More Like this

1. Byzantine-tolerant federated Gaussian process regression for streaming data

   Zhang, Xu; Yuan, Zhenyuan; Zhu, Minghui (December 2022, Conference on Neural Information Processing Systems)

   In this paper, we consider Byzantine-tolerant federated learning for streaming data using Gaussian process regression (GPR). In particular, a cloud and a group of agents aim to collaboratively learn a latent function where some agents are subject to Byzantine attacks. We develop a Byzantine-tolerant federated GPR algorithm, which includes three modules: agent-based local GPR, cloud-based aggregated GPR and agent-based fused GPR. We derive the upper bounds on prediction error between the mean from the cloud-based aggregated GPR and the target function provided that Byzantine agents are less than one quarter of all the agents. We also characterize the lower and upper bounds of the predictive variance. Experiments on a synthetic dataset and two real-world datasets are conducted to evaluate the proposed algorithm. 

   more » « less
2. Airborne LiDAR Point Cloud Data, Drained Lake Basin Study Areas on the North Slope of Alaska, 22-27 July 2021

   https://doi.org/10.18739/A2TD9N940  

   Larsen, Christopher; Jones, Benjamin (January 2023, NSF Arctic Data Center)

   The 2021 airborne Light Detection and Ranging (LiDAR) data was acquired by the University of Alaska Geophysical Institute for an National Science Foundation (NSF) funded project focused on catastrophic arctic lake drainage in northern Alaska. The data was acquired at six key field study sites that included drained lake basins north of Teshekpuk Lake, perched lakes on the Ikpikpuk River Delta, tapped lakes at Drew Point, the large drained lake basin complex at the Pik Dunes, a lake and drained lake basin complex at the Anaktuvuk River tundra fire site, and the cascading lake drainage events along the Chipp and Alaktak Rivers. This dataset encompasses 300 square kilometers of terrestrial and aquatic tundra settings in northern Alaska. The data were acquired between 22 July and 27 July 2021 at an estimated density of 16-20 points per square meter (ppm) using an Reigl VQ-580ii LiDAR system flying at an altitude of 750 meters above ground level. The vertical accuracy of this dataset is 10 centimeters. The data have been post-processed to WGS84 UTM Zone 5 North in Ellipsoid Heights (meters). 

   more » « less
3. Ground Penetrating Radar Investigation of Late Pleistocene Shorelines of Pluvial Lake Clover, Elko County, Nevada, USA

   https://doi.org/10.3390/quat3010009  

   Munroe, Jeffrey S. (March 2020, Quaternary)

   Beach ridges constructed by pluvial Lake Clover in Elko County, Nevada during the Late Pleistocene were investigated with ground-penetrating radar (GPR). The primary objective was to document the internal architecture of these shorelines and to evaluate whether they were constructed during lake rise or fall. GPR data were collected with a ground-coupled 400-Mhz antenna and SIR-3000 controller. To constrain the morphology of the ridges, detailed topographic surveys were collected with a Topcon GTS-235W total station referenced to a second class 0 vertical survey point. GPR transects crossed the beach ridge built by Lake Clover at its highstand of 1725 m, along with seven other ridges down to the lowest beach at 1712 m. An average dielectric permittivity of 5.0, typical for dry sand and gravel, was calculated from GPR surveys in the vicinity of hand-excavations that encountered prominent stratigraphic discontinuities at known depths. Assuming this value, consistent radar signals were returned to a depth of ~3 m. Beach ridges are resolvable as ~90 to 150-cm thick stratified packages of gravelly sand overlying a prominent lakeward-dipping reflector, interpreted as the pre-lake land surface. Many ridges contain a package of sediment resembling a buried berm at their core, typically offset in a landward direction from the geomorphic crest of the beach ridge. Sequences of lakeward-dipping reflectors are resolvable beneath the beach face of all ridges. No evidence was observed to indicate that beach ridges were submerged by higher water levels after their formation. Instead, the GPR data are consistent with a model of sequential ridge formation during a monotonic lake regression. 

   more » « less
4. The Lake Tahoe Basin Stream Catchment Database: A resource for water quality monitoring in the basin

   https://doi.org/10.17632/sn5dpnzc83.1  

   Dilts, Thomas; Blaszczak, Joanna (May 2025, Mendeley Data)

   Although the Lake Tahoe Basin and its receiving waterbody, Lake Tahoe, are intensively monitored, managed, and studied, there has been no centralized resource for evaluating variation in environmental characteristics among watersheds (i.e., catchments). To address this opportunity, we compiled and calculated 161 variables for 60 non-overlapping contiguous watersheds draining to Lake Tahoe . Watershed-scale variables include climatic, topographic, vegetation, edaphic, hydrologic, and anthropogenic characteristics. Data were downloaded from publicly-available sources including: the National Elevation Dataset, USDA SSURGO soils, Calfire FRAP dataset of fire perimeters, the National Land Cover Dataset, and the Rangeland Analysis Platform. We compiled data in a Geographic Information System at the scale of the watershed. Existing and custom scripts were used to process data and derive variables that could not be obtained from existing databases. These data will be useful for environmental managers and scientists who work in the Lake Tahoe Basin and can assist with future site selection intended to span environmental gradients. 

   more » « less
5. North Temperate Lakes LTER: Pelagic Prey - Sonar Data 2001 - current

   https://doi.org/10.6073/pasta/5eacf039425717bc9bc766962489e184  

   Magnuson, John; Carpenter, Stephen; Stanley, Emily (January 2022, Environmental Data Initiative)

   Total pelagic fish abundance data were collected annually in mid-summer using sonar along a set of transects in each of eight lakes (Allequash, Big Muskellunge, Crystal, Sparkling, Trout, Mendota, Monona, and Fish), from 1981-1999, and in Lakes Monona and Fish from 1995-1999. This data is not available online (contact gahler@wisc.edu). No data was collected in 2000. In 2001, collection resumed on Crystal, Sparkling, and Trout. In 2005, collection resumed on Lake Mendota. This data is included in this dataset as CSV files. The data represent lake-wide density estimates for abundant pelagic prey species in each lake. The sampling on each lake was conducted in depths greater than 5 meters to avoid hazards to equipment. In addition, because of the near field acoustic effects, the upper 2 meters of the water column is not represented in the data. Although they were rare, large targets representing predatory species were excluded from the density estimation for pelagic prey species using the proportion of large targets identified during single target analysis on each lake. Densities for Sparkling, Crystal and Mendota are for the entire basin of each lake. The data shown for Trout Lake represent densities in only the south basin. Number of sites: 4 

   more » « less