Bio-optical Measurements, Algorithms & Analyses
Summarized by W.M. Balch
Inherent Optical Properties (Zaneveld)
For the measurement of absorption, this group recommends that filter-pad measurements be made in parallel with ac-9 measurements in order to verify spectral shape, and also to understand the relative importance of particulate and dissolved absorption. