5 Case study results

There has been a strong emphasis on the investigation of synergistic relationships between the observations from different instrument in the Clean Seas project. It was not possible to plan in advance what would appear once all three classes of sensor were finally operating together and therefore a flexible approach was taken to respond to opportunities noted in the data acquired on a routine basis. This work has provided the results presented in the following sections and demonstrates the advantages of this flexible and responsive approach.

5.1 Multi-sensor Algal Bloom monitoring

Data from four different satellite sensors have been used to test the possibilities of a multi-sensor algal bloom monitoring approach. The satellite scenes were acquired within less than two hours over the same sea area during a cyanobacterial (blue-green algae) bloom. The data includes: the Thematic Mapper (TM) on Landsat-5, the synthetic aperture radar (SAR) aboard the Second European Remote Sensing Satellite (ERS-2), the Wide Field Scanner (WiFS) aboard the Indian IRS-1C satellite, and the Advanced Very High Resolution Radiometer (AVHRR) flown on the NOAA-14 satellite. Using this extensive data set, for the first time, an investigation of the usefulness of the fusion of different remote sensing data (particularly by incorporating radar data) for monitoring of an ongoing algal bloom has been performed. The results of the analysis show that the fusion of optical and microwave remote sensing data (even with different spatial resolutions) can yield more information about the biological and oceanic characteristics of sea surface areas.

The summer (July - August) blooms in the open Baltic Sea are often dominated by the species Nodularia spumigena and Aphanizomenon sp. During the later phase of the bloom, algae start to flocculate and accumulate on the sea surface in large quantities, thereby becoming visible even with non-optimal satellite sensors. The possibility to detect the surface accumulations by a satellite sensor is most often limited to the visible spectral bands, but in areas of intensive surface accumulation, the near infrared band can also be used. The surface accumulations of cyanobacteria can in some cases cause local increases in the satellite-derived sea surface temperature (SST) due to increased absorption of sunlight by higher phytoplankton pigment concentration (Kahru et al. 1993).

The use of satellite data, such as the Advanced Very High Resolution Radiometer (AVHRR), is limited to cloud-free days because of the used optical and infrared bands, and to large-scale patterns due to its low spatial resolution. Moreover, the spectral resolution is also not optimal for algae detection and monitoring. Nevertheless, for long time-series comparison of the dynamics of algae blooms, and near real-time monitoring of ongoing blooms, AVHRR has proven to be a useful data source, due to its high temporal resolution.

Recently, a large variety of different sensors have been operational, thus allowing for synoptic studies concerning the monitoring of the same oceanic (and atmospheric) phenomena, such as ongoing algae blooms. Therefore a search of the satellite data archives has been carried out in order to find a data set of satellite images acquired by different sensors within a short time period. Analysis of such data sets may provide a better understanding of oceanic (and atmospheric) phenomena and their imaging by the different sensors working at different electromagnetic frequencies.

On July 15, 1997, a day with extensive cyanobacterial blooms in large parts of the Baltic Sea, especially the northern Baltic Proper, data from four different satellite sensors have been acquired over the same sea surface area. Details about the different sensors, the satellite platforms, spatial resolutions (pixel sizes), and acquisition times are given in Table 5-1.

The WiFS sensor, with only two spectral bands, one visible (0.62-0.68 µm) and one near infrared (0.77-0.86 µm) also identifies most of the surface patterns in the area. The resolution of 188 metres makes it an interesting and useful contribution between the high (TM) and low (AVHRR) resolution imagery.

The AVHRR/2 sensor uses five spectral bands (0.63 to 12.5 µm), but because of its low spatial resolution (see Table 5-1) it is only capable of registering meso- and large-scale surface patterns. AVHRR data have been thoroughly used for monitoring algae blooms such as the cyanobacteria accumulations in the Baltic Sea (Kahru et al. 1994).

ERS-SAR images are two-dimensional maps of the radar backscattering at 5.3 GHz (C-band, 5.7 cm), here from the ocean surface. This backscattering is caused by small-scale surface waves, which have wavelengths comparable to the radar wavelength (7.2 cm; Bragg waves). These waves are often damped by the surface films produced by the (blooming) algae, so that the natural slicks, apart from other oceanic and atmospheric phenomena, are visible on SAR images as dark irregular patches (Gade et al. 1998). The advantages of SAR sensors are their independence of daytime and weather conditions and their high spatial resolution.

