Niagara River Pollution 1986-96


Data Source: 

Data Interpretation Group


Abdel El-Shaarawi of the National Water Research Institute, Canada Centre for Inland Waters



A water quality monitoring program involving the collection of ambient water and suspended solids samples at the head (Fort Erie) and mouth (Niagara-on-the-Lake) of the Niagara River has been undertaken by Canada, the United States, New York State and the Province of Ontario in accordance with the Niagara River Toxics Management Plan and the Declaration of Intent (1987). The program is operated by Environment Canada and uses sampling and analytical methods agreed upon by all four parties.

Monitoring data were interpreted jointly by a group of scientists representing Environment Canada, the United States Environmental Protection AgencyOntario Ministry of the Environment and New York State Department of Environmental Conservation, known as the Data Interpretation Group. Their annual reports have the consensus of all members. The information given here has been excerpted from the report for 1988-89 [1].

Station Location and Sampling Protocol

The Data Interpretation Group assumes that the locations of the Fort Erie and Niagara-on-the-Lake stations are adequate to monitor, respectively, chemical inputs to the Niagara River from the Eastern basin of Lake Erie and chemical outputs from the river into Lake Ontario.

Site selection was also dictated by operational considerations such as acquisition and availability of property, accessibility, availability of electrical power, and site security. Since there was no single best site, sites were chosen to serve as the best possible compromise between acceptable representativeness and operational constraints.

The sample collection system at both locations consists of an intake structure, intake line, pump, continuous-flow centrifuge, and Goulden Large Sample Extractor (GLSE). The intake structure, intake line and pump are permanently submerged in the river while the pump controllers, centrifuge and GLSE are housed on the river bank. Intakes were positioned in the water column such that they were sufficiently far from the bottom to avoid sampling bedload yet far enough below the surface so as not to constitute a hazard to navigation.

Generally, sample collection at Fort Erie commenced on Tuesday afternoon of each week while sample collection at Niagara-on-the Lake was started on Wednesdays. The delay of 15 to 18 hours between sample collection at Fort Erie and Niagara-on-the Lake was introduced to account for the time required for water to travel the length of the river from Fort Erie to Niagara-on-the-Lake. Staggered sampling times could not account for storage and release of water from the Robert Moses and Sir Adam Beck power plant reservoirs. However, the introduction of time-delayed sampling provides a closer approximation of the river's hydrologic regime.

The 1988-89 report of the Data Interpretation Group [1] gives a map, diagrams illustrating the configuration of the sampling stations, and flow charts for the sampling procedures and chemical analyses and preprocessing of the data.

Research Question: 

The objective is to see if the reduction in pollution targeted in the Declaration of Intent (1987) has been achieved. While this can be viewed as testing for trend in a multivariate time series, there are issues of calibration, upper and lower detection limits, missing data, changes in measurement techniques over time, seasonal fluctuations, point sources of pollution, and unevenly spaced observations.


For the Case Studies in Data Analysis, only the following variables have been included in the data file. Missing values are indicated by "NA". There are 965 observations taken more or less weekly at each station from April 1986 to March 1996.



STATION FE = Fort Erie, NOTL = Niagara-on-the-Lake
DATE Calendar Date
DISCHARGE Flow in cfs
SEDIMENT Sediment concentration, mg/l
DIE_H2O Dieldrin dissolved in water, ng/l.
DIE_H2O_DL An L in this column indicates that the preceding value was censored at the upper or lower limit of detection.
DIE_SOL Dieldrin in solids, ng/g.
DIE_SOL_DL An L in this column indicates that the preceding value was censored at the upper or lower limit of detection.
PCB_H2O Polychlorinated biphenyl (PCB) dissolved in water, ng/l.
PCB_H2O_DL An L in this column indicates that the preceding value was censored at the upper or lower limit of detection.
PCB_SOL PCB in solids, ng/g.
PCB_SOL_DL An L in this column indicates that the preceding value was censored at the upper or lower limit of detection.


Data Access: 

To get the data, as a plain text file or as an Excel worksheet, click here. If you have difficulty downloading these files, contact Peter Macdonald at

Some Useful Links



Veuillez contacter Dr El-Shaarawi si vous avez besoin de photocopies de l'un de ces documents.

[1] Data Interpretation Group (1990). Joint Evaluation of Upstream/Downstream Niagara River Monitoring Data 1988-1989. A joint publication of EC, USEPA, MOE, and NYSDEC.

[2] El-Shaarawi, A.H. (1989). Inferences about the mean from censored water quality data. Water Resources Research 25(4): 685-690.

[3] El-Shaarawi, A.H. & A. Naderi (1991). An approximation to the likelihood function with application to environmental data. Environmetrics 2(2): 227-240.

[4] Yip, Wai-Siu Jason (1989). Inferences from Censored Data with Applications to Water Quality. Unpublished M.Sc. Project, Graduate Program in Statistics, McMaster University. iv + 65 pp.

[5] Yu, Yongmin (1998). Regression under Transformation for Censored Data with Application to Water Quality.Unpublished M.Sc. Project, Graduate Program in Statistics, McMaster University. v + 59 pp.