Remote sensing of river temperatures in eastern Canada

Collaborators

Ms Anik Daigle, Cégep Garneau and project principal investigator

Gespe’gewa’gi Institute of Natural Understanding (GINU)

Direction principale de l’expertise sur la faune aquatique (MELCCFP)

Funding

Natural Sciences and Engineering Research Council of Canada (NSERC) Applied Research and Development (ARD) Program

Objectives

Water temperature in rivers and lakes is a determining variable in water quality and aquatic habitats, affecting dissolved oxygen, the toxicity of certain pollutants, the proliferation of invasive species and the metabolism of aquatic organisms, among other things.

Although the importance of monitoring the temperature of our rivers and lakes is now recognized, existing monitoring networks in Canada remain limited in time (recent) and space. As a result, our knowledge of thermal regimes is still largely fragmentary, with important consequences for our ability to characterize water quality and aquatic habitats, assess the duration of ice and ice-free seasons, their effects on spring freshet, on the phenology of aquatic species, on fishing, and so on. Without in-depth knowledge of these different processes, it is also difficult to anticipate the effect of global warming on each of them.

We propose here to substantially enrich the thermal portrait of Eastern Canada’s hydrographic network using satellite remote sensing.

Methodology

This project aims firstly to automate the identification of “water” pixels in Landsat images, and then to use the information extracted from these pixels to

1. Characterize the thermal regime at each pixel.
2. Assess its evolution over time. End-users will be able, via a user-friendly, public interface, to view and download temperature chronicles at their sites of interest, as well as several descriptors (e.g. mean thermal regime, trends, descriptive metrics of thermal habitat quality, etc.) corresponding to their needs in terms of aquatic habitat preservation and restoration.

Thermal infrared imagery measures the temperature of a surface by exploiting the link between temperature, spectral radiance and emissivity of the surface in question. This imagery is available on Google Earth Engine (GEE), but to meet the requirements of the project, thermal information should only be extracted from pixels that are 100% within a watercourse (so-called “sentinels”), so as to avoid false measurements and inducing noise in the thermal profiles.

Identifying “sentinel” pixels in Landsat thermal images

To do this, different sources of information are cross-referenced to assign a probability of “water” membership to each pixel likely to represent a stretch of water.

Longitudinal interpolation

Sentinel pixels, identified as “pure”, enable temperature measurements at several points on lakes and rivers. These measurements are typically spatially disjointed, as they are separated by river sections where pure pixels could not be identified. Interpolation is carried out along the river’s longitudinal axis using conventional methods (one-dimensional linear or cubic interpolations), or using regressions based on local characteristics likely to influence temperature (e.g. elevation, order, river km, width, orientation, slope).

Field validation

Some historical data acquired in medium to large rivers and included in the RivTemp (rivtemp.ca) and DataStream Atlantic (atlanticdatastream.ca) databases are used to validate the temperature values obtained by the sentinel pixels identified, and to assess their accuracy and potential bias.

Spatio-temporal characterization of thermal regimes

Various types of information are extracted, and some measured, from the thermal profiles derived from Landsat images, such as :

1. The probability of “water” belonging to the pixel
2. Temperature series available at this pixel
3. The mean interannual profile of the thermal regime (mean interannual daily temperature as a function of the day of the year)
4. A simplified three-parameter model of this thermal regime (parameters representing the annual maximum, the date of occurrence of this maximum and the duration of the warm season) and the temporal trends in these parameters.
5. Where statistically significant, the trend (°C/year) in mean annual and monthly temperatures
6. A list of 60 descriptive statistics of interest for assessing thermal habitat quality, including the thermal growth potential index for juvenile Atlantic salmon for rivers with a population of this species.

Setting up the distribution interface

GEE offers efficient online deployment options through an integrated solution called Apps Engine, which supports real-time interaction with data, enabling users to explore and analyze geospatial information dynamically, and access interactive map, graph and dashboard displays to present analysis results intuitively.

Equipments

• Google Earth Engine
• Python and Javascript programming
• Thermographs

Domains

Resource management

Environmental monitoring