Evaluating the Water Level Gage Network Density of New York City

Written by: Alex Truby and Amy Piscopo

Introduction

Water level data from riverine and coastal gaging stations is used broadly across an array of industries, including in transportation and shipping, municipal permitting, and industrial processing, to name a few. In the US, most gages are installed and maintained by the US Geological Survey (USGS) and the National Oceanographic and Atmospheric Administration (NOAA). These government entities provide public access to the data, with measurements recorded at 6-minute increments on average. Since the late 1800s, the USGS network has grown to over 11,800 riverine gaging stations [1] and NOAA currently operates over 300 coastal gaging stations.??

Water data has enabled researchers to study the various impacts of climate change, such as sea level rise and the increasing frequency and severity of storms and droughts. Water data also underlies various climate change adaptation efforts, including early-warning alerts for flood conditions, strategic reservoir operations, vulnerability assessments, and flood disclosure requirements. In many of these cases, the existing density of gages, and therefore, the density of data is less than ideal for the intended adaptation effort. The USGS outlines that their current stream gage network covers less than 1% of rivers and streams in the US, not nearly enough to adequately understand the risks posed from flood and drought [1].? Even in regions not historically prone to flooding, such as the greater NYC area, storm events such as Hurricane Sandy illustrate the need for a widespread, finely tuned understanding of flood risk.?

Gage locations were historically selected based on proximity to shipping routes and for irrigation, not necessarily to provide the most accurate flood forecast or to protect public infrastructure from storm surge. For example, flood alerts based on gage data can offer life-saving early warnings, although their accuracy tends to diminish with distance from the gage. Would a denser deployment of gages provide alert recipients with more site-specific notice, thereby reducing skepticism caused by false positives? And offer a more accurate estimate of the flood level???

In this white paper, we propose an approach from spatial statistics to evaluate the adequacy of gage density for climate adaptation use cases. We also explore different criteria that should be considered when deciding where to place additional gages. While the definition of adequate will depend on the specific adaptation problem at hand, we offer this technique as an objective, data-driven option. We focus on New York City as a case study.?

Current Gage Coverage in NYC

Currently, there are five publicly available water level monitoring gages in the New York City metropolitan area. Figure 1 shows each gage’s location and Table 1 provides key information for each gage, including the start date of data collection and whether the data includes any gaps in record. We did not include any inactive gages in this analysis, i.e., gages not yielding water level data at the time of release of this white paper.?

Extending north on the Hudson River, there are an additional 9 active gages over roughly 300 river miles. In comparison to other major rivers in the US, the Hudson River ranks lowest when considering the ratio of adjacent population to gage count, as depicted in Figure 2 below. The adjacent population was defined as the 2020 Census population within fifteen miles on either side of the river, measured perpendicular to the river. Within fifteen miles of the Hudson River, the total population exceeds 11 million.??

While the current coverage in the NYC area is surprisingly sparse for such a densely populated area, this comparison with other major rivers is offered only as a reference point, not as a standard to achieve.? None of the major rivers represent an ideal ‘gage density model’ to follow for use cases like improving flood model accuracy because gages were not installed for this purpose. The question of ‘coverage sufficiency’ depends on the use case and associated goals.? In the next section, we demonstrate an approach for establishing a gage density threshold to support use cases where a correlation of time series data between adjacent gages is expected.

A Variogram Approach to Consider Gage Density??

Variogram models are commonly used in spatial statistics to evaluate spatial variability. These models are relevant for geospatial datasets in which the properties of points closer together have a higher degree of correlation. In this whitepaper, we fit variogram models to water level data from different gage networks to estimate a distance at which water level measurements from gages at varying distances become unrelated.??

Using Divirod’s global database of water level gages, we identified coastal cities (Table 2) with a high density of water level gages and at least one year of recorded data. We focused on gage networks along coastlines and connected estuaries where (a) we could assume some degree of hydrologic connectivity and (b) any engineered influence on water levels was unlikely (e.g., levees). Three cities were evaluated with enough gages to conduct this analysis, and at that, we acknowledge that the cities have fewer gages than ideal for fitting a variogram.??

We normalized the water levels following Divirod’s Water Level Index (WLI) methodology [2] to ensure a reasonable comparison across location types. For each of the coastal cities, we fit a variogram to the WLI timeseries for the last year of the identified gages using a gaussian model to investigate the spatial variance of the WLI timeseries at different lag distances (see example for Vancouver, Canada in Figure 3). As the distance between each pairwise set of gages in each network increases, the correlation between the WLI values decreases until values are no longer correlated, known as the sill of the variogram. The distance at which point the WLI value is no longer representative, known as the variogram range, is reported in Table 2.

