Summary:
The U.S. Geological Survey (USGS) New York Water Science Center (NYWSC) operates a monitoring network of streamgages in New York collecting water-resources information, reported in near-real time, which can be used for computing long-term low-flow statistics. This study aims to determine low-flow and mean annual flow statistics for all continuous streamgages in New York State. Regional regression equations will be developed to estimate the computed statistics for ungaged locations in New York from the relation of computed statistics at gaged locations to various physical and climate characteristics of the drainage areas contributing flows. The equations and basin characteristics will be available at U.S. Geological Survey StreamStats for New York web application. There is currently no easy and reliable way to compute these low-flow statistics for ungaged locations in New York.
Reliable information about the magnitude, frequency, and duration of low-flows are critical for water-supply management; reservoir design; waste-load allocation; and the preservation of water quality and quantity for irrigation, recreation, and ecological conservation purposes. Low-flow statistics, including n-day flow frequency, are used by Federal, State, and local entities for setting regulatory standards and establishing water-quality and water-supply management goals, particularly during dry periods when demand often exceeds the supply of water (Ries, 2006). The New York State Department of Environmental Conservation (DEC) often uses the annual minimum 7-day average flow that likely will occur once every 10 years (7Q10), to establish water use permitting criteria for New York streams and to determine critical flow levels to maintain adequate water supply for human uses and ecosystem health. Regionalization estimation methods have been used in many States including Georgia (Gotvald, 2017), Iowa (Eash and Barnes, 2017), North Dakota (Williams-Sether and Gross, 2016) and Minnesota (Ziegeweid and others, 2015) due to the recognized importance of low-flow statistic and duration estimates to aid water resources management efforts during critical low-flow periods.
Problem:
Low-flow frequency and duration information is particularly important in regions and States that are periodically affected by drought, including New York. As part of their mission to protect public health and aquatic ecosystems, state agencies, such as the New York State Department of Environmental Conservation (DEC), need accurate and representative streamflow statistics to establish realistic and applicable criteria for both water quality and water quantity. Historically, low-flow statistics, such as the annual minimum 7-day average flow that likely will occur once every 10 years (7Q10), have been used by water-resource managers and planners as a threshold criterion for applying the chronic aquatic life criteria for regulatory measures such as determining waste-load allocations for point sources, total maximum daily loads (TMDL) for streams, and the quantity of water that can be safely withdrawn from a particular stream.
Because of the importance of these applications, it is critical to effectively measure and document base-flow data for use in updating low-flow frequency relations on a regular basis, preferably every 10 years, and especially after periods of extreme low flow (Feaster and Lee, 2017). Recent drought periods in New York from 2000-2019 (figure 1.), and currently during 2020, have heightened the need for pertinent low-flow information for State and local agencies to make critical water-resources decisions. It is prudent to update low flow statistics because changes may be occurring to low flows in New York over time (Suro and Gazoorian, 2011).
Though outside the scope of this project, trends at long-term unregulated streamgages may result from changes in climatic cycles, land use, changes in upstream diversions for water supply or waste treatment purposes, groundwater pumpage, or other practices that may affect groundwater levels (Weaver, 2015). Regulated streamgages, may also be influenced by changes in climatic cycles, land use, diversions, and groundÂwater activities, but those changes can be mitigated, enhanced, or even offset by changes in regulation patterns (Gotvald, 2016). Trend assessments in the flow patterns of regulated streamgages are also useful to help determine the suitability of frequency analysis (Feaster and Guimaraes, 2014).
Figure 1. Recent drought periods in New York (NIDIS, 2019)
Low-flow statistics in New York have not been updated statewide since 1975 (Eissler, 1978). Regression equations to estimate low-flow statistic values at ungaged locations in New York were made by Randall (2010) and Randall and Freehafer (2017), but only included data for the Susquehanna River Basin and the lower Hudson River Basin, respectively. Additionally, these studies did not use basin characteristics that could be incorporated into StreamStats for low-flow frequency statistics prediction equations to be applied for users.
To provide better documentation of flow characteristics for New York streams, the U.S. Geological Survey (USGS) proposes a cooperative two-phase investigation to (1) update low-flow statistics at selected active and discontinued continuous-record streamflow gaging stations, and (2) develop regional regression equations for selected low-flow characteristics for unaltered streams.