All satellite images were geo-referenced to the same projection (UTM, zone 34) with the same corner co-ordinates as the TM scene, the smallest of the four satellite scenes. The total RMS error in the transformation between the images is about 0.5 pixels for the optical sensors and 1 pixel for the SAR scene.

Apart from satellite imagery, bathymetry data, wind, and modelled current data have also been available for the analysis; however, in this work the focus has been on the comparison of the satellite imagery.

Figure 5-1. Composite AVHRR image (visible, near and thermal infrared as RGB) taken on July 15, 1997 at 11:01 UTC over the northern Baltic Proper between Sweden (left) and Finland (upper right). The image dimensions are 300km by 300km and the white rectangle denotes the locationof the subsections shown in Figure 5-1.

The bright patches in the centre of the AVHRR composite shown in Figure 5-1 are due to algae accumulations, but also due to some clouds and contrails (see the elongated lines).

In order to study the imaging of the ongoing algae bloom by the different sensors we have chosen a sub-section marked by the rectangle in Figure 5-1. The corresponding parts of the WiFS band 2, SAR, and TM band 4 images are shown in Figure 5-2.

Figure 5-2. Subsections (29km x 45km) of the TM band 4 (left), SAR (middle), and WiDS band 2 (right) images (the location is denoted in Figure 5-1). All images were acquired within 2 hours on July 15, 1997, and were resampled to a pixel size of 50m to allow for better comparison. The algae bloom accumulations are visible in all images. The dashed line corresponds to the profiles shown in Figure 5-3.

Figure 5-3. The three graphs show the relative pixel values for the TM band 4 (top, red), SAR (center, green), and WiFS band 2 (bottom, blue graph) along the profiles indicated in Figure 5-2. Low values in the SAR image, due to the dampening of the small-scale surface waves, correlates well with the peaks in the TM band 4 and WiFS band 2.

Cyanobacteria accumulations occur in large quantities in most of the investigated area. The spatial variations that exist give rise to the typical patterns seen in each of the panels in Figure 5-2. These patterns show up in all visible and near-infrared images, with the most obvious patterns in the visible bands (not shown herein). The reason for this is the ability of the TM 2 sensor (0.52-0.60 µm) to detect sub-surface algae. In the near infrared band (TM4) only algae at the surface are visible, and can therefore be detected only in some areas.

The widespread cyanobacterial accumulation that can be seen in the near infrared indicates that in those areas the accumulations are very close to or at the sea surface. Most surface patterns are clearly visible in both the high-resolution (30 metres) TM scene (left panel of Figure 5-2) and the medium-resolution (188 metres) WiFS scene (right panel). The densest accumulations are also visible in the AVHRR imagery (Figure 5-1), but in less detail compared to imagery with higher spatial resolution. The dark signatures visible in the SAR image (middle panel of Figure 5-2), which are usually attributed to the occurrence of slicks on the sea surface, are very well correlated with the high reflectance areas seen in TM 4. This indicates that the information from the SAR sensor correlates with the presence of the cyanobacteria seen on the sea surface.

We have calculated profiles along the dashed lines included into each panel of Figure 5-2. The results are shown in Figure 5-3, where the upper (red) curve corresponds to the relative pixel values of the TM band 4 image, the middle (green) curve corresponds to the pixel values of the SAR image, and the bottom (blue) curve corresponds to the relative pixel values of the WiFS band 2 image. Note that the middle curve (SAR) is plotted in logarithmic scales for better visualisation.

As can be seen from the profiles in Figure 5-3, there is an obvious correlation between low values in the SAR image and peaks in the TM band 4 and WiFS band 2. At a relative distance of about 15 km the single spike in the WiFS band 2 (bottom curve) corresponds very well with the singular drop of the SAR image intensity (middle curve) and with a respective peak in the TM band 4 (upper curve). A similar good correlation can be found at distances of about 18 km and about 40 km, where all sensors detected a triple spike. This good correlation is a strong indicator that both the SAR and the near infrared sensors detected cyanobacterial accumulations at the very surface of the sea.

