MONITORING NORTH SEA COASTAL WATERS:
FROM RADIANCE AT SENSOR DATA TO A WEB MAPPING SERVICE

M.A. Eleveld (1), A.J. Wagtendonk (2), R. Pasterkamp (1) & A.Q.A. Omtzigt (2)

(1) Programme Unit Remote Sensing (PERS), Institute for Environmental Studies (IVM), Vrije Universiteit Amsterdam (NL)
(2) SPatial INformation laboratory (SPINlab), Vrije Universiteit Amsterdam (NL)

Abstract
Monitoring is a prerequisite for the assessment of changes in the state of the North Sea. Various ocean colour satellite sensors collect data sets of the North Sea every day. These data sets need to be processed to extract information on water quality parameters. Processing needs to be optimised when using remote sensing for monitoring purposes. This paper presents a processing chain for the monitoring of Total Suspended Matter (TSM) in the southern North Sea with data from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS). All SeaWiFS data of the research area for the year 2001 were acquired through the Internet and processed using the SeaWiFS Data Analysis System (SeaDAS 4.0) with MUMM's turbid water extended atmospheric correction algorithm. Subsequently, the POWERS TSM algorithm was used to derive TSM concentration (in mg l­1) from SeaWiFS sub-surface irradiance reflectance, R(0­), in band 5. This resulted in 491 TSM products: on average more than one image a day. Seasonal variation in TSM concentration was extracted from composites, and statistics on TSM concentrations were produced for any location within the research area. The data were exported as tables with comma separated values. In ArcView 3.3 with the Spatial Analyst extension, these values were imported as Event themes into a View with Universal Transverse Mercator (UTM) WGS 84 projection. The point data were converted to grid using the Nearest Neighbour algorithm, and exported as georeferenced images. Scripts were used to incorporate the grids into the IVM map layout for an atlas. The grids were also put on CD to enable further GIS analysis. Images from the previous year (2000) were used in an ArcIMS application. The monitoring data are currently available in different formats for different users: as a hardcopy atlas accompanied by a CD with the data in ArcView format for the coastal management authorities that commissioned the work, and (for the 2000 data) as a Web mapping service that allows anyone with an Internet connection to interactively combine the TSM data with other geographical information. The IVM processing chain appeared to work well, because we were able to process all southern North Sea data sets from SeaWiFS for the year 2001 within a few weeks time. The result is useful for monitoring because of its strength in both 2D spatial and temporal coverage: an average sampling of 107 data sets per pixel over the total of 491 TSM data sets for 2001. Unfortunately, the processing chain cannot easily be adapted for the processing of data from other ocean colour sensors, but certain components can be plugged into new processing chains. Although the introduction of ocean colour remote sensing data for the monitoring of water quality parameters was successful, their use has not yet been fully incorporated into the regular monitoring practices of the water managers.

