skip to main content


Search for: All records

Award ID contains: 1916573

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. null (Ed.)
    Neural methods are state-of-the-art for urban prediction problems such as transportation resource demand, accident risk, crowd mobility, and public safety. Model performance can be improved by integrating exogenous features from open data repositories (e.g., weather, housing prices, traffic, etc.), but these uncurated sources are often too noisy, incomplete, and biased to use directly. We propose to learn integrated representations, called EquiTensors, from heterogeneous datasets that can be reused across a variety of tasks. We align datasets to a consistent spatio-temporal domain, then describe an unsupervised model based on convolutional denoising autoencoders to learn shared representations. We extend this core integrative model with adaptive weighting to prevent certain datasets from dominating the signal. To combat discriminatory bias, we use adversarial learning to remove correlations with a sensitive attribute (e.g., race or income). Experiments with 23 input datasets and 4 real applications show that EquiTensors could help mitigate the effects of the sensitive information embodied in the biased data. Meanwhile, applications using EquiTensors outperform models that ignore exogenous features and are competitive with "oracle" models that use hand-selected datasets. 
    more » « less
  2. null (Ed.)
  3. null (Ed.)
  4. null (Ed.)
  5. null (Ed.)
  6. Millions of Californians access drinking water via domestic wells, which are vulnerable to drought andunsustainable groundwater management. Groundwater overdraft and the possibility of longer droughtduration under climate change threatens domestic well reliability, yet we lack tools to assess the impact ofsuch events. Here, we leverage 943 469 well completion reports and 20 years of groundwater elevationdata to develop a spatially-explicit domestic well failure model covering California’s Central Valley. Ourmodel successfully reproduces the spatial distribution of observed domestic well failures during the severe2012–2016 drought(n=2027). Next, the impact of longer drought duration(5–8years)on domestic wellfailure is evaluated, indicating that if the 2012–2016 drought would have continued into a 6 to 8 year longdrought, a total of 4037–5460 to 6538–8056 wells would fail. The same drought duration scenarios withan intervening wet winter in 2017 lead to an average of498 and 738 fewer well failures. Additionally, wemap vulnerable wells at high failure risk andfind that they align with clusters of predicted well failures.Lastly, we evaluate how the timing and implementation of different projected groundwater managementregimes impact groundwater levels and thus domestic well failure. When historic overdraft persists until2040, domestic well failures range from 5966 to 10 466(depending on the historic period considered).When sustainability is achieved progressively between 2020 and 2040, well failures range from 3677 to6943, and from 1516 to 2513 when groundwater is not allowed to decline after 2020. 
    more » « less