Abstract. In this study, we validate and compare elevation accuracy and geomorphic metrics of satellite-derived digital elevation models (DEMs) on the southern Central Andean Plateau. The plateau has an average elevation of 3.7 km and is characterized by diverse topography and relief, lack of vegetation, and clear skies that create ideal conditions for remote sensing. At 30 m resolution, SRTM-C, ASTER GDEM2, stacked ASTER L1A stereopair DEM, ALOS World 3D, and TanDEM-X have been analyzed. The higher-resolution datasets include 12 m TanDEM-X, 10 m single-CoSSC TerraSAR-X/TanDEM-X DEMs, and 5 m ALOS World 3D. These DEMs are state of the art for optical (ASTER and ALOS) and radar (SRTM-C and TanDEM-X) spaceborne sensors. We assessed vertical accuracy by comparing standard deviations of the DEM elevation versus 307 509 differential GPS measurements across 4000 m of elevation. For the 30 m DEMs, the ASTER datasets had the highest vertical standard deviation at > 6.5 m, whereas the SRTM-C, ALOS World 3D, and TanDEM-X were all < 3.5 m. Higher-resolution DEMs generally had lower uncertainty, with both the 12 m TanDEM-X and 5 m ALOS World 3D having < 2 m vertical standard deviation. Analysis of vertical uncertainty with respect to terrain elevation, slope, and aspect revealed the low uncertainty across these attributes for SRTM-C (30 m), TanDEM-X (12–30 m), and ALOS World 3D (5–30 m). Single-CoSSC TerraSAR-X/TanDEM-X 10 m DEMs and the 30 m ASTER GDEM2 displayed slight aspect biases, which were removed in their stacked counterparts (TanDEM-X and ASTER Stack). Based on low vertical standard deviations and visual inspection alongside optical satellite data, we selected the 30 m SRTM-C, 12–30 m TanDEM-X, 10 m single-CoSSC TerraSAR-X/TanDEM-X, and 5 m ALOS World 3D for geomorphic metric comparison in a 66 km2 catchment with a distinct river knickpoint. Consistent m∕n values were found using chi plot channel profile analysis, regardless of DEM type and spatial resolution. Slope, curvature, and drainage area were calculated and plotting schemes were used to assess basin-wide differences in the hillslope-to-valley transition related to the knickpoint. While slope and hillslope length measurements vary little between datasets, curvature displays higher magnitude measurements with fining resolution. This is especially true for the optical 5 m ALOS World 3D DEM, which demonstrated high-frequency noise in 2–8 pixel steps through a Fourier frequency analysis. The improvements in accurate space-radar DEMs (e.g., TanDEM-X) for geomorphometry are promising, but airborne or terrestrial data are still necessary for meter-scale analysis.
Abstract. Grain-size distributions are a key geomorphic metric of gravel-bed rivers. Traditional measurement methods include manual counting or photo sieving, but these are achievable only at the 1–10 m2 scale. With the advent of unmanned aerial vehicles and increasingly high-resolution cameras, we can now generate orthoimagery over hectares at sub-cm resolution. These scales, along with the complexity of high-mountain rivers, necessitate different approaches for photo sieving. As opposed to other image segmentation methods that use a watershed approach to automatically segment entire images, our open-source algorithm, PebbleCounts, relies on k-means clustering in the spatial and spectral domain and rapid manual selection of well-delineated grains. The result is improved grain-size estimates for complex river-bed imagery, without any post processing. In a second step, we develop a fully automated method, PebbleCountsAuto, that relies on edge detection and filtering suspect grains, without the k-means clustering or manual selection steps. The algorithms are tested in controlled indoor conditions on three arrays of pebbles and then applied to 12 × 1 m2 orthomosaic clips of high-energy mountain rivers collected with a camera-on-mast setup (akin to a low-flying drone). A 20-pixel b-axis length lower truncation is necessary for attaining accurate grain-size distributions. For the k-means PebbleCounts approach, average percentile bias and precision are 0.03 and 0.09 ψ, respectively, for ~ 1.16 mm/pixel images, and 0.07 and 0.05 ψ for one 0.32 mm/pixel image. The automatic approach has higher bias and precision of 0.13 and 0.15 ψ, respectively, for ~ 1.16 mm/pixel images, but similar values of −0.06 and 0.05 ψ for one 0.32 mm/pixel image. For the automatic approach, only at best 70 % of the grains are correct identifications, and typically around 50 %. PebbleCounts operates most effectively at the 1 m2 scale, where the algorithm can be rapidly applied in ~ 5 minutes in many small areas to acquire accurate grain-size data over 10–100 m2 areas. These data can be used to validate PebbleCountsAuto applied at the scale of entire survey sites (102–104 m2). We synthesize results and recommend best practices for image collection, orthomosaic generation, and grain-size measurement using both algorithms.