1 Introduction
Monitoring the North Sea is required to support the management and environmental protection of the North Sea, and to comply with laws and international conventions. Many of these monitoring activities are being performed by order of national management authorities. In the Dutch case this is Rijkswaterstaat, the Ministry of Transport Public Works and Water Management. Remote sensing is ideal for monitoring practices. In the past, it has been used succesfully for creation of individual maps, because the measurements do not disturb the study object, and because they can cover a large area. Nowadays, remote sensing is also increasingly being used for monitoring, since a steady data stream has become available, and computing power for automated processing has increased, resulting in the required high spatial and temporal coverage.
This paper focuses on the monitoring of North Sea water quality parameters. These parameters exhibit important spatio-temporal variability that can be well detected with ocean colour remote sensing monitoring techniques. Remote sensing gives information about some water quality parameters because light is subject to various optical effects such as absorption and scattering when it illuminates water bodies. Sensors, consisting of spectro-radiometers suitable for determining ocean colour measure light radiating from water bodies as water-leaving radiance. Such ocean colour data from various sensors are currently sent to ground stations, so that most areas, such as the North Sea are covered at least daily. Some of the best-known ocean colour satellite sensors currently collecting data are: the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) that has been collecting data since 1997, and the MODderate resolution Imaging Spectrometer (MODIS) on the Terra and Aqua satellites, and the MEdium Resolution Imaging Spectrometer (MERIS), which have been collecting data since February 2002, May 2002, and March 2003, respectively.
Although ocean colour data is available, water managers are currently hardly incorporating information derived from remote sensing techniques in their standard monitoring practice, because the information is not easily accessible, and is often disparate and frequently presented without context. These problems have been solved for the monitoring of Total Suspended Matter (TSM) in the North Sea, because for some years Dutch coastal zone management authorities (Rijkswaterstaat) have commissioned the production of an atlas of Total Suspended Matter (TSM) for the North Sea based on satellite imagery (Pasterkamp et al., 2002, in press), ensuring some continuity in the information flow. In addition, presentation of the results, and integration and interactive analysis with other data is nowadays possible through our Web mapping service (http://www.feweb.vu.nl/gis/SPINlab/mapservice/Noordzee_atlas/
noordzeeatlas.asp).
For these projects, the Institute for Environmental studies (IVM) has been using its own processing chain to efficiently handle large volumes of data in a short time, as required for monitoring purposes. In this chain, a stand-alone GIS, ArcView, was used for re-projection and conversion, and scripts were used to incorporate the grids into the IVM map layout for an atlas. A network-based GIS, ArcIMS, was used for the widespread distribution of TSM maps and for allowing access to, and comparison with data on a variety of related subjects.
This paper aims to show how to set up a processing workflow for the production of water quality products from remote sensing data, and to illustrate the introduction of ocean colour remote sensing data into the monitoring practices of water managers.

2 Online ocean colour data and processing tools
Information on ocean colour satellites, ocean colour data, and even some ocean colour processing software are available on the Internet. These elements can be used in the construction of processing chains, which are required for monitoring purposes. Therefore, a general overview of available material for some well-known ocean colour satellites, SeaWiFS, MODIS and MERIS, is given first, followed by elaboration of a processing chain for SeaWiFS in section 3.

2.1 SeaWiFS
Information on SeaWiFS is available through the SeaWiFS Project homepage (http://seawifs.gsfc.nasa.gov/SEAWIFS.html). SeaWiFS overpass information is available from NASA's Satellite Overpass Predictor (http://earthobservatory.nasa.gov/MissionControl/overpass.html). SeaWIFS data sets can be requested from NASA's Distributed Archive Center (http://daac.gsfc.nasa.gov/data/dataset/index.html). For processing, the standard SeaWiFS Data Analysis System (SeaDAS) source code and binaries are available through the SeaDAS homepage (http://seadas.gsfc.nasa.gov/). Additional software for atmospheric correction for turbid water regions (Ruddick et al, 2000) is available from MUMM's Ocean colour site (http://www.mumm.ac.be/OceanColour/ocrmumm.htm).

2.2 MODIS
Many of the tools available for SeaWiFS can also be used for MODIS: MODIS Terra and Aqua overpasses can be predicted with NASA's overpass predictor (http://earthobservatory.nasa.gov/MissionControl/overpass.html), MODIS data can be requested from NASA's Distributed Archive Center (http://daac.gsfc.nasa.gov/data/dataset/index.html), and SeaDAS (http://seadas.gsfc.nasa.gov/) can display MODIS data. MODIS is especially interesting for monitoring purposes because of its direct broadcast capability: in addition to storing data for later download at designated intervals, MODIS immediately broadcasts the raw data it collects (http://modis.gsfc.nasa.gov/data/directbrod.html). NASA's overpass predictor can be used to determine when there will be a Terra or Aqua spacecraft overpass at any location that may have a MODIS Direct Broadcast receiving station. A comprehensive, up-to-date preview of MODIS data is available through the MODIS Rapid Response System (http://rapidfire.sci.gsfc.nasa.gov/realtime/). Every granule that covers some area of land appears on this site in near-real time (i.e. with a delay of a few hours after acquisition).

