The Baltic Algae Watch System – a remote sensing application for monitoring cyanobacterial blooms in the Baltic Sea
Nowcasting of harmful algal blooms is important both for the public and for environmental management purposes. In the Baltic Sea summer blooms of nitrogen-fixing cyanobacteria are regular phenomena but the past years intense and widespread blooms have caused major environmental concern due to its nuisance, increased nitrogen input and toxicity. The Swedish Meteorological and Hydrological Institute have developed a monitoring application that use satellite data to detect blooms and that assembles essential sources of information at one website. A supervised classifications algorithm has been applied to NOAA-AVHRR data during 1997-2006 and the dataset collected has been evaluated and analyzed. Subsequently, definitions of normalized bloom duration, extent and intensity have been developed to enable comparison between different years. Results suggest that the most intense blooms during 1997-2006 were recorded in 2006, while both 2005 and 2006 had the longest duration. The largest extent was noted in 1998. The experience from the 10 years of monitoring has shown that the combination of satellite imagery, observations, forecasts and tools assembled at one website, is a powerful nowcasting tool for monitoring and prediction of cyanobacterial blooms.
The aim of this paper is to describe our satellite based system for monitoring of cyanobacterial blooms and to present the method we used for evaluating and analyzing the dataset collected during the period 1997-2006. To compare different year and see trends normalized index for bloom duration, extent and intensity were developed.
Cyanobacterial blooms, usually dominated by the toxic Nodularia sp. and non toxic Aphanizomenon sp., are commonly found during summer in the Baltic Sea. Although blooms have occurred in the Baltic Sea for at least 7000 years B.P. , it has been debated if the areal extent and intensity of the blooms have increased because of anthropogenic eutrophication . In recent years there has been an increasing interest from both the public and authorities since the peak of the bloom usually coincide with the summer holidays when people reside near the coast. The thick surface accumulation of cyanobacteria is a major nuisance, especially when the accumulations aggregate along the coast, releasing bad odour and prevent swimming. At worst, toxic substances are released from Nodularia that can kill pets and cattle drinking polluted water. Humans and especially children who get in contact with contaminated water usually suffer from skin irritation, stomach problems and symptoms similar to the flu .
Another environmental concern with the cyanobacterial blooms in the Baltic Sea is the increasing input of nitrogen to the water. Most cyanobacteria use dissolved molecular nitrogen (N2) as an additional nutrient source, which allows them to bloom during summer when growths of other phytoplankton are limited by nitrogen . Estimates of nitrogen fixation by cyanobacteria may be as high as 400 000-700 000 ton/year . In 2000, rivers transported ~660 000 ton nitrogen to the Baltic Sea and ~300 000 ton/year due to atmospheric deposition . Hence, it is likely that nitrogen fixation is the single largest source of nitrogen to the Baltic Sea.
The signature of cyanobacterial blooms in the Baltic Sea is distinct even if the character can differ depending on the vertical position of blooms in the water column and its concentration. Cyanobacteria have the ability to move vertically to find favorable light or nutrient conditions, subsurface blooms often cause a discoloring. Towards the end of a bloom cyanobacteria can loose their buoyancy control and aggregate at the surface forming dense accumulations. The blooms are formed into meandering structures or patches (Langmuir circulation) by winds and currents (see Fig. 1).
Since cyanbacterial blooms have a high spatial and temporal variability it is difficult to provide accurate near real-time (NRT) information using only conventional shipborn sampling. Collecting data using remote sensing techniques enables monitoring of vast areas on a regular basis in NRT. The main disadvantage is that most sensors cannot ”see” through clouds, which limits the monitoring ability at high latitudes, which have a high presence of clouds . Remote sensing data is also limited to the top meters of the sea surface; blooms at greater depths are not visible .
Material and method
Detection of cyanobacteria blooms
Observations of cyanobacteria blooms forming heavy bloom or scum at or close to the sea surface in the Baltic Sea is associated with high optical scattering properties rather than discolouring by absorption properties. This experience emanates from experimental work using satellite data and airborne or ship based reconnaissance (visual inspection and photos). Early attempts to demonstrate this experience can be found in .
