OBSERVATIONS OF PHYTOPLANKTON DYNAMICS IN LAKE ONTARIO USING ATSR-2 IMAGERY

ABSTRACT

ATSR-2 gridded brightness temperature data have been used to estimate chlorophyll concentrations in Lake Ontario, North America. Lake truth data were collected during July and October 1995 as part of the Lake Ontario Trophic Transfer (LOTT) project. Epilimnetic chlorophyll a concentration was estimated using in situ fluorometer apparatus and traditional solvent extraction techniques. Nadir and forward looking pixel data were used to derive a chlorophyll estimation algorithm based on a multiple regression of 555, 670, 870 nm and 1.6 m wavelengths and fluorometer values (r²>0.87, p<0.01). ATSR-2 scenes from July 1995 were processed using the algorithm to derive lake wide chlorophyll concentration and distribution. Phytoplankton production generally followed the pattern of surface current circulation, with the highest levels recorded in the south western part of the lake. Images showed chlorophyll distributions changing more rapidly than could be detected by the tri-annual LOTT cruises. Further work incorporating ATSR-2 lake surface temperature maps is planned.

1. INTRODUCTION

Lake Ontario is the fourth largest of the Laurentian Great Lakes with a mean volume of 1635 km³ and a maximum depth of 245 m. The lake is located on the US - Canadian border between 76 and 80^oW, and 43 and 44^oN. The lake forms part of a valuable shipping route between cities including Toronto and Chicago and the St Lawrence sea way through to the Atlantic. The commercially important fisheries in the lake have historically been based on the yellow perch (Perca flavescens). Colonisation by alewife (Alosa pseudoharengus), which were first discovered in 1873 and rainbow smelt (Osmerus mordax) have accelerated the collapse of yellow perch stocks. In the mid-1960s , stocking of native and introduced predatory fishes (salmon and trout) was begun to reduce alewife and rainbow smelt and develop economically important sport fisheries (Mills et al., 1995).

The lake has been classified as mesotrophic being more productive than the oligotrophic lakes Superior, Huron and Michigan and less productive than the shallow eutrophic Lake Erie, located upstream (Beeton, 1965). Lake wide water quality surveys have been carried out tri-annualy since the early 1960s with measurements for Chl a, dissolved organic carbon, dissolved oxygen, suspended sediment, light attenuation and temperature profile being recorded at over 40 stations on six north south transects. The phytoplankton are dominated by the Diatomae, Chlorophyta and Cryptophyceae. Epilimnetic phytoplankton biomass vary between averages of 1 and 5.3 g l^-1, for spring and summer respectively (Munawar et al., 1987).

Bukata and Bruton (1974) first used ERTS-1 (Landsat 1) imagery estimating Chlorophyll a (Chl a)and suspended solid concentrations in different regions of the lake. This work has been followed up in a series of papers illustrating the use of optical cross sections in association with high spectral resolution satellite (Coastal Zone Colour Scanner) or airborne (Airborne Thematic Mapper) imagery for the estimation of water quality components (Bukata et al 1981; 1991 and Jerome et al., 1994). Since the failure of the Coastal Zone Colour Scanner satellite observations of the Great Lakes have been limited to optical observations using TM and SPOT data (Lathrop & Lillesand, 1986; 1989) and thermal observations using AVHRR imagery (Bolgrien & Brooks, 1992).

Bukata's method of modelling optical cross sections has the potential for accurate chlorophyll estimations if a suitable sensor were available. This will not however be the case until the successful launch of SeaWiFS some time in 1997. High spatial resolution Landsat TM and SPOT are also capable of producing accurate biomass estimates for freshwater systems, however they lack the synoptic cover and temporal resolution for monitoring intraseasonal variability in large lakes such as Ontario.

ATSR-2 has forward and nadir viewing visible wavelengths which correspond to Green (545-565 nm) and Red (649-669 nm) colours of the spectrum. In addition to these it has near- (855-875 nm), mid- (1.58-1.64 & 3.55-3.93 m) and thermal infrared (10.4-11.3 & 11.5-12.5 m) wavelengths. The thermal wavelengths are suitable for accurate calculation of sea surface temperatures (Barton et al., 1989 & Gohil et al., 1994) but there has been no investigation into the use of the visible channels for chlorophyll biomass estimation.

ATSR-2 has a 512 km swath width, a three day repeat cycle over the Laurentian Great Lakes and 1 km pixel size, providing adequate synoptic cover, temporal and spatial resolution for phytoplankton observations in Lake Ontario. This paper reports on the initial investigations into how ATSR-2 visible and near-infrared wavelengths can be used to estimate phytoplankton concentration in Lake Ontario and illustrates the heterogeneity of phytoplankton distributions over relatively short time periods.

