Oceanic Productivity: Applications for Remote Sensing and GIS

GEO 565 Annotated Bibliography (Winter 2007)
by Karen A. Breitlow

An Overview

*Note: The annotated bibliography that follows is ordered in the fashion as described in the introductory paragraphs for ease of understanding why the papers discussed velow were chosen for this assignment.

The application of satellite imagery to oceanography is still in its infancy as evidenced by the many algorithms and methods for spatial analysis of chlorophyll alone. Waters have been divided into two major cases, each needing its own category of often multi-band algorithms in different wavelengths. A number of satellite sensors are currently being employed, each with its own strengths and limitations that need to be taken into consideration during analysis such as spatial and spectral resolution. In fact, while chlorophyll is generally used to give estimates of net primary productivity in the oceans, recent studies have proposed alternatives using backscattering coefficients to measure phytoplankton carbon to give a much more accurate estimate of productivity.

The remote sensing community has recognized the need for standardized methods and software compatibility and has pushed discussion of such matters into the forefront. However, much research needs to be done before a consensus on optimal methodology for even the more basic measurements can be reached. In the meantime, empirical descriptions and the development of various indices have been used in an effort to categorize basic regions according to their characteristics as is the case for “bioregions.” This technique is being applied to various GIS applications, but only at a very general level due to the difficulties mentioned previously when looking at data in more detail or at a smaller spatial scale. Also discussed are future applications of GIS with further implementation of complex data models and software currently under construction such as ArcMarine that will hopefully minimize the gap between the mathematical based scripts used in MATLAB by researchers and the commercial GIS packages used by marine management and conservation agencies.

An Overview of MODIS Capabilities for Ocean Science Observations

Remote sensing analysis has come to the forefront in terms of measuring global oceanic productivity. It was the Coastal Zone Color Scanner that allowed researchers to study the distribution and abundance of phytoplankton and its importance in various biogeochemical cycles. Since then, the Moderate Resolution Imaging Spectroradiometer (MODIS) has been developed and provides scientists with sea surface temperature (SST) and ocean color at a resolution of 1km. It also uses narrower color bands than SeaWiFS, which increases the atmospheric correction and a center wavelength of 531nm to give better signals for accessory pigments. Through various algorithms, the concentration of chlorophyll a and phaeopigments, chlorophyll a alone, the diffuse attenuation coefficient, and amount of suspended solids can be determined [for case 1 waters].

Phycoerythrin is a type phaeopigment that contains phycoerythrobilin (PEB), which is found in all phycoerythrin-containing marine cyanobacteria. The PEB concentration can be determined using the numerical radiance model inversion, and validated with airborne PEB fluorescence and published values from laboratory experiments. This information is critical in determining the global distribution and abundance of cyanobacteria such as Trichodesmium, a major nitrogen fixer in the open ocean, where nitrogen has been shown to limit phytoplankton production. In some areas, the carbon production by cyanobacteria is greater than that of larger phytoplankton. On a more general note, the identification of bioregions describing the dominant species found with seasonal and geographic changes would allow more accurate analysis of the optical properties of those areas since different species exhibit different absorption and scattering characteristics. The delineation of bio-optical domains improves the error in chlorophyll a estimation from 50% to 30%.

MERIS for Coastal Zone Applications: Past Achievements, Future Prospects

The coastal upwelling area of Benguela between Cape Town and Orange River was studied in October of 2002 to determine whether empirical analysis of remotely sensed data can be used to characterize the distribution of phytoplankton functional types (PFTs) such as diatoms, dinoflagellates, flagellates, and prokaryotes. While MERIS data was used to derive phytoplankton functional types, biomass, productivity, and photosynthesis parameters and rates, the authors discussed the advantages to using a real time application such as DISMAR (Data Integration System for MARine pollution) to study oceanic productivity. DISMAR is an interactive GIS system that gives a seven day rolling archive and allows coincident viewing of multiple data sets in the same image but is currently used to study marine pollution rather than chlorophyll concentrations.

