Ice roughness estimation via remotely piloted aircraft and photogrammetry

The monitoring of fluvial ice covers can be time intensive, dangerous, and costly if detailed data are required. Ice covers on a river surface cause resistance to water flow, which increases upstream water levels. Ice with a higher degree of roughness causes increased flow resistance and therefore even higher upstream water levels. Aerial images collected via remotely piloted aircraft (RPA) were processed with structure from motion photogrammetry to create a digital elevation model (DEM) and then produce quantitative measurements of surface ice roughness. Images and surface ice roughness values were collected over 2 years on the Dauphin River in Manitoba, Canada. It was hypothesized that surface ice roughness would be indicative of subsurface ice roughness. This hypothesis was tested by comparing RPA-measured surface ice roughness values to those predicted by the Nezhikhovskiy equation, wherein subsurface ice roughness is proportional to ice thickness. Various statistical metrics were used to represent the roughness height of the DEMs. Strong trends were identified in the comparison of RPA-measured ice surface roughness to subsurface ice roughness values predicted by the Nezhikhovskiy equation, as well as with comparisons to ice thickness. The standard deviation and interquartile range of roughness heights were determined to be the most representative statistical metrics and several properties of the DEMs of fluvial ice covers were calculated and observed. No DEMs were found to be normally distributed. This first attempt at using RPA-derived measurements of surface ice roughness to estimate river ice flow resistance is shown to have considerable potential and will hopefully be verified and improved upon by subsequent measurements on a wide variety of rivers and ice covers.

Ice Roughness Estimation via Remotely Piloted Aircraft and Photogrammetry

Structure-from-Motion Photogrammetry conducted with images obtained via Remotely Piloted Aircraft (RPA) has revolutionized the field of land surface monitoring. RPA-Photogrammetry can quickly and easily capture a full 3D representation of a study area. The result of this process is a high-definition Digital Elevation Model (DEM) representing the land surface of a given study area. It is particularly useful in applications where land surface data collection would otherwise be expensive or dangerous. The monitoring of fluvial ice covers can be time-intensive, dangerous, and costly, if detailed data are required. Fluvial ice roughness is a sensitive parameter in hydraulic models and is incredibly difficult to measure directly using traditional field methods. This research hypothesized that the surface roughness of a newly-frozen fluvial ice cover is indicative of subsurface roughness. The hypothesis was tested through a comparison of ice roughness determined through the statistical analysis of RPA-photogrammetry DEMs to ice roughness values predicted by the Nezhikhovskiy equation. The Nezhikhovskiy equation is a widely used empirical method for estimating ice roughness based on observed ice thickness. Hydraulic and topographic data were collected over two years of field research on the Dauphin River in Manitoba, Canada. Various statistical metrics were used to represent the roughness of the DEMs. Strong trends were identified in the comparison of ice cover roughness values determined through RPA-photogrammetry and those calculated via the Nezhikhovskiy equation, as well as with ice thickness. The inter-quartile range of observed roughness heights was determined to be the most representative roughness metric. The maximum peak value performed better in some cases, but the fact that this metric would be heavily influenced by outliers led to it being rejected as a representative metric. Three distinct forms of surface ice roughness were noted: rough, smooth, and ridged. Statistical properties of the DEMs of fluvial ice covers were calculated. No DEMs were found to be normally distributed. k-means clustering analysis was used to group sampled data into two categories, which were interpreted as rough and smooth ice. The inter-quartile range of the smooth and rough categories were found to be 0.01–0.05 meters and 0.07–0.12 meters, respectively. RPA-photogrammetry was concluded to be a suitable method for the monitoring of fluvial ice covers. Other applications of RPA-photogrammetry for the characterization of fluvial ice covers are proposed.

Collocated OLCI optical imagery and SAR radar altimetry from Sentinel3 for enhanced sea ice surface classification

<p>The Copernicus operational Sentinel-3A since February 2016 and Sentinel-3B since April 2018 build on the CryoSat-2 legacy in terms of their synthetic aperture radar (SAR) mode altimetry providing high-resolution radar freeboard elevation data over the polar regions up to 81N. This technology combined with the Ocean and Land Colour Instrument (OLCI) imaging spectrometer offers the first space-time collocated optical imagery and radar altimetry dataset. We use these joint datasets for validation of several existing surface classification algorithms based on Sentinel-3 altimeter echo shapes. We also explore the potential for novel AI techniques such as convolutional neural networks (CNN) for winter and summer sea ice surface classification (i.e. melt pond fraction, lead fraction, sea ice roughness). For lead surface classification we analyse the winters of 2018/19 and 2019/20 and for summer sea ice feature classification we focus on the Sentinel-3A &3B tandem phase of the summer 2018. We compare our CNN models with other existing surface classification algorithms.</p>

Characterizing winter landfast sea-ice surface roughness in the Canadian Arctic Archipelago using Sentinel-1 synthetic aperture radar and the Multi-angle Imaging SpectroRadiometer

Two satellite datasets are used to characterize winter landfast first-year sea-ice (FYI), deformed FYI (DFYI) and multiyear sea-ice (MYI) roughness in the Canadian Arctic Archipelago (CAA): (1) optical Multi-angle Imaging SpectroRadiometer (MISR) and (2) synthetic aperture radar Sentinel-1. The Normalized Difference Angular Index (NDAI) roughness proxy derived from MISR, and backscatter from Sentinel-1 are intercompared. NDAI and backscatter are also compared to surface roughness derived from an airborne LiDAR track covering a subset of FYI and MYI (no DFYI). Overall, NDAI and backscatter are significantly positively correlated when all ice type samples are considered. When individual ice types are evaluated, NDAI and backscatter are only significantly correlated for DFYI. Both NDAI and backscatter are correlated with LiDAR-derived roughness (r = 0.71 and r = 0.74, respectively). The relationship between NDAI and roughness is greater for MYI than FYI, whereas for backscatter and ice roughness, the relationship is greater for FYI than MYI. Linear regression models are created for the estimation of FYI and MYI roughness from NDAI, and FYI roughness from backscatter. Results suggest that using a combination of Sentinel-1 backscatter for FYI and MISR NDAI for MYI may be optimal for mapping winter sea-ice roughness in the CAA.

Arctic Sea Ice Surface Roughness Estimated from Multi-Angular Reflectance Satellite Imagery

Sea ice surface roughness affects ice-atmosphere interactions, serves as an indicator of ice age, shows patterns of ice convergence and divergence, affects the spatial extent of summer meltponds, and affects ice albedo. We have developed a method for mapping sea ice surface roughness using angular reflectance data from the Multi-angle Imaging SpectroRadiometer (MISR) and lidar-derived roughness measurements from the Airborne Topographic Mapper (ATM). Using an empirical data modeling approach, we derived estimates of Arctic sea ice roughness ranging from centimeters to decimeters within the MISR 275-m pixel size. Using independent ATM data for validation, we find that histograms of lidar and multi-angular roughness values were nearly identical for areas with a roughness < 20 cm, but for rougher regions, the MISR-estimated roughness had a narrower range of values than the ATM data. The algorithm was able to accurately identify areas that transition between smooth and rough ice. Because of its coarser spatial scale, MISR-estimated roughness data have a variance about half that of ATM roughness data.