Due to surface currents in the area there is a small displacement of the surface features identified in the various satellite images. From this displacement, and from the time periods between the image acquisitions, mean current velocities between 10 cms^-1 and 30 cms^-1 were calculated. From the results we infer that the mean surface current along the scan line in the southern (lower) part of the panels in Figure 5-2 was lower at the time between the first and second image acquisition (14 cms^-1, derived from TM and SAR data, respectively), than at the time between the second and third image acquisition (20 cms^-1, derived from SAR and WiFS data, respectively). However, by comparing the right-handed parts of the profiles shown in Figure 5-3 (corresponding to the northern (top) parts of the panels in Figure 5-2) we inferred that the mean surface current along the scan line in this particular area was changing from 10 cms^-1 in southerly direction (derived from TM and SAR data) to about 30 cms^-1 in northerly direction (derived from SAR and WiFS data). We attribute this finding to the fact that at the time of the image acquisitions an inflow of cold water from the Bothnian Sea in the north into the northern Baltic Proper in the south occurred. The front line of this inflow passes through the image subsections of Figure 5-2, thus causing an overall turbulent current field in that particular area.

In order to better demonstrate the spatial displacement with time of the observed features, we have generated a three-colour composite shown in Figure 5-4 by using the TM band 4 (red channel), SAR (green channel), and WiFS band 2 (blue channel) images thereby denoting the time order of acquisition.

Figure 5-4. Geo-referenced three-colour composite based on TM band 4 (red), SAR (green), and WiFS band 2 (blue) images. For better visibility the SAR data has been inverted so that areas of reduced radar backscatter appear in bright green. Image dimensions are 90km x 90km.

The image dimension is 90 km x 90 km, and all data have been re-sampled to a pixel size of 50 m. Note that for a better visualisation of the interesting features we have inverted the SAR image so that areas of reduced radar backscatter appear as bright green patches. The oceanic front is clearly visible in the SAR image (see the bright green line reaching towards the upper left image corner), and the observed displacements can best be seen in the image centre, north of the oceanic front. Contrails visible in the WiFS band 2 image can be seen as elongated blue lines in the upper right image corner.

The obvious 'front' in the middle of the image (reaching from north-west to south-east) is visible in all investigated images and spectral bands. The homogeneous area to the left (west) of this front is due to somewhat colder and cyanobacteria-free water flowing from the Bothnian Sea southwards into the Baltic Proper. An area of high surface concentration of algae, which is likely to coincide with a locally low wind field, causes strong reduction of the radar backscattering (visible as a bright green patch in the image centre).

In these results of the synoptic study of the imaging of an ongoing cyanobacterial bloom by different optical and microwave sensors, it turned out that the subsurface algae accumulations were best detected by the visual TM band 2 (not shown herein), but that the best correlation between the different imagery was between the SAR and the near infrared images. These findings reflect the fact that both sensors (radar and near infrared) are influenced by phenomena occurring at the very sea surface, like the damping of the small-scale surface waves (seen by the SAR) and the accumulation of cyanobacteria at the sea surface (seen by the near infrared bands of TM and WiFS). Whether the SAR detects the presence of any oily/fatty substances released by the algae and then concentrated along the observed patterns, or whether it detects the physical presence of the cyanobacteria accumulation on the sea surface is an issue to be investigated further.

Noteworthy is also the fact that the distinct ‘north-west-south-east’ front line is displaced in the SAR image with respect to the other sensors, which cannot be explained by a different density of cyanobacteria on the water surface and which is still under investigation.

The conclusion is that the investigation has shown that the combination of satellite imagery of different (optical and microwave) sensors may help to better understand and monitoring those phenomena which are often observed in satellite imagery of the ocean surface, in particular, ongoing algae blooms.

The results described above have been presented at the IGARSS’98 symposium (Gade et al. 1999b) and at the EARSeL’98 conference (Gade et al. 1999b), as well as at the IGARSS’99 symposium in Hamburg in June 1999 (Rud and Gade, 1999a), and also at the Oceans’99 conference in Seattle in September 1999 (Rud and Gade, 1999b).