The scattering properties are however intrinsically difficult to measure with existing in situ instruments simultaneous with remote sensing data. Both spatial texture and near surface accumulations of heavy blooms are the main problem for in situ sampling, whereas atmospheric correction methods generally fail for satellite data . Hence, validation of in situ scattering and absorption characteristics of these heavy phytoplankton blooms are still unsolved, mainly due to lack of adequate methodology and technology.
We therefore conclude that heavy blooms of cyanobacteria close to or at the sea surface can be detected by their scattering characteristics in the red and near infrared part of the electromagnetic spectrum by satellite sensors such as NOAA-AVHRR, MODIS, MERIS and Hyperion (cf. ).
Detection of algae blooms using optical satellite data is restricted to cloud free areas, since direct sun radiation is absorbed and reflected by clouds, hindering radiation information from the upper water column to reach the satellite sensor. Time series of optical satellite data can therefore be biased and/or aliased, concerning spatial and temporal coverage of blooms. These effects remain to be proven but are subjects outside the scope of this investigation. Here we can provide information based on the ten years of time series data (1997 - 2006). We find that algae blooms are detected in 20 to 60 % of the cloud free satellite data. Typically cloud free days occur as often as cloud covered days during July to August, but varying from year to year. Furthermore, the temporal coverage of blooms varies between a few days up to 20 days, always considerably less than available number of cloud free days. We conclude that heavy blooms are not under sampled in cloud free areas and these areas are on average evenly distributed over the open Baltic Sea during summer months.
Our hypothesis is therefore that optical satellite data, available several times a day, in NRT, (e.g. NOAA-AVHHR) can be used to estimate the occurrence of these blooms in time and space.
The method to process NOAA-AVHRR data to detect cyanobacteria is basically a supervised classification algorithm that relies on multiple thresholding and difference in visible, near infrared and thermal channels to distinguish cyanobacteria accumulations from other common features as land, clear water, clouds, haze, sun glint and error pixels. The method includes manual interpretation and correction of each satellite scene .
A common problem when using NOAA/AVHRR data is the relatively coarse pixel size (~1 km2), which make bloom detection impossible near the coast. Hence, the method can only be applied for monitoring of offshore blooms. Hindcasting with MERIS or MODIS satellite data with a spatial resolution of 300m (Available at ESA, ENVISAT Web File Server) and 250-500m respectively (Available at NASA, Level 1 and Atmosphere Archive and Distribution System) are therefore used to visually validate the daily bloom maps.
To get an estimate of cloud-free observations in relation to bloom and cloud observations the cloud cover is screened for every analyzed image. A cloud mask produced by the NoWCasting Satellite Application Facility (NWC-SAF)  is loaded during the image processing and the cloud cover over the Baltic Sea is extracted. To further minimize the cloud problem hindcasting with composite maps of the daily bloom maps during the past 7 days are produced showing number of observations in each pixel. Hence, even if cloud cover makes it impossible to detect any blooms, the composite image can picture the algal situation .
The monitoring system
The web based monitoring system was constructed to provide a direct overview of the blooming situation at the default view but also offering more thoroughly information about the sea state. Maps showing the extent of cyanobacterial bloom are updated daily throughout the blooming season (June-September). During cloud free conditions one map per day is enough but if the cloud situation changes rapidly revealing new information maps can be updated up to three times per day. Enclosed to each map a short text is presented describing the current bloom situation and the 7 days composite. The latest sea surface temperature image from Ocean & SeaIce – Satellite Application Facility (O&SI-SAF) is also presented. The default view of the website is presented in Fig. 2.
 provides photosynthetic photon density from the last 24 hours. Another application incorporated in the system is a drift model  adapted and operated by SMHI for forecasts and dispersion calculations of oil-spill but also usable on cyanobacterial blooms. Drift forecast are produced on detected blooms to predict if or if not the blooms will reach coastal areas.