2. METHODOLOGY

2.1 Water quality data collection

Lake Ontario is surveyed three times a year; in the spring, summer and autumn. Data are recorded on water quality, benthic and pelagic fauna as part of the Lake Ontario Trophic Transfer Study. Data are collected from 41 stations located along 6 transects spaced equally throughout the lake. The summer survey was carried out between the 10th and 26th of July in 1995.

Chlorophyll concentration data were collected from within the epilimnion using an integrated sampling bottle, after location of the thermocline using multiple CTD casts. In addition, on the 21st (Fig. 2.2) and 22nd of July, over 120 km of cruise track were sampled using a ship-borne fluorometer, in a continuous lake water flow arrangement (Fig. 2.1). Only 70 km of the transect data could be used due to missing sensor scans over the northern region of the lake from the forward look angle of the 15th and 21st of July images.

Figure 2.1. Fluorometer in flow through vessel as used for transects.

Water was pumped from a depth of 2 m below the surface, at a rate of 50 l min^-1. The flow was passed from the bottom through a vessel containing the submerged fluorometer and allowed to flow out of the top. Residence time in the vessel was estimated at being less than 3 minutes. Fluorometer measurements were recorded every 5 seconds and down loaded to a computer. Water flow was dictated by speed of the ship (average 11 knots). Transects were truncated to allow for the speeding up and slowing down of the ship at the ends and turns of the transect. The fluorometer was calibrated with known concentrations Chl a before and after the transect (r² = 0.97, p = <0.01).

2.2 Image collection and processing

Three ATSR-2 gridded brightness temperature scenes of Lake Ontario for the 15th, 18th and 21st of July were acquired from The Rutherford Appleton Laboratory. Images were atmospherically corrected using a 5S procedure to top of the atmosphere reflection values and were geometrically corrected using geolocation data within the products (RMS <0.3 for all images). A crude water mask using the 11 m wavelength was used before extraction of pixel values from transect lines and individual LOTT stations.

Fluorometer data were averaged to produce mean 1 km values for correlation with the extracted reflectance values. Averaged fluorometer values were correlated with different combinations of forward and nadir looking bands.

Figure 2.2. Location of the 120 km of fluorometer cruise track recorded on 21st July 1995

3. RESULTS & DISCUSSION

3.1 Wavelength and chlorophyll concentration relationships

Table 3.1 shows the results for the regressions of fluorometer transect and LOTT station chlorophyll values against image pixel data. All band combinations tested showed significant relationships. There were better correlations between forward looking wavelengths and chlorophyll values than between nadir wavelengths and chlorophyll values. The best relationship was exhibited using both nadir and forward looking wavelengths. Figure 3.1 illustrates the low variability in the relationship over the range of chlorophyll values observed.

Table 3.1. R values for different band combination relationships.

The concept of the forward looking wavelengths being better correlated with chlorophyll concentration is difficult to explain. There are however two ideas that have been suggested. The first is that the larger pixel view of the forward look is on a more similar scale to the natural spatial variability of phytoplankton patches. A second possible explanation is that the solar elevation at this latitude and variability in lake surface roughness produces a range of reflectance angles, more strongly detected at the ATSR-2 forward look angle. Further work is required to investigate this feature.

Figure 3.1. Plot of predicted versus recorded Chl a concentrations (g l^-1) for fluorometer transect data.

Figure 3.2. Processed ATSR-2 GBT images from the (a) 15th, (b) 18th and (c) 21st of July 1995. Chlorophyll estimation derived from a multiple regression relationship incorporating the 555, 670 & 870 nm and 1.6 m wavelengths from both the forward and nadir views. Each image section is approximately 300 km wide.

3.2 Phytoplankton distribution

Figure 3.2 shows the ATSR-2 derived chlorophyll maps for the 15th, 18th and 21st of July. There are marked differences over the 6 day period.

On the 15th of July the lake appears virtually completely mixed, with chlorophyll concentrations of between 1.5 and 2.5 g l^-1 across the lake. There is a tongue of elevated chlorophyll concentration in the far west of the lake, spreading from the Hamilton harbour area, approximately 70 km towards the centre of the lake. There are also small elevations in phytoplankton concentrations originating at the mouths of the Niagara (~15 km) and Genesee (~5 km) rivers on the southern shore. There is a small area of lower productivity in the centre more deeper region of the lake.

The 18th of July image is characterised by high levels of cloud fringing the lake on both the north and south shores. This cloud has resulted in large areas of the western end of the lake being obscured. There is an average increase in chlorophyll concentration across the lake. There appears to be an increase in the western arm tongue observed on the 15th spreading out towards the centre of the lake. The plume emanating from the Niagara river also appears to have increased in size with the Genesee plume obscured by cloud. There is an increase in chlorophyll concentration in the eastern arm on the lake, with lower levels in the centre and in the entrance to the St Lawrence river.

