skip to main content

Title: Stream Boundary Detection of a Hyper-Arid, Polar Region Using a U-Net Architecture: Taylor Valley, Antarctica
Convolutional neural networks (CNNs) are becoming an increasingly popular approach for classification mapping of large complex regions where manual data collection is too time consuming. Stream boundaries in hyper-arid polar regions such as the McMurdo Dry Valleys (MDVs) in Antarctica are difficult to locate because they have little hydraulic flow throughout the short summer months. This paper utilizes a U-Net CNN to map stream boundaries from lidar derived rasters in Taylor Valley located within the MDVs, covering ∼770 km2. The training dataset consists of 217 (300 × 300 m2) well-distributed tiles of manually classified stream boundaries with diverse geometries (straight, sinuous, meandering, and braided) throughout the valley. The U-Net CNN is trained on elevation, slope, lidar intensity returns, and flow accumulation rasters. These features were used for detection of stream boundaries by providing potential topographic cues such as inflection points at stream boundaries and reflective properties of streams such as linear patterns of wetted soil, water, or ice. Various combinations of these features were analyzed based on performance. The test set performance revealed that elevation and slope had the highest performance of the feature combinations. The test set performance analysis revealed that the CNN model trained with elevation independently received a precision, recall, and F1 score of 0.94±0.05, 0.95±0.04, and 0.94±0.04 respectively, while slope received 0.96±0.03, 0.93±0.04, and 0.94±0.04, respectively. The performance of the test set revealed higher stream boundary prediction accuracies along the coast, while inland performance varied. Meandering streams had the highest stream boundary prediction performance on the test set compared to the other stream geometries tested here because meandering streams are further evolved and have more distinguishable breaks in slope, indicating stream boundaries. These methods provide a novel approach for mapping stream boundaries semi-automatically in complex regions such as hyper-arid environments over larger scales than is possible for current methods.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Remote Sensing
Page Range / eLocation ID:
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Advanced deep learning methods combined with regional, open access, airborne Light Detection and Ranging (LiDAR) data have great potential to study the spatial extent of historic land use features preserved under the forest canopy throughout New England, a region in the northeastern United States. Mapping anthropogenic features plays a key role in understanding historic land use dynamics during the 17th to early 20th centuries, however previous studies have primarily used manual or semi-automated digitization methods, which are time consuming for broad-scale mapping. This study applies fully-automated deep convolutional neural networks (i.e., U-Net) with LiDAR derivatives to identify relict charcoal hearths (RCHs), a type of historical land use feature. Results show that slope, hillshade, and Visualization for Archaeological Topography (VAT) rasters work well in six localized test regions (spatial scale: <1.5 km2, best F1 score: 95.5%), but also at broader extents at the town level (spatial scale: 493 km2, best F1 score: 86%). The model performed best in areas with deciduous forest and high slope terrain (e.g., >15 degrees) (F1 score: 86.8%) compared to coniferous forest and low slope terrain (e.g., <15 degrees) (F1 score: 70.1%). Overall, our results contribute to current methodological discussions regarding automated extraction of historical cultural features using deep learning and LiDAR. 
    more » « less
  2. Abstract

    Streams in the McMurdo Dry Valleys (MDVs) of Antarctica moderate an important hydrologic and biogeochemical connection between upland alpine glaciers, valley‐bottom soils, and lowland closed‐basin lakes. Moreover, MDV streams are simple but dynamic systems ideal for studying interacting hydrologic and ecological dynamics. This work synthesizes 20 years of hydrologic data, collected as part of the MDVs Long‐Term Ecological Research project, to assess spatial and temporal dynamics of hydrologic connectivity between glaciers, streams, and lakes. Long‐term records of stream discharge (Q), specific electrical conductance (EC), and water temperature (T) from 18 streams were analysed in order to quantify the magnitude, duration, and frequency of hydrologic connections over daily, annual, and inter‐annual timescales. At a daily timescale, we observe predictable diurnal variations in Q, EC, and T. At an annual timescale, we observe longer streams to be more intermittent, warmer, and have higher median EC values, compared to shorter streams. Longer streams also behave chemostatically with respect to EC, whereas shorter streams are more strongly characterized by dilution. Inter‐annually, we observe significant variability in annual runoff volumes, likely because of climatic variability over the 20 record years considered. Hydrologic connections at all timescales are vital to stream ecosystem structure and function. This synthesis of hydrologic connectivity in the MDVs provides a useful end‐member template for assessing hydrologic connectivity in more structurally complex temperate watersheds. Copyright © 2016 John Wiley & Sons, Ltd.