The Swedish national marine monitoring programme performs monthly water sampling cruises in the Baltic Sea. In situ measurements of phytoplankton are analyzed onboard and an algae report is also produced shortly after the cruise giving a brief indication on the phytoplankton species composition, which can be used for sea-truthing of the satellite products. These reports together with cruise reports are available at the website.
The main sources of information assembled at the Baltic Algae Watch System can be found in Fig. 3.
To be able to compare the blooms between different years definitions of bloom normalized duration (T), extent (A) and intensity (I) have been developed. Based on the yearly summaries where the area (ai) is equal to the extent that is covered by surface accumulations of blooms during (i) number of days, the normalized duration, Eq. (1), and extent, Eq. (2), is given. Where (i) ranges from 1 to the maximum number of days with bloom observations during the current year. The intensity, Eq. (3) is given in “extent days” or km2days .
The highest normalized intensity was recorded during the bloom 2006, while both 2005 and 2006 had the longest duration. The largest extent was noted in 1998 but the bloom was interrupted by cold and windy weather and the duration of the bloom was short, see Table 1. Ten days of data is missing from the year 1999 due to antenna malfunction. It is likely that blooms were present during this time gap since blooms were detected both before and after the interruption which would have given both a larger extent and longer duration e.i. higher intensity. It has also been suggested  that the bloom during 1999 had the highest cumulative blooming area during the period 1979-1984, 1998-2006.
Table 1. Presents a comparison of normalized extent, duration and intensity of cyanobacterial blooms between the period 1997-2006. The results are based on the yearly summaries. Year 2001 is missing due to satellite antenna problems.
Based on the annual summaries and the normalized indexes for extent, duration and intensity hindcasting products are produced after the blooming season. These products are part of the marine environmental Indicator Fact Sheets  that Helsinki Commission is publishing at http://www.helcom.fi to provide information about recent state of and trends in the Baltic marine environment.
Summary and conclusion
The Baltic Algal Watch System has shown that that the combination of satellite imagery, observations, forecasts, models and tools assembled at one website, is a powerful nowcasting tool for monitoring and prediction of cyanobacterial blooms. The system has been appreciated and is frequently used by the Marine Information Offices, authorities, environmental managers, media and public. The number of visitors of the website has increased rapidly since the system became operational in 2002. During 2004 there were about ~43 000 visitors, which increased during the strong bloom in 2005 to ~120 000 visitors. In 2006, which had the highest normalized intensity during the 10 years analyzed the number of visitor increased further too over 149 000. About 70-80% of the visits at the website were made during July and August when the blooms are most frequent and tourists are spending their holidays near the coasts of the Baltic Sea.
Future work will focus on incorporating NRT satellite data from MERIS and MODIS with higher spatial and radiometric resolution than NOAA-AVHRR and obtaining homogenous data sets of the mapping procedure using different sensors. Algorithms for certain cyanobacterial pigments e.i. phycocyanin  yet not evaluated in the Baltic Sea could improve the mapping procedure.
Further evaluations on the impact of clouds on the mapping procedure together with NRT in situ validations with ferrybox systems in the Baltic region  are required.
The main conclusions and experience from 5 years of operational monitoring of cyanobacterial blooms are:
- NOAA-AVHHR is suitable for nowcasting purposes of algal blooms because of high number (10-12) overpasses per day, NRT-access to data, availability of relatively long compiled time series were data continuously are being added.
- The cyanobacterial blooms showed large inter-annual variability. Years with massive blooms can be followed by years with only minor blooms.
- Weather conditions are an important factor for the formation of cyanobacterial blooms. Warm, sunny and calm weather in combination with high concentrations of phosphorus provides suitable conditions for cyanobacterial blooms.
- Cloud cover, which is a major problem when using NOAA-AVHRR data, can be minimized using 7-days composite as a complement to the daily bloom maps.
- The increase of the blooms both in area extent and intensity that have been debated is difficult to prove. Satellite data has been available for about 3 decades, and this period is too short to show trends or natural variability of the phenomena. Though, low N:P ratios observed in the Baltic Sea since 2002 could be one explanation to the high intensity bloom during recent years.