A project called BENCAL took field samples from Benguela at multiple depths and used pigment analysis to determine the phytoplankton functional types used later in the analysis. This information was then used to calibrate and validate the MERIS images and analyses, which were for the most part within acceptable ranges. With the detailed information regarding phytoplankton assemblages from the field studies, bio-optical traits of the different groups of phytoplankton can be examined and used to map the general distributions of these groups from satellite imagery. The results from water samples correlated well with optical analysis of satellite imagery based on scattering properties of each phytoplankton type, and the degree of accuracy was quantified. Mismatches occurred only for dynamic regions exhibiting mixed populations, and the accuracy of analysis can be improved with further partitioning of groups based on their optical characteristics. Empirical algorithms were then used to relate chlorophyll with photosynthesis quantum efficiency (PQE) and the absorption cross-section for photosynthesis system 2 (Sigm). PQE values were derived for low light conditions so quenching effects were minimal and the quasi-steady state of the phytoplankton assemblage could be observed. Using these photosynthesis parameters, maps were produced showing the distribution of phytoplankton types while incorporating information from depths now “seen” by the satellites.

Results showed diatoms dominated in areas of high biomass, which were also the areas close to shore experiencing the strongest upwelling with cold, nutrient-rich water. Offshore areas have relatively low nutrient concentrations and were dominated by flagellates. Prokaryotes never dominated any of the sampled areas, but did consistently make up approximately 10-20% of the assemblage in the offshore waters.

Optical remote sensing of chlorophyll a in case 2 waters by use of an adaptive two-band algorithm with optimal error properties

A single approach to oceanic remote sensing does not work because different bodies of water have different optical properties and therefore cannot be treated the same. Two broad categories currently exist: case 1 and case 2 waters. Case 1 refers to a volume of water for which variation of optical properties is dominated by phytoplankton and other associated materials. A number of successful algorithms have been applied to this situation, but the same is not the case for case 2 waters. The latter describes a volume of water with high concentrations of inorganic suspended matter and colored dissolved organic matter. Unfortunately, there has been little success in finding chlorophyll a concentrations for these waters, which are typical of waters experiencing high eutrophication and coastal upwelling areas. The blue-green two band ratio algorithms used for case 1 studies are not appropriate for case 2 conditions.

While a number of multi-band algorithms from which total suspended matter, chlorophyll a, and CDOM concentrations can be calculated have been developed to address this issue, this paper studied a new two band red-near infrared approach where the second wavelength is not static. This way, the algorithm can be adapted to reduce errors resulting from imperfect atmospheric correction. When the second wavelength gives the same reflectance as the first, the algorithm shows optimal performance. This paper compared field results from IJssel Lagoon and two cruises in Belgian coastal waters in April of 1998 and 2000 with the results from the algorithm discussed here using airborne sensors. The in situ and calculated values for the concentration of chlorophyll a were quite close. The authors found 672:704nm worked the best for atmospheric correction because the reflectance errors are relatively flat over this section, while the more distant MODIS bands show increased error. The particulate backscatter of case 2 waters actually improves the signal to noise ratio using the red-near infrared approach.

A number of corrections and calculations are required to determine chlorophyll a concentrations from satellites for upwelling areas: atmospheric correction, air-sea interface correction, bio-optical modeling, and the conversion of the phytoplankton absorption coefficient to the concentration of chlorophyll a. The biggest problem for case 2 studies resides in the estimation of the phytoplankton absorption coefficient at a specific wavelength from “subsurface irradiance reflectance.” An accurate chlorophyll a measurement depends on the phytoplankton species composition and trophic state, so field work is still an integral part of remote sensing analysis. For the proposed algorithm to work, the optimal second wavelength must first be determined. This requires a high wavelength accuracy and spectral resolution in the 700-740nm range. At the writing of this paper, airborne imaging spectrometers are required, but some satellite sensors (CHRIS) hold great potential.