Objective and Scope:
Phase I
The objective of the first phase of this study is to perform a statistical low-flow analysis for active and discontinued continuous record streamflow gaging stations that have a minimum 10-year period of record of daily mean flows (approximately 300 streamgages). The scope of this phase will include regulated and unregulated streams in New York, except for Long Island, and hydrologically connected streams in adjoining states. Tidally influenced streams will be excluded. Priority will be given to stations at which at least 3 years of streamflow data were collected after 1990 and new stations with at least 10 years of available data. In addition, a flow-duration analysis will be completed, and selected exceedance percentiles will be provided.
Low-flow statistics will be calculated using consecutive minimum 1-day and 7-day average streamflows and will be provided for recurrence interval of 10 years. Additionally, annual mean flows will be computed for the selected gages.
Phase II
The objective of the second phase of this study will be to develop regional regression equations to estimate low-flow statistics at ungaged locations on unregulated streams in New York, excluding Long Island. Using data computed for unregulated continuous-record streamflow gaging stations during phase I of this investigation, regional regression equations will be developed to estimate the annual minimum 7-day average flow with a 10-year recurrence interval (7Q10) and the annual minimum 1-day average flow with a 10-year recurrence interval (1Q10). Once published, the regression equations will be publicly accessible on USGS StreamStats.
Historically, regional regression equations for low-flow statistics often resulted in large uncertainty in the estimates (Zalants, 1991; Lara, 1979) making them less useful than similar regression equations for flood-frequency estimates. However, availability and improvements in geographic information system (GIS) data layers along with improved regression methodologies have reduced the uncertainty in low-flow regression estimates and has made them more widely used (Gotvald, 2017, Eash and Barnes, 2012; Koltun and Kula, 2013; Martin and Arihood, 2010; Law and others, 2009; and Funkhouser and others, 2008).
Relevance & Benefits:
Low-flow statistics for New York have not been systematically updated for over 40 years (Eissler, 1978) and New York currently does not have regional regression equations for low-flow statistics that can be easily applied. This study will provide New York water-resource managers with more precise low-flow-frequency streamflow estimates at the USGS streamflow gaging stations. In addition, having regional regression equations for estimating low-flow statistics at ungaged locations will be of significant benefit where streamflow is not continuously monitored. The low-flow equations will be incorporated into the USGS New York StreamStats application providing efficiency and consistency with respect to applying the equations (http://streamstats.usgs.gov/ss). The information can be used by New York water-resource managers for planning, management, and permitting decisions to help ensure adequate water for consumptive use, water-quality standards, recreation, and aquatic habitat protection.
Low-flow frequency statistics provided in this report support USGS priority water-resource issues at both the National and State levels. By providing updated low-flow data and regional regression equations, the USGS will fulfill its mission of providing the hydrologic data and procedures necessary to increase the understanding of the Nation's water resources, advancing knowledge of the regional hydrologic system, advancing understanding of hydrologic processes, furnishing hydrologic data or information that contribute to protection of life and property, and providing standardized, quality-assured data to National data bases available to the public that will be used to advance the understanding of regional and temporal variations in hydrologic conditions. In addition, development of these statistics will address the priority water-resources issues of mitigating hydrologic hazards, enhancing hydrologic data-collection networks, and assisting states with data needed for TMDL development and drought preparedness (Evenson and others, 2013). This project will address the natural hazards strategic action under the long-term mission goals of the USGS to enhance understanding of the linkages among natural hazards, the environment, climate, and society, and the ways by which climate variability and change influence the frequency and intensity of natural-hazard events (U.S. Geological Survey, 2007). The study also addresses the USGS Cooperative Water Program priorities to support data-collection activities and to establish comprehensive, uniform, and accurate data on surface water, groundwater, water quality, sediment, and water use (U.S. Geological Survey, 2013).
Approach:
Phase I
Phase I of the study will cover a 1-year period beginning on or about October 1, 2022 and ending September 30, 2023. The Phase I results will be published in a USGS Scientific Investigation Report. In addition, the statistics will be included in USGS StreamStats making them easily accessible for the gaged locations and assisting in transferring the statistics to appropriate ungaged locations (Ries and others, 2017).
The statistical analyses of streamflow data in this study will be based on three categories of stations: (1) long-term record stations on unregulated streams; (2) short-term record stations on unregulated streams; and (3) regulated stations.