We would like to acknowledge MISTRA RESE-miljömål for financially supporting the development of the Baltic Algae Watch System and Ove Rud for providing satellite data from 1997-2000 and getting us started. We would also like to acknowledge the Swedish Coastguard for providing us with aerial photographs of cyanobacterial blooms.
- ↑ T. S. Bianchi, E. Engelhaupt, P. Westman, T. Andren, C. Rolff, and R. Elmgren, "Cyanobacterial blooms in the Baltic Sea: natural or human-induced?" Limnol. Oceanogr. 45(3), 716-726 (2000).
- ↑ E-L. Poutanen and K. Nikkilä, "Carotenoid pigments as tracers of Cyanobacterial blooms in recent and post-glacial sediments of the Baltic Sea," Ambio 30(4), 179-183 (2001) [doi:10.1639/0044-7447(2001)030[0179:CPATOC]2.0.CO;2].
- ↑ 3,0 3,1 M. Kahru, U. Horstmann, and O. Rud, "Satellite detection of increased Cyanobacteria blooms in the Baltic Sea: natural fluctuation or ecosystem change?" Ambio 23(8), 469-471 (1994).
- ↑ T. Finni, K. Kononen, R. Olsonen, and K. Wallström, "The history of Cyanobacterial blooms in the Baltic Sea," Ambio 30(4), 172-178 (2001) [doi:10.1639/0044-7447(2001)030[0172:THOCBI]2.0.CO;2].
- ↑ K. Kononen, "Eutrophication: harmful algal blooms and species diversity in Phytoplankton communities, examples from the Baltic Sea," Ambio 30 (4), 184-189 (2001) [doi:10.1639/0044-7447(2001)030[0184:EHABAS]2.0.CO;2].
- ↑ J. Degerholm, "Ecophysiological characteristics of the Baltic Sea N2-fixing cyanobacteria Aphanizomenon and Nodularia," PhD Thesis, Department of Botany, University of Stockholm, Stockholm (2002).
- ↑ L. Edler, E. Willén, T. Willén, and G. Ahlgren, "Skadliga alger i sjöar och hav," Naturvårdsverket Report 4447 (1995).
- ↑ P. A. Wheeler and D. L. Kirchman, "Utilization of inorganic and organic nitrogen by bacteria in marine systems," Limnol. Oceanogr. 31(5), 998-1009 (1986).
- ↑ S. Fonselius, "Västerhavets och Östersjöns Oceanografi," Swedish Meteorological and Hydrological Institute Oceanografi report, ISBN 91-87996-07-3 (1995).
- ↑ N. Wasmund, G. Nausch, B. Schneider, K. Nagel and M. Voss, "Comparison of nitrogen fixation rates determined with different methods: a study of the Baltic Proper," Mar. Ecol. Prog. Ser. 297, 23-31 (2005) [doi:10.3354/meps297023].
- ↑ HELCOM, Helsinki Commission, "The Baltic marine environment 1999-2002," Baltic Sea Environment Proc. 87, ISSN 0357-2994, (2003).
- ↑ B. Håkansson, "Satellite remote sensing of coastal oceans: water quality and algae blooms," Seas at the Millenium: An Environmental Evaluation, C. Sheppard, Ed., Elsevier Science Ltd. 3, 293-302 (2000).
- ↑ J. Gordon, M. Antoine, C. Francois, M. Calatuyud, V. Barale, H. M. Snaith, O. Rud, M. Ishii, M. Gade, J. M. Redondo, and A. Platinov, "The Clean Seas Project - final report," Contract number ENV4-CT96-0334, (1999).
- ↑ U. Horstmann, "Eutrophication and mass production of blue-green algae in the Baltic," in Proc. 3rd Baltic Symposium on Marine Biology 239, pp. 83-99 (1975).