2.3 MERIS
MERIS overpasses can be predicted with Display Earth Remote Sensing Swath Coverage for Windows (DESCW) (http://earth.esa.int/descw/). Data products can be browsed through the EOLI Envisat catalogue (http://cat.envisat.esa.int/), which is an early release of the online multi-mission catalogue that will give access to all ENVISAT, ERS and ESA third party missions. At present (August 2003), it allows you to browse the metadata and quicklook images of the currently available ENVISAT products (i.a., MERIS Full and Reduced Resolution), and view the planned and potential acquisitions for these instruments; it does not yet support automatic ordering. The MERIS L0, L1 and L2 products (MER) are delivered in a special ESA format (see MERIS product Handbook http://envisat.esa.int/dataproducts/meris/CNTR.htm). Regular image processing software will not directly be able to handle these product formats, but several tools were developed by ESA to read, process and analyse the MERIS data products (http://envisat.esa.int/services/tools_table.html). EnviView2.0 is a free application that allows Envisat data users to open any Envisat data file, examine its contents, and export the data in Hierarchical Data Format (HDF) (http://hdf.ncsa.uiuc.edu/) for use in other software packages. The Basic ERS & Envisat (A)ATSR and MERIS Toolbox (BEAM) is a collection of executable tools and an application programming interface (API) which has been developed to facilitate the utilisation, viewing and processing of ESA MERIS, (A)ATSR and ASAR data. BEAM software is open source and comes with full source code. One of the main components of BEAM is VISAT, a visualisation, analysis and processing software tool entirely written in Java. The MERIS and (A)ATSR Toolbox provides functions and C routines for direct ingestion of MERIS data into commercial software. In VISAT 1.1 data can be exported in MATBX-DIMAP (.dim, ESA format), or HDF 5 format. These characteristics enable BEAM to be implemented in processing chains. Such a chain can be relatively short, because virtually all users will start from Level 2 products that contain, i.a., reflectance for 13 bands, and standard water quality products, such as TSM and CHL (chlorophyll) concentrations.

3 IVM's SeaWiFS processing chain
To produce regular TSM atlasses from SeaWiFs data IVM developed a processing chain (Fig. 1). It starts with the SeaWiFS radiance at sensor data that were available through the Internet. For our atlases of the SeaWiFS data from the Dundee ground station were used.
First the data were obtained free of charge from NASA by File Transfer Protocol (FTP). The files contained radiance counts for the eight SeaWiFS bands, additional calibration and navigation data, instrument and spacecraft telemetry, and ancillary data on wind, surface pressure, humidity and ozone. Then the data sets that have the southern North Sea in a central position on the images were selected based on filtering on filenames, which carry information about the time of overpass.
Subsequently this data set is pre-processed. Radiance at sensor, Lrs, is a measurement of light reaching the sensor from water and the atmosphere. The atmosphere over coastal regions differs from the ocean atmosphere, and the amount of TSM in coastal waters differs from the amount in the ocean. Therefore, an extension to the standard SeaWiFS Data Analysis System (SeaDAS) atmospheric correction method (Ruddick et al., 2000) had to be used to get values for the calibration parameters, MUMM-epsilon (e) and MUMM-alpha. Using these parameters subsequent processing in SeaDAS generates atmospherically corrected Level 2 data, subsurface irradiance reflectance R(0-) of the southern North Sea.
Based on this intermediate product water quality parameters can be derived. In this case a one-band algorithm based on band 5 (545-565 nm wavelength) was applied to derive TSM concentrations (Van der Woerd et al., 2000; Van der Woerd & Pasterkamp, in press). In addition to the TSM data, TSM quick-looks were generated, and the percentage of cloud cover over the southern North Sea was provided for each image.
The TSM data were reprojected to a rectangular co-ordinate system. Based on this georeferenced TSM data set, further analysis has been performed. Individual images are almost never 100% cloud-free for the entire southern North Sea. This is one of the reasons that composites of images were also made. Statistical analysis in Matlab provided mean, standard deviation, median, and number of samples per grid cell. The data have also been exported as tables with comma separated values.
In ArcView 3.1 with the Spatial Analyst extension, these values were imported as Event themes into a View with Universal Transverse Mercator (UTM) WGS 84 projection, interpolated to grids, and exported as georeferenced images.
These images were used in an ArcIMS application (URL http://www.feweb.vu.nl/gis/SPINlab/mapservice/
Noordzee_atlas/noordzeeatlas.asp) that also contains additional information layers on current activities on the North Sea and plans for the Dutch North Sea coast. First the layers were built with ArcIMS Author; the georeferenced TSM images were added manually. Then the ArcIMS MapService was created in ArcIMS Administrator. Finally the layout of Web site was designed with ArcIMS Designer. This resulted in a Web mapping service that allows users to interactively combine the TSM data with other information on the North Sea.