Typically, low-flow statistics are computed for streams if at least 10 years of continuous daily record are available; however, computing the statistics from long-term records is preferred because they are likely more representative of a broader range of hydrologic conditions. Thus, long-term streamflow data are better suited for trend assessments and statistical estimates. The USGS typically considers 30 years of streamflow record to designate long-term streamgages (U.S. Geological Survey, 2009). For stations with short-term records (those which have at least 10 years of record but less than 30 years), low-flow statistics can be improved by using record extensions or augmentation methods (Hirsch, 1982) based on correlations with long-term gages. This approach is particularly beneficial if the streamflow data at the short-term record station were collected during an unusually dry, wet, or otherwise unrepresentative period. For this study, streams for which multiple USGS streamgages are located and represent a mix of long-term and short-term gages, assessments will be made to determine if suitable correlations exist, and conditions warrant, for the record augmentation of the short-term gages.
There are three main steps involved in estimating low-flow frequency discharges at gaging stations using the log-Pearson Type III (LPIII) distribution method (Riggs, 1972).
1. The low-flow streamflow data for a station are retrieved, compiled, and quality assured.
2. The data are divided into climatic year periods of record (April 1 to March 31). A climatic year is a continuous 12-month period selected for the presentation of data relative to a hydrological or meteorological phenomenon of interest and is usually designated by the calendar year during which most of the 12 months occur (Langbein and Iseri, 1960).
3. The low-flow statistics are then computed, ranked, and analyzed for frequency of occurrence.
For this investigation, the frequency analysis of the streamflows will be made using LPIII distributions (figure 2; Riggs, 1972; U.S. Geological Survey, 1979). The frequency curves will be reviewed to verify that the LPIII distribution adequately fits the station’s data. Low-flow statistics will be computed for the annual minimum 1-day and 7-day flows for recurrence interval of 10 years.
For gaging stations on regulated streams, the low-flow characteristics also will be assessed for long-term trends. If the assessment shows that the regulation patterns have been reasonably consistent and the LPIII distribution provides a reasonable fit of the data, low-flow statistics will be computed for that period using similar techniques for the unregulated streamflow gaging stations (Riggs, 1972). In cases where regulation patterns are shown to be highly variable and (or) where the LPIII distribution does not reasonably represent the data, tables of exceedance percentiles for consecutive 7-day average flows will be generated in place of a frequency analysis for the stations. These exceedance percentiles should not be construed to be a representation of a low-flow frequency, but as an empirical representation of recorded 7-day average flows. The data will be useful for assessing the flow conditions for the period of record.
The techniques described previously for augmenting short-term records are only applicable to natural, unaltered streamflows and therefore, will not be applied to regulated streams. In addition, the low-flow statistics for regulated streams are relevant to similar future regulation patterns and would not be applicable if the future regulation patterns are significantly different than for the period analyzed.
Figure 2. Log-Pearson type III distributions of 7-day minimum flows for two USGS streamgages in New York.
Phase II
Phase II of the study will cover a 1-year period beginning on or about January 1, 2024 and ending September 30, 2025. This phase of the investigation will include the development of regional regression equations to estimate the annual minimum 7-day average flow with a 10-year recurrence interval (7Q10), and the annual minimum 1-day average flow with a 10-year recurrence interval (1Q10) at ungaged locations on unregulated streams. For the stations in Phase I that are located in unregulated basins, a set of explanatory (independent) variables will be developed using GIS data layers for basin characteristics such as drainage area, slope, elevation, and percent impervious area, various geology and soils indices, hydrologic characteristics such as streamflow variability or base-flow recession indices, and climatic characteristics, such as mean annual precipitation (Smakhtin, 2001; Eash and Barnes, 2012).
The explanatory variables will then be tested in the regression analysis using a combination of ordinary least squares (OLS), weighted least squares (WLS) and generalized least squares (GLS) regression techniques. The OLS techniques will be used to select the initial set of explanatory variables and to determine the regionalization appropriate for the state. The WLS and (or) GLS techniques will be used to compute the final regression coefficients and to compute the accuracy statistics for the final equations (Eng and others, 2009). In addition, for regions where zero flows are a possibility, logistic regression will be used to develop a preliminary equation for determining the probability of zero flow (Funkhouser and others, 2008). Once the regional regression equations are published, the USGS StreamStats application will be updated to include the capability of generating the explanatory variables needed for the low- flow regression equations. The low-flow regression equations will be tested for accuracy and incorporated into StreamStats. The estimated low-flow statistics will be available to anyone using the USGS StreamStats web application.
