Search for: All records

Coastal urban areas like New York City (NYC) are more vulnerable to urban pluvial flooding particularly because the rapid runoff from extreme rainfall events can be further compounded by the co-occurrence of high sea-level conditions either from tide or storm surge leading to compound flooding events. Present-day urban pluvial flooding is a significant challenge for NYC and this challenge is expected to become more severe with the greater frequency and intensity of storms and sea-level rise (SLR) in the future. In this study, we advance NYC’s assessment of present and future exposure to urban pluvial flooding through simulating various storm scenarios using a citywide hydrologic and hydraulic model. This is the first citywide analysis using NYC’s drainage models focusing on rainfall-induced flooding. We showed that the city’s stormwater system is highly vulnerable to high-intensity short-duration “cloudburst” events, with the extent and volume of flooding being the largest during these events. We further showed that rainfall events coupled with higher sea-level conditions, either from SLR or storm surge, could significantly increase the volume and extent of flooding in the city. We also assessed flood exposure in terms of the number of buildings and length of roads exposed to flooding as well as the number of the affected population. This study informs NYC’s residents of their current and future flood risk and enables the development of tailored solutions to manage increasing flood risk in the city.

 

Compound dry-hot extreme (CDHE) events pose greater risks to the environment, society, and human health than their univariate counterparts. Here, we project decadal-length changes in the frequency and duration of CDHE events for major U.S. cities during the 21st century. Using the Weather Research and Forecasting (WRF) model coupled to an urban canopy parameterization, we find a considerable increase in the frequency and duration of future CDHE events across all U.S. major cities under the compound effect of high-intensity GHG- and urban development-induced warming. Our results indicate that while GHG-induced warming is the primary driver of the increased frequency and duration of CDHE events, urban development amplifies this effect and should not be neglected. Furthermore, We show that the highest frequency amplification of major CDHE events is expected for U.S. cities across the Great Plains South, Southwest, and the southern part of the Northwest National Climate Assessment regions.

 

Spatial data volumes have increased exponentially over the past couple of decades. This growth has been fueled by networked observational devices, remote sensing sources such as satellites, and simulations that characterize spatiotemporal dynamics of phenomena (e.g., climate). Manual inspection of these data becomes unfeasible at such scales. Fitting models to the data offer an avenue to extract patterns from the data, make predictions, and leverage them to understand phenomena and decision-making. Innovations in deep learning and their ability to capture non-linear interactions between features make them particularly relevant for spatial datasets. However, deep learning workloads tend to be resource-intensive. In this study, we design and contrast transfer learning schemes to substantively alleviate resource requirements for training deep learning models over spatial data at scale. We profile the suitability of our methodology using deep networks built over satellite datasets and gridded data. Empirical benchmarks demonstrate that our spatiotemporally aligned transfer learning scheme ensures ~2.87-5.3 fold reduction in completion times for each model without sacrificing on the accuracy of the models. 

Spatial data volumes have grown exponentially over the past several years. The number of domains that spatial data are extensively leveraged include atmospheric sciences, environmental monitoring, ecological modeling, epidemiology, sociology, commerce, and social media among others. These data are often used to understand phenomena and inform decision-making by fitting models to them. In this study, we present our methodology to fit models at scale over spatial data. Our methodology encompasses segmentation, spatial similarity based on the dataset(s) under consideration, and transfer learning schemes that are informed by the spatial similarity to train models faster while utilizing fewer resources. We consider several model fitting algorithms and execution within containerized environments as we profile the suitability of our methodology. Our benchmarks validate the suitability of our methodology to facilitate faster, resource-efficient training of models over spatial data. 

Geospatial data collections are now available in a multiplicity of domains. The accompanying data volumes, variety, and diversity of encoding formats within these collections have all continued to grow. These data offer opportunities to extract patterns, understand phenomena, and inform decision making by fitting models to the data. To ensure accuracy and effectiveness, these models need to be constructed at geospatial extents/scopes that are aligned with the nature of decision-making — administrative boundaries such as census tracts, towns, counties, states etc. This entails construction of a large number of models and orchestrating their accompanying resource requirements (CPU, RAM and I/O) within shared computing clusters. In this study, we describe our methodology to facilitate model construction at scale by substantively alleviating resource requirements while preserving accuracy. Our benchmarks demonstrate the suitability of our methodology. 

We describe our methodology to support time-series forecasts over spatial datasets using the Prophet library. Our approach underpinned by our transfer learning scheme ensures that model instances capture subtle regional variations and converge faster while using fewer resources. Our benchmarks demonstrate the suitability of our methodology. 

Climate change, population growth, urbanization, and interactions thereof may alter the water supply‐demand balance and lead to shifts in water shortage characteristics at different timescales. This study proposes an approach to improve the vulnerability assessments of U.S. river basins to the shortage at the interannual to decadal timescales by characterizing shifts in intensity, duration, and frequency (IDF) of water shortage events from current (1986–2015) to future (2070–2099) periods. The results indicate that under the driest future climate projection, the frequency and intensity of over‐year (D > 12 months) events at the monthly scale and decadal (D > 10 years) events at the annual scale tend to increase in the Southwest, Southern, middle Great Plain, and Great Lakes regions. Conversely, the frequency of interannual (D < 12 months) events at the monthly scale and annual (D > 1 year) and multi‐year (D > 3 years) events at the annual scale is likely to increase in the West Coast regions. Besides, river basins with a higher rate of aridification are likely to experience more frequent over‐year (D > 12 months) events, while river basins with a decrease in aridification were projected to undergo more frequent interannual (D < 12 months) events due to an increase in the variability of extreme weather anomalies within a year. The findings of this study provide new insights to understand and characterize vulnerability to water shortage under current and future water supply‐demand conditions and can inform the development of effective mitigation and/or adaptation strategies.