    more » « less
  3. Abstract

    Stream fishes are restricted to specific environments with appropriate habitats for feeding and reproduction. Interactions between streams and surrounding landscapes influence the availability and type of fish habitat, nutrient concentrations, suspended solids, and substrate composition. Valley width and gradient are geomorphological variables that influence the frequency and intensity that a stream interacts with the surrounding landscape. For example, in constrained valleys, canyon walls are steeply sloped and valleys are narrow, limiting the movement of water into riparian zones. Wide valleys have long, flat floodplains that are inundated with high discharge. We tested for differences in fish assemblages with geomorphology variation among stream sites. We selected rivers in similar forested and endorheic ecoregion types of the United States and Mongolia. Sites where we collected were defined as geomorphologically unique river segments (i.e., functional process zones; FPZs) using an automated ArcGIS‐based tool. This tool extracts geomorphic variables at the valley and catchment scales and uses them to cluster stream segments based on their similarity. We collected a representative fish sample from replicates of FPZs. Then, we used constrained ordinations to determine whether river geomorphology could predict fish assemblage variation. Our constrained ordination approach using geomorphology to predict fish assemblages resulted in significance using fish taxonomy and traits in several watersheds. The watersheds where constrained ordinations were not successful were next analyzed with unconstrained ordinations to examine patterns among fish taxonomy and traits with geomorphology variables. Common geomorphology variables as predictors for taxonomic fish assemblages were river gradient, valley width, and valley slope. Significant geomorphology predictors of functional traits were valley width‐to‐floor width ratio, elevation, gradient, and channel sinuosity. These results provide evidence that fish assemblages respond similarly and strongly to geomorphic variables on two continents.

    more » « less
  4. Abstract Predicting the edges of species distributions is fundamental for species conservation, ecosystem services, and management decisions. In North America, the location of the upstream limit of fish in forested streams receives special attention, because fish-bearing portions of streams have more protections during forest management activities than fishless portions. We present a novel model development and evaluation framework, wherein we compare 26 models to predict upper distribution limits of trout in streams. The models used machine learning, logistic regression, and a sophisticated nested spatial cross-validation routine to evaluate predictive performance while accounting for spatial autocorrelation. The model resulting in the best predictive performance, termed UPstream Regional LiDAR Model for Extent of Trout (UPRLIMET), is a two-stage model that uses a logistic regression algorithm calibrated to observations of Coastal Cutthroat Trout ( Oncorhynchus clarkii clarkii ) occurrence and variables representing hydro-topographic characteristics of the landscape. We predict trout presence along reaches throughout a stream network, and include a stopping rule to identify a discrete upper limit point above which all stream reaches are classified as fishless. Although there is no simple explanation for the upper distribution limit identified in UPRLIMET, four factors, including upstream channel length above the point of uppermost fish, drainage area, slope, and elevation, had highest importance. Across our study region of western Oregon, we found that more of the fish-bearing network is on private lands than on state, US Bureau of Land Mangement (BLM), or USDA Forest Service (USFS) lands, highlighting the importance of using spatially consistent maps across a region and working across land ownerships. Our research underscores the value of using occurrence data to develop simple, but powerful, prediction tools to capture complex ecological processes that contribute to distribution limits of species. 
    more » « less
  5. Abstract

    The glacial meltwater streams in the McMurdo Dry Valleys (MDVs), Antarctica only flow during the austral summer and contain abundant algal mats which grow at the onset of flow. Their relative abundance in stream channels of this polar desert make the streams biogeochemical hot spots. The MDVs receive minimal precipitation as snow, which is redistributed by wind and deposited in distinct locations, some of which become persistent snow patches each year. Previous studies identified that MDV streamflow comes from a combination of glacier ice and snow, although snow was assumed to contribute little to the overall water budget. This study uses a combination of satellite imagery, terrain analysis, and field measurements to determine where snow patches accumulate and persist across MDV watersheds, and to quantify the potential hydrologic and biogeochemical contributions of snow patches to streams. Watersheds near the coast have the highest snow‐covered area and longest snow persistence. Many of these snow patches accumulate within the stream channels, which results in the potential to contribute to streamflow. During the summer of 2021–2022, stream channel snow patches had the potential to contribute anywhere between <1% and 90% of the total annual discharge in Lake Fryxell Basin streams, and may increase with different hydrometeorological conditions. On average the potential inputs from snow patches to streamflow was between 12% and 25% of the annual discharge during the 2021–2022 season, as determined by snow area and SWE. Snow patches in the majority of the watersheds had higher nitrogen and phosphorous concentrations than stream water, and six streams contained snow with higher N:P ratios than the average N:P in the stream water. This suggests that if such patches melt early in the summer, these nutrient and water inputs could occur at the right time and stoichiometry to be crucial for early season algal mat growth.

    more » « less