Abstract. Grain-size distributions are a key geomorphic metric of gravel-bed rivers. Traditional measurement methods include manual counting or photo sieving, but these are achievable only at the 1–10 m2 scale. With the advent of drones and increasingly high-resolution cameras, we can now generate orthoimagery over hectares at millimeter to centimeter resolution. These scales, along with the complexity of high-mountain rivers, necessitate different approaches for photo sieving. As opposed to other image segmentation methods that use a watershed approach, our open-source algorithm, PebbleCounts, relies on k-means clustering in the spatial and spectral domain and rapid manual selection of well-delineated grains. This improves grain-size estimates for complex riverbed imagery, without post-processing. We also develop a fully automated method, PebbleCountsAuto, that relies on edge detection and filtering suspect grains, without the k-means clustering or manual selection steps. The algorithms are tested in controlled indoor conditions on three arrays of pebbles and then applied to 12 × 1 m2 orthomosaic clips of high-energy mountain rivers collected with a camera-on-mast setup (akin to a low-flying drone). A 20-pixel b-axis length lower truncation is necessary for attaining accurate grain-size distributions. For the k-means PebbleCounts approach, average percentile bias and precision are 0.03 and 0.09 ψ, respectively, for ∼1.16 mm pixel−1 images, and 0.07 and 0.05 ψ for one 0.32 mm pixel−1 image. The automatic approach has higher bias and precision of 0.13 and 0.15 ψ, respectively, for ∼1.16 mm pixel−1 images, but similar values of −0.06 and 0.05 ψ for one 0.32 mm pixel−1 image. For the automatic approach, only at best 70 % of the grains are correct identifications, and typically around 50 %. PebbleCounts operates most effectively at the 1 m2 patch scale, where it can be applied in ∼5–10 min on many patches to acquire accurate grain-size data over 10–100 m2 areas. These data can be used to validate PebbleCountsAuto, which may be applied at the scale of entire survey sites (102–104 m2). We synthesize results and recommend best practices for image collection, orthomosaic generation, and grain-size measurement using both algorithms.
Quantitative geomorphic research depends on accurate topographic data often collected via remote sensing. Lidar, and photogrammetric methods like structure-from-motion, provide the highest quality data for generating digital elevation models (DEMs). Unfortunately, these data are restricted to relatively small areas, and may be expensive or time-consuming to collect. Global and near-global DEMs with 1 arcsec (∼30 m) ground sampling from spaceborne radar and optical sensors offer an alternative gridded, continuous surface at the cost of resolution and accuracy. Accuracy is typically defined with respect to external datasets, often, but not always, in the form of point or profile measurements from sources like differential Global Navigation Satellite System (GNSS), spaceborne lidar (e.g., ICESat), and other geodetic measurements. Vertical point or profile accuracy metrics can miss the pixel-to-pixel variability (sometimes called DEM noise) that is unrelated to true topographic signal, but rather sensor-, orbital-, and/or processing-related artifacts. This is most concerning in selecting a DEM for geomorphic analysis, as this variability can affect derivatives of elevation (e.g., slope and curvature) and impact flow routing. We use (near) global DEMs at 1 arcsec resolution (SRTM, ASTER, ALOS, TanDEM-X, and the recently released Copernicus) and develop new internal accuracy metrics to assess inter-pixel variability without reference data. Our study area is in the arid, steep Central Andes, and is nearly vegetation-free, creating ideal conditions for remote sensing of the bare-earth surface. We use a novel hillshade-filtering approach to detrend long-wavelength topographic signals and accentuate short-wavelength variability. Fourier transformations of the spatial signal to the frequency domain allows us to quantify: 1) artifacts in the un-projected 1 arcsec DEMs at wavelengths greater than the Nyquist (twice the nominal resolution, so > 2 arcsec); and 2) the relative variance of adjacent pixels in DEMs resampled to 30-m resolution (UTM projected). We translate results into their impact on hillslope and channel slope calculations, and we highlight the quality of the five DEMs. We find that the Copernicus DEM, which is based on a carefully edited commercial version of the TanDEM-X, provides the highest quality landscape representation, and should become the preferred DEM for topographic analysis in areas without sufficient coverage of higher-quality local DEMs.
