness effects were reduced by using observations at P-band.
查看更多>>摘要：Compared with disturbance maps produced at annual or multi-year time steps, monthly mapping of forest harvesting can provide more temporal details needed for studying the socio-economic drivers (e.g., differentiating salvage logging and slash-and-burn from other timber harvesting) of harvesting and characterizing the associated intra-annual carbon and hydrological dynamics. Frequent cloud cover limits the application of optical remote sensing in timely mapping of forest changes. The freely available Sentinel-1 synthetic aperture radar (SAR) sensor provides an unprecedented opportunity to achieve more frequent mapping of forest harvesting than ever before (i.e., at monthly interval). The unique landscape pattern of forest harvesting from Sentienl-1 data (i. e., how a harvested patch contrasts to surrounding intact forests) holds critical information for harvesting mapping but have not been fully explored. In this study, we propose a deep learning-based (i.e., U-Net) approach using the landscape pattern from Sentinel-1 data to produce monthly maps of forest harvesting in two deforestation hotspots -California, USA and Rondo<SIC>nia, Brazil - for as long as three years. Our results show that (1) our proposed approach is reliable (mean F1 scores (the geometric mean of user's and producer's accuracies) 0.74-0.78; mean IoU (the area of intersection over union between the prediction part and target part) 0.59-0.65) for monthly forest harvesting mapping with Sentinel-1 data, outperforming the traditional object-based approach (0.38-0.43 in IoU). The varying harvesting pattern from Sentinel-1 data can be recognized by the U-Net bottleneck block as whole entities, which is the key advantage of our proposed approach; (2) multi-temporal SAR filtering is helpful for improving the accuracies of our proposed approach (increased F1 and IoU for 0.04 and 0.06, respectively); (3) our proposed model can be trained using samples collected during a particular time period over one location and be fine-tuned using sparse local samples from a new area to achieve optimal performance, and hence can greatly reduce training data collection effort when applied to new study sites; (4) forest harvesting maps produced using our approach revealed substantial variations in monthly harvesting activities: in Rondo<SIC>nia, most of the forest harvest occurred in July/August (the dry season) and about 14% of the dry season harvesting were followed by fires (i.e., slash-and-burn); in California, the rates of forest harvesting were relatively stable, but abnormally high values could occur due to salvage logging after big fires. Our novel approach for mapping forest harvesting at monthly interval represents an important step towards timely monitoring of forest harvesting and assisting stakeholders in developing sustainable strategy of forest management, especially for regions with frequent cloud cover.
NSTL
Elsevier