The results indicate that some correlation exists between WLI values at gage spacings less than 6-8 km. We consider this variogram range as the representative distance beyond which WLI values become unrelated, and thus, can be interpreted as the radius of coverage around each gage. Given the lack of dense gage networks and the limited number of gages within the three networks we identified, these results should be considered preliminary. Additional local factors such as the interdependence of water bodies, whether the water system is natural or influenced by engineered interventions should also be considered when evaluating the appropriate number of gages for an area of interest.?? Qualifiers aside, we contribute this white paper to demonstrate the approach, and more importantly, to draw attention to the global scarcity of dense gage networks. At present, there are very few regions – both coastal and inland – with a high density of water level monitoring stations. At Divirod, we intend to continuously revisit and update this analysis as we install more dense deployments around the world.

NYC Gage Placement Criteria

The ideal density of a gage deployment should be coupled with location-specific gage siting criteria to target areas most vulnerable to flooding. NYC has conducted extensive flood mapping, as reported in their NYC Hurricane Evacuation Zone Dataset [3]. Should NYC increase its gage density, we recommend targeting these evacuation zones which also represent areas without any current monitoring infrastructure. One benefit of deploying in areas most prone to flooding is the ability to send localized flood alerts with greater accuracy than current systems. More accurate flood alerts can limit false positives and strengthen community trust in the alerts.?

Many of these evacuation zones have been identified by the NYC Mayor’s Office of Climate & Environmental Justice (EJNYC) as environmental justice areas (EJ areas). EJ areas often have fewer resources to manage the impacts of flooding and are defined by EJNYC as any “geographic area that has experienced disproportionate negative impacts from environmental pollution due to historical and existing social inequities without equal protection and enforcement of environmental laws and regulations” [4].? EJ areas have not only historically been underfunded for risk monitoring systems, but they are often located in low-lying regions of the city, increasing the risk of severe flood damage when a flood occurs. As such, we recommend including these areas, as defined in the Environmental Justice Areas and Disadvantaged Communities [5] dataset, when evaluating supplemental gage locations in NYC.? ?

Proximity to critical infrastructure is another criterion to consider when selecting locations for water level monitoring sensors. Hospitals, utilities, and schools would benefit from a collocated water level monitoring sensor. Gages can also be sited to establish baseline water levels prior to watershed replenishment projects, such as restoring wetlands, to measure whether the project is having the intended impact over time. Gage placement will need to account for water-adjacent host availability in the areas of interest.

Variogram Results Applied to NYC

Based on the results of the variogram analysis, NYC would benefit from additional gages to ensure representative coverage along the coastline with a focus on areas susceptible to flooding. Using the identified ‘radius of coverage’ of 4 km, we superimposed the variogram range on the gage locations in NYC along with the intersection NYC’s flood evacuation zones [3] and environmental justice areas [5] (Figure 4a).

Major gaps exist between the coverage area of each gage, including the entire waterfront along the Hudson River north of the Battery, a stretch of more than 13 miles. Much of this stretch consists of both flood evacuation zones and environmental justice areas. The southern coastline of Brooklyn also lacks coverage.?

An ideal distribution of gages would minimize these gaps. Figure 4b shows an example deployment. The 11 supplemental gage locations recommended by Divirod and illustrated in Figure 4b are intended to accomplish three goals: 1) alleviate the current gaps in coverage, 2) provide additional data points in Hurricane Evacuation Zones, 3) provide additional data points to quantify flood risk in underserved communities.

Conclusion?

In this white paper, we demonstrated an approach for establishing a gage density threshold to support use cases where a correlation of time series data between adjacent and proximate coastal and inlet water gages is expected. To conduct this analysis, we used Divirod’s water database to identify available dense water level monitoring gage networks in geographic regions similar to NYC. This effort exposed major gaps in dense gage networks, likely due to the intended use, installation and maintenance costs of legacy water level monitoring gages. Legacy gages have previously limited the density of water data collection, which has far-reaching impacts on how society plans, regulates, adapts, and insures. While certain NYC-based groups like FloodNet, originally a grassroots collaboration between NYU, CUNY, and the Jamaica Bay Science and Resilience Institute, have improved water data coverage on roadways, additional work is needed for coastal and estuary monitoring in NYC.?Looking ahead, we expect that increasing the density of sensor networks while making data accessible and uniform will fundamentally change the way we manage water and water risks, not only in NYC but broadly across the globe.??

References?

[1] U.S. Geological Survey. (n.d.). Past, present, and future: USGS streamgages. USGS. Retrieved November, 2024, from https://www.usgs.gov/news/featured-story/past-present-and-future-usgs-streamgages?

[2] Divirod. (n.d.). Water level index methodology. Retrieved November, 2024, from https://wli.divirod.com/methodology.pdf?

[3] Esri. (n.d.). NYC Hurricane Evacuation Zone [Data set]. Retrieved November, 2024, from https://services5.arcgis.com/GfwWNkhOj9bNBqoJ/arcgis/rest/services/NYC_Hurricane_Evacuation_Zone/FeatureServer/0/query?

[4] City of New York. (n.d.). EJNYC: A study of environmental justice issues in New York City. Retrieved November 22, 2024, from https://climate.cityofnewyork.us/ejnyc-report/?

[5] Esri. (n.d.). Environmental Justice Areas and Disadvantaged Communities [Data set]. Retrieved November, 2024, from