Development of a model of the uncertainty associated with the traditional measurement of water surface profiles during floods for rivers in southern Québec

Following the flooding that occurred in the spring of 2017, the Quebec government began examining the management of flood hazards throughout the province in a context of climate change. Those undertaking this work recognized the need for updated flood zone mapping to better define the risks associated with flooding. Flood risk modelling and mapping requires a thorough understanding of the uncertainty in the water level data used to calibrate river hydraulic models.

Traditionally, water level measurements are made using a Global Navigation Satellite System (GNSS) at the river’s edge. However, there are several sources of uncertainty that can affect the quality of the water level measurement. The detailed knowledge of the sources of uncertainties and their quantification makes it possible to define calibration targets for the hydraulic models.

• Identify sources of uncertainty associated with traditional water surface profile measurements using real-time kinematic positioning (GNSS-RTK).

• Assess and quantify sources of uncertainty through an exploratory field campaign to target the greatest sources of uncertainty.

• Develop a sampling plan to cover the main sources of  uncertainty in water level measurement by means of traditional and bathymetric buoy methods.

• Develop and validate an uncertainty model that provides the distribution of the probabilities of the true water level based on measured water levels. Provide recommendations to reduce sources of uncertainty in traditional water level measurements.

• Multiple measurement campaigns during spring floods, with a variety of measurement sites, field conditions and operators

• Repetition by varying one source of uncertainty in a controlled manner

• Semi-controlled repetitions by taking advantage of the natural variation of the sources of uncertainty related to the measurement sites

• Uncontrolled repetitions based on the precise context at the time of measurement

• Creation of an uncertainty model using field data

The data collected at about 15 sites in the Eastern Townships and in Outaouais covered a range of measurement conditions common in southern Quebec. This data set was used to develop four uncertainty models that can be used to estimate the confidence interval of any water level measurement. The choice of model depends on the level of information accompanying the water level measurement. It has been shown that several sources of uncertainty related to environmental conditions (vegetation, structures, entrenchment, etc.), which affect the quality of satellite reception, are fairly well reflected in the calculation of the vertical error performed by the GNSS itself. 

Therefore, this vertical positioning error estimated by the device should be included in the calculation of the uncertainty of the water level measurements and should be systematically transmitted by agents. The other two significant sources of uncertainty in the proposed models are operator error and error related to GNSS range. The analysis of the data set has made it possible to define four uncertainty models that can be used depending on the level of information available:

1. The first and most general model requires only an elevation measurement.

2. The second model is based on the estimated VRMS at the time of the elevation measurement.

3. The third model differentiates the uncertainty using the VRMS associated with the elevation measurement and the GNSS range.

4. The fourth proposed model defines the uncertainty associated with a given site by collecting a series of measurement repetitions.

Following the various analyses of the data set, recommendations were made to improve the measurement protocol transmitted to MELCC agents who collect and transmit water level measurements. These recommendations are intended to minimize the uncertainty associated with water level measurement.