Figure 1. IVM's processing chain.

4 Output
The resulting IVM processing chain enabled us to process the 2001 data set to TSM products, and to compile an atlas within a few weeks time. The 2001 TSM Atlas of the North Sea contained various daily, two-monthly and seasonal images. It was delivered in hardcopy, accompanied by a CD with the data in ArcView format for the coastal management authorities that commissioned the work. Fig. 2 gives an example of one of the maps in this atlas. A copy of the Atlas of 2000 can be downloaded in pdf-format from Watermarkt, a portal on water monitoring by the Dutch Ministry of Transport, Public Works, and Water Management) (http://www.watermarkt.nl/digiproducts/noordzee-atlas%202000.pdf).
As a test, maps of the 2000 atlas were used in a Web mapping service (Fig. 3) that also contains additional information on current activities on the North Sea from Rijkswaterstaat and preliminary research results on potential locations for a possible airport in the North Sea. The data are presented as static GIS-based maps that can be used interactively. Therefore, the Web maps classify as static interactive maps (Kraak & Brown, 2000).

Figure 2. Example of a map from the North Sea atlas for 2001. Two-monthly mean of TSM concentrations in January and February 2001 (based on unclouded pixels of 76 individual images).

Figure 3. Screen-dump showing the ArcIMS service. Dredge spoil locations and sand winning areas (active layer), are being examined together with mean Total Suspended Matter (TSM) in the period January-February 2000. In addition, several features on land, e.g., rivers, cities, borders are visible. The results of a query on one of the sand winning sites is presented just below the map.