Remote sensing of the coastal zone: an overview and priorities for future research

Remote sensing is unique in that it delivers data at multiple scales, but much work still needs to be done. Basic information regarding spatial and temporal variations of optical properties in the water column is still needed and the methods for processing data are still being developed and tested. In addition, many of the satellite sensors being used have limited spectral capabilities so some of the wavelengths that could serve as useful signatures are lost. Moderate resolution is typical of many satellite sensors being used (i.e. MODIS) and is recommended only for coarse descriptive mapping. Classification accuracy does increase with higher resolution and the use of contextual classifiers and neural networks, so accuracy should improve over the coming years as these techniques are further honed by researchers.

The emergence of biological provinces to divide the ocean waters into more manageable regions according to certain parameters (such as chlorophyll concentration) and/or indices has proven useful to policy makers. An example is the determination of optimal regions for designation as marine reserves.

Carbon-based ocean productivity and phytoplankton physiology from space

Recent research has proposed an alternative to using chlorophyll concentration as a direct measurement of net primary production using satellite ocean color. Remote determination of phytoplankton carbon biomass has been difficult, so net primary production (NPP) estimates were based on chlorophyll and empirical descriptions to account for variability. This approach has given poor results when compared with local field measurements. This is due to physiological changes in phytoplankton where pigment concentrations change in response to changes in light, nutrients, and temperatures.

It has been shown through carbon tracer experiments that the particulate beam attenuation coefficient at 660nm can be used as a ratio with chlorophyll to give an index of phytoplankton carbon:chlorophyll. While this coefficient is not a product of remote sensing, an alternative has been suggested by the authors of this paper.

Recent advances allow the separation of light absorbing and scattering components in the water column, so the chlorophyll and particulate backscattering coefficients can be estimated. This backscattering coefficient can be converted to a measure of phytoplankton carbon biomass as long as non-algal particles co-vary with phytoplankton biomass (studies suggest this is the case for many areas). The chlorophyll to carbon ratios (which are constant for different sets of light, nutrient, and temperature conditions) can be used to calculate the phytoplankton growth rates. The growth rate and carbon biomass can be used to calculate the net primary productivity.

Carbon-based and chlorophyll-based models yield different seasonal cycles in net primary productivity, where cycles were damped in areas of low variance and stronger for areas of high variance. Remotely sensed optical properties continue to be used to gather more information that could change the parameters used for determining such major oceanographic measurements as net primary production. This paper demonstrates the many as-of-yet unexplored uses of remote sensing in the field of oceanography and the ever-changing landscape of methodology and analyses.

A GIS-based Method for Retrieving Ocean Environmental Parameters of Fishing Grounds

Chlorophyll concentration is an environmental parameter commonly used for correlation purposes with fisheries. In the case of this paper, the relationships between fishing grounds in the NW Pacific Ocean for the squid Ommastrephes bartrami and various parameters including chlorophyll, temperature, temperature gradients, temperature anomalies, temperature fronts (often characterized by changes in chlorophyll), and salinity were explored using ArcGIS, SPSS, and SQL Server.

Catch data and related data from the literature were used to create indices over a time series at various time scales. The other features were treated according to their data types. Isoline data underwent multi-section interpolations and in-situ point data underwent Kriging and TIN-based interpolations to create grids. Some data were collected from multiple sources, such as sea surface temperature (SST), and were combined in a single data layer using the statistical Bayes method and a neural network.

Categorical Mapping of Marine Eutrophication Based on Ecological Indices

Eutrophication levels have been characterized by several indices including: the number of phytoplankton species, total number of individuals, Margalef’s index, Menhinick’s index, Odum’s species per thousand individuals, Shannon’s diversity index, and evenness index. An inverse distance weighted interpolation displays the spatial distribution for each index. The end result is a map showing the spatial distribution of marine eutrophication described by four categories: eutrophic, upper-mesotrophic, lower-mesotrophic, and oligotrophic through unsupervised and supervised classifications as well as simple overlay within a GIS interface.