- ↑ B. Öström, "Fertilization of the Baltic by nitrogen fixation in the blue-green alga Nodularia Spumigena," Rem. Sens. Environ. 4, 305-310 (1976) [doi:10.1016/0034-4257(75)90026-7].
- ↑ M. Moberg and B. Håkansson, "Monitoring of remotely sensed blue-green phytoplankton in the Baltic Proper during summer 1991," Nordisk Romvirksomhet 5(93), 14-16 (1993).
- ↑ B. Håkansson and M. Moberg , "The algal bloom in the Baltic during July and August 1991, as observed from the NOAA weather satellite," Int. J. Rem. Sens. 15(5), 963-965 (1994) [doi:10.1080/01431169408954127].
- ↑ 18,0 18,1 M. Karhu, O. P. Savchuk, and R. Elmgren, "Satellite measurements of cyanobacterial bloom frequency in the Baltic Sea: interannual and spatial variability," Mar. Ecol. Prog. Ser. 343, 15-23 (2007) [doi: 10.3354/meps06943].
- ↑ 19,0 19,1 A. Reinart and T. Kutser, "Comparison of different satellite sensors in detection cyanobacterial bloom events in the Baltic Sea," Rem. Sens. Environ. 102, 74-85 (2006) [doi:10.1016/j.rse.2006.02.013].
- ↑ M. Kahru, "Monitoring algal blooms: new techniques for detecting large-scale environmental change," Photosynthet. 36(1-2), 50 (1999) [doi: 10.1023/A:1007072215898].
- ↑ A. Thoss and A. Dybbroe, "Scientific user manual for the AVHRR/AMSU cloud and precipitation products of the SAFNWC/PPS," EUMETSAT Satellite Application Facility to support NowCasting & Very Short Range Forecasting, Swedish Meteorological and Hydrological Institute (2005) http://produkter.smhi.se/saf/.
- ↑ 22,0 22,1 M. Hansson, "Cyanobakterieblomningar i Östersjön, resultat från satellitövervakning 1997-2005," Swedish Meteorological and Hydrological InstituteOceanografi Report 82, ISSN 0283-7714 (2006).
- ↑ T. Landelius, W. Josefsson, and T. Persson, "Swedish Meteorological and Hydrological Institute (STRÅNG): A system for modelling solar radiation parameters with mesoscale spatial resolution," Swedish Meteorological and Hydrological Institute, Report 96 (2001).
- ↑ J. Mattsson, O. Nordblom, and O. Liungman, "The new web based oil drift forecasting system: Seatrack web 2.0,"in Modelling in Support of Environmental Management for Resource Management ,Environmental Risk Assessment, and Impact Mitigation, 8th Int. Mar. Environ. Mode. Sem. (IMEMS), Helsinki, Finland, (2005).
- ↑ P. Anderson and B. Karlson, "Harmful algal blooms expert system, final report," Delft Hydraulics 81866 (2004).
- ↑ HELCOM, Helsinki Commission, "Cyanobacterial blooms in the Baltic Sea" Indicator reports, URL: http://www.helcom.se (2006).
- ↑ S. G. H. Simis, A. Ruiz-Verdu, J. A. Dominguez-Gomez, R. Pena-Martinez, S. W. M. Peters, and H. J. Gons, "Influence of phytoplankton pigment composition on remote sensing of cyanobacterial biomass," Rem. Sens. Environ. 106, 414-427 (2007) [doi:10.1016/j.rse.2006.09.008].
- ↑ S. G. H. Simis, S. W. M. Peters, and H. J. Gons, "Remote sensing of the cyanobacterial pigment phycocyanin in turbid inland water," Limnol. Oceanogr. 50, 237−245 (2005).
- ↑ E. R. Rantajärvi, R. Olsonen, S. Hällfors, J.-M. Leppanen, and M. Raateoja, "Effect of sampling frequency on detection of natural variability in phytoplankton: unattended high-frequency measurements on board ferries in the Baltic Sea," J. Mar. Sci. 55(4), 697-704 (1998) [doi: 10.1006/jmsc.1998.0384].
Please note that others may also have edited the contents of this article.