The image from the 21st of July is incomplete due to forward look angle data being omitted from the northern shore of the lake. This image does however, show the greatest heterogeneity of chlorophyll concentration across the lake. There is a plume emanating from the Genesee river mouth which appears to move westwards along the southern shore towards the mouth of the Niagara river. A similar increase in chlorophyll concentration is seen originating from the Niagara river mouth moving north eastwards towards the centre of the lake. These plumes appear most concentrated in two regions (~3 g l^-1) approximately 15 km north of the south shore.

Immediately to the north of these regions is an area of elevated concentration scrolling northwards and eastwards towards the centre of the lake. The central northern region shows the lowest phytoplankton concentration of approximately 2.5 g l^-1. There are higher concentrations with no apparent structure in the eastern end of the lake.

The clearly defined eddy structure in the mid-western part of the lake in the 21st of July image appears to mirror the surface circulation as described by Harrington (1894). The water inflows from the Niagara and Genesee rivers appear to be important factors in determining the distribution of phytoplankton in the western end of the lake. This may be as a result of river outflow driven circulation or as a result of elevated nutrient levels in river water.

4. CONCLUSIONS

It is evident from this short study, that the phytoplankton distribution in Lake Ontario is more dynamic than can be detected during a triannual sampling regime. Changes over the six day period of these images illustrate changes in both concentration and distribution of phytoplankton over the whole lake.

Satellite derived chlorophyll maps cannot show what is causality, but they can be used as indicators of anomalous phytoplankton patterns. Further work is required to validate these wavelength/water quality parameter relationships for other large lakes and time periods. These data can then be used in conjunction with surface temperature maps, and radar derived products to produce integrated water quality management tools.

This study would not have been possible without the support of the scientists and crew associated with the Lake Ontario Trophic Transfer study at the Canada Centre for Inland Waters, Burlington Ontario. Specific thanks goes out to Mr. C. Wenghofer, Dr. M. Munawar, Mr. S. Millard, Ms. D. Myles and W.G. Sprules (University of Toronto). Thanks to Dr. G. Mackay for providing the code for the atmospheric correction.

This work was completed while GMJ was in receipt of NERC studentship GT12/94/ATSR2/6.

4. REFERENCES

BARTON, I.J., ZAVODY, A.M., O'BRIEN, D.M., CUTTEN, D.R., SAUNDERS, R.W., LLEWLYN-JONES, D.T., (1989) Theoretical algorithms for satellite-derived sea surface temperatures. Journal of Geophysical Research, 94, 3365-3375.

BEETON, A.M., (1965) Eutrophication of the St. Lawrence Great Lakes. Limnology and Oceanography, 10, 240-254.

BOLGRIEN, D.W., and BROOKS, A.S., (1992) Analysis of thermal features of Lake Michigan from AVHRR satellite images. Journal of Great Lakes Research, 18, 259-226.

BUKATA, R.P., and BRUTON, J.E., (1974) ERTS-1 digital classification of the water regimes comprising Lake Ontario. Proceedings of the Second Canadian Symposium on Remote Sensing, University of Guelph, Guelph, Ontario, April 29 - May 1, 1974. pp 627-631.

BUKATA, R.P., JEROME, J.H., and BRUTON, J.E., (1991) Validation of a five-component optical model for estimating chlorophyll a and suspended mineral concentrations in Lake Ontario. Applied Optics, 20, 3472-3474.

BUKATA, R.P., JEROME, J.H., KONDRATYEV, K.K., and. POZDNYAKOV, D.V., (1981) Estimation of organic and inorganic matter in inland waters: optical cross sections of Lakes Ontario and Ladoga. Journal of Great Lakes Research, 17, 461-469.

GOHIL, B.S., MATHUR, A.K., and PANDY, P.C., (1994) An algorithm for sea surface temperature estimation from ERS-1 ATSR using moisture dependent coefficients: a simulation study. International Journal of Remote Sensing, 15, 1161-1167.

HARRINGTON, M.W., (1894) Currents of the Great Lakes as deduced from the movements of bottled papers during the seasons of 1892 and 1893. U.S. Weather Bureau, Washington, D.C., 20 pp.

JEROME, J.H., BUKATA, R.P., WHITFIELD, P.H., and ROUSSEAU, N., (1994) Colours of natural waters: 1. factors controlling the dominant wavelength. Northwest Science, 68, 43-52.

MILLS, E.L., O'GORMAN, R., ROSEMAN, E.F., ADAMS, C., and OWENS, R. W., (1995) Planktivory by alewife (Alosa pseudoharengus) and rainbow smelt (Osmerus mordax) on microcrustacean zooplankton and dreissenid (Bivalvia: Dreissenidae) veligers in souther Lake Ontario. Canadian Journal of Fisheries and Aquatic Sciences, 52, 925-935.

MUNAWAR, M., MUNAWAR, I.F., and McCARTHY, L.H., (1987) Phytoplankton ecology of large eutrophic and oligotrophic lakes of North America: Lakes Ontario and Superior. Proceedings of an international symposiu