GoMEx – An Experimental GIS System for the Gulf of Mexico Region using SAR and Additional Satellite and Ancillary Data

Synthetic aperture radar, SAR, has been used to detect algal blooms because the algae produce a slick which reduces the surface waves and leads to decreased backscattering. While other factors can also cause lower backscattering values, the usefulness of SAR is not contingent upon sunlight or cloudiness.

The feasibility of using SAR to detect algal blooms was tested by comparing its results with in situ measurements and SEAWIFS for the September 2001 Karenia brevis bloom. All three techniques were found to correlate well, so in some instances, SAR could be used to fill in the gaps in SEAWIFS chlorophyll data due to cloudiness. The described correlation was for an algal bloom dominated by one species rather than the more typical assemblage of several phytoplankton types, each with its own optical characteristics.

SAR data, along with GOES images, SEAWIFS, MODIS, AVHRR, and scatterometer data, are being archived and stored by NOAA in a large database to be accessed via a GIS platform called GoMEx. The program is being run through the World Wide Web Image Processing Environment and in the future will also incorporate data layers such as topography, bathymetry, buoy data, wind barbs, and ship positions. The ability to view multiple datasets in one application would be extremely useful to a wide range of investigators and managers with specific interests.

GIS and Coastal Remote Sensing

Biogeochemical and physical oceanographic processes at regional scales can be modeled using GIS, especially with remotely sensed data from an assortment of satellite sensors. The Sea Viewing Wide Field-of-View System (SeaWiFS) and the Moderate-Resolution Imaging Spectroradiometer (MODIS) provide a wealth of information regarding chlorophyll concentrations in the ocean that demonstrate productive regions in the oceans. Open ocean chlorophyll is the easiest to characterize, but GIS tends to be used more for near-shore studies where satellite-based chlorophyll estimates are much more complicated.

Once a generally accepted protocol for SeaWiFS and MODIS interpretation has been reached, GIS will likely be the next step in evaluating remote sensing data. The advantage to displaying multiple datasets on a single platform means one is able to evaluate relationships of oceanic productivity to terrestrial processes such as rain (which influences amount of suspended sediment and therefore light penetration) and river runoff (which influences nutrient concentrations in coastal waters). In addition, chlorophyll is already being used as a parameter for evaluating fishery productivity, red tides, temperature fronts, and consequences of climate change and GIS could help increase understanding of such links.

Progress and Grand Challenges of Marine GIS

”Spatial reasoning” has become the center of attention in the marine GIS community dealing with dynamic oceanic processes such as chlorophyll and primary production as measured from satellite sensors. In other words, spatial analysis techniques are being explored rather than being limited to the input and display of data in a GIS application. Universal data standards and techniques (i.e. algorithms) need to be developed, and the Integrated Ocean Observing System has been created as a resource in moving toward this goal.

Another problem resides with the many different programs being used by different groups of people. Researchers using SeaWiFS and MODIS data tend to use mathematical scripting languages such as MATLAB while those in the management and conservation sectors use commercial GIS packages. Software still needs to be developed to allow information to be input and accessed by these different programs.

A program called ArcMarine is currently being developed and is designed to temporally reference data structures in order to display spatially and temporally dynamic marine processes. This program would provide a common starting point for complex analyses of oceanic data such as chlorophyll concentrations for a multitude of users. Data models are also expected to play an integral role in this jump from relatively static datasets to datasets that change with time and geography.

Wright, Dawn J. and Patrick N. Halpin. Progress and Grand Challenges of Marine GIS. Geospatial Application Papers. Website accessed on March 8, 2007. http://www.gisdevelopment.net/application/nrm/ocean/marinea.htm