5 Discussion and conclusions
The IVM processing chain appeared to work well, because we were able to do all processing within a few weeks time. The result is useful for monitoring because of its strength in both 2D spatial and temporal coverage (in our study: an average sampling of 107 data sets per pixel over a total of 491 TSM data sets for 2001). Thus we have, on average, more than one TSM image a day for the year 2001. This generates new information about large scale processes in the North Sea (Eleveld et al., in press). Although we optimised processing for SeaWiFS to a great extent, this processing is not generic, because it cannot automatically be applied to other ocean colour sensors. Further standardisation and interoperability, concepts that are already being implemented in the GIS world, are also developing within the remote sensing society, e.g. in the Sensor Intercomparison and Merger for Biological and Interdisciplinary Oceanic Studies (SIMBIOS) initiative (http://simbios.gsfc.nasa.gov/). Nonetheless, currently different processing lines have to be developed for different sensors. This might be worthwhile for monitoring purposes, because of the increase in the amount of available information.
Following original user requirements (Van der Woerd et al., 2000), the products resulting from the IVM processing chain were delivered as a hardcopy atlas accompanied by a CD with the data in ArcView format for the coastal management authorities that commissioned the work. We have also created and maintained a Web mapping service, on our initiative. Currently, the hardcopy atlas is frequently consulted, whereas the ArcView files are less frequently used. The hardcopy atlases are mostly used by national coastal managers and policy advisors, and by researchers of ocean colour of the North Sea. The Web mapping service has attracted 10723 visitors in 32 months, equalling an average visiting rate of 335 visitors per month. Web statistics show that the information is accessed from all continents, and from university, business and industry, as well as private accounts.
Since the start of the TSM atlas projects in 2000, the use of ocean colour remote sensing data for water quality monitoring practices has to a large extent been accepted and encouraged by Rijkswaterstaat. The TSM results are available, affordable, and reliable. Especially the high spatio-temporal coverage at low cost offered by remote sensing is appreciated in times of reduction of monitoring stations and optimisation of measurement schemes, although discussion about the comparability of TSM values derived from remote sensing with TSM from in situ measurements is continuing. In situ measurements have the advantage that they are fully developed methods, that allow for testing against standards (http://www.nen.nl/). In addition, they can often also give the information for different water depths. Therefore, remote sensing is mostly seen as complimentary to in situ measurements. Nonetheless, although ocean colour remote sensing is envisioned as a new measurement technique that needs to be introduced into the regular monitoring practices, this has not yet been formalised. The use of optical remote sensing for the monitoring of water quality parameters has yet to be fully incorporated into the regular monitoring practices of Rijkswaterstaat.
Steps required for implementation comprise:

  1. Creating user support;
  2. Optimising bulk processing of remote sensing data, and (statistical) validation for specific topics and areas;
  3. Performing risk-analysis to support strategic planning and management (e.g., an analysis of Strengths, Weaknesses, Opportunities and Threats (SWOT));
  4. Improving the service (or technique);
  5. Implementing.

We have been working on steps 1 and 2, and might continue towards step 3 in collaboration with the water managers.
In parallel with the creation of our North Sea Web mapping service, the use of the Internet for dissemination of measurement results of Rijkswaterstaat is increasing, and there is also an interest in Web mapping applications for the North Sea (e.g., http://www.noordzeeatlas.nl). Watermarkt (http://www.watermarkt.nl/) is an example of a Rijkswaterstaat site that aims to make information and knowledge available to the public. Aggregated data, long-term monitoring data, and near-real time monitoring data are available from Waterstat (http://www.waterstat.nl), WaterBase (http://www.waterbase.nl/) and the near-real time water data page (http://www.actuelewaterdata.nl/), respectively. There is even a site specifically dedicated to in situ measurements of a specific water quality parameter, phytoplankton (http://www.fytoplankton.nl/). In addition to increasing the volume of online information, Rijkswaterstaat is also involved in new technical developments that determine how the information is given, such as, CoastBase (http://www.coastbase.org), which is an Internet-based distributed system that aims to improve search, access and manipulation of data and information within Europe (Eleveld et al., 2003).
Currently, coastal mangers mainly acknowledge the strengths of ocean colour remote sensing in combination with online data access and processing on demand, for determination of the spatial extent of events, and for explanation through hindcasting. Ocean colour remote sensing can be used to track pressures on the North Sea, and to observe impacts. Analysis of long time series of these spatial data sets enables to unravel natural variations from human impact. However, the full potential of ocean colour remote sensing as a monitoring tool for regional seas has yet to be explored. We think that this Dutch case shows an example of scientists and water managers jointly trying to optimise monitoring using up-to-date remote sensing, GIS, and Internet technology.

Acknowledgements
We want to thank NASA for providing the SeaWiFS data and SeaDAS processing software, MUMM for providing the atmospheric correction algorithm, and the Survey Department (MD) and the National Institute for Coastal and Marine Management (RIKZ) of Rijkswaterstaat for the financial means that have enabled us to work with the data. Hans van der Woerd (VU-IVM) and Kees van Ruiten (RIKZ) are thanked for useful discussions.

References


URLs