It has long been recognized that the grain-size distribution in mixed sand-and gravel-bed channels-and the deposits they form-encode end transport conditions (e.g.,
Abstract. In the arctic and high mountains it is common to measure vertical changes of ice sheets and glaciers via digital elevation model (DEM) differencing. This requires the signal of change to outweigh the noise associated with the datasets. Excluding large landslides, on the ice-free earth the land-level change is smaller in vertical magnitude and thus requires more accurate DEMs for differencing and identification of change. Previously, this has required meter to submeter data at small spatial scales. Following careful corrections, we are able to measure land-level changes in gravel-bed channels and steep hillslopes in the south-central Andes using the SRTM-C (collected in 2000) and the TanDEM-X (collected from 2010 to 2015) near-global 12–30 m DEMs. Long-standing errors in the SRTM-C are corrected using the TanDEM-X as a control surface and applying cosine-fit co-registration to remove ∼1/10 pixel (∼3 m) shifts, fast Fourier transform (FFT) and filtering to remove SRTM-C short- and long-wavelength stripes, and blocked shifting to remove remaining complex biases. The datasets are then differenced and outlier pixels are identified as a potential signal for the case of gravel-bed channels and hillslopes. We are able to identify signals of incision and aggradation (with magnitudes down to ∼3 m in the best case) in two >100 km river reaches, with increased geomorphic activity downstream of knickpoints. Anthropogenic gravel excavation and piling is prominently measured, with magnitudes exceeding ±5 m (up to >10 m for large piles). These values correspond to conservative average rates of 0.2 to >0.5 m yr−1 for vertical changes in gravel-bed rivers. For hillslopes, since we require stricter cutoffs for noise, we are only able to identify one major landslide in the study area with a deposit volume of 16±0.15×106 m3. Additional signals of change can be garnered from TanDEM-X auxiliary layers; however, these are more difficult to quantify. The methods presented can be extended to any region of the world with SRTM-C and TanDEM-X coverage where vertical land-level changes are of interest, with the caveat that remaining vertical uncertainties in primarily the SRTM-C limit detection in steep and complex topography.
<p>Synthetic Aperture Radar (SAR) amplitude measurements from spaceborne sensors are sensitive to surface roughness conditions near their radar wavelength. These data can be exploited to measure gravel-to-sand transitions and downstream gradients in grain size related to geomorphic setting in tectonically active high mountain environments at large spatial scales. The bedload of mixed sand- and gravel-bed rivers can be considered mixed smooth (compacted sand) and rough (gravel) surfaces. Here, we assess backscatter gradients over a large high-mountain alluvial river with aerially exposed sand and gravel bedload using X-band TerraSAR-X/TanDEM-X, C-band Sentinel-1, and L-band ALOS-2 PALSAR-2 radar scenes. In a first step, we compare backscatter response over vegetation-free endmember surfaces within the dry channel bed to assess expected responses and limitations of SAR roughness measurements. We then develop methods to extract smoothed backscatter gradients downstream along the channel using kernel density estimates. In a final step, the presence of sand and gravel bars is analyzed using Fourier frequency analysis, by fitting stretched exponential and power-law models to the power spectrum. We find a large range in backscatter depending on the heterogeneity of contiguous smooth- and rough-patches of bedload material. The SAR signal responds primarily to the fraction of smooth-sand bedload, but is further modified by gravel elements. The sensitivity to gravel is more apparent in longer wavelength (L-band) radar. Because the spatial extent of smooth sand bars is typically < 50 m, only higher resolution sensors (e.g., TerraSAR-X/TanDEM-X) are useful for power spectrum analysis. Our results show the potential for mapping sand-gravel transitions and local geomorphic complexity using SAR amplitude at the scale of large high mountain catchments with aerially exposed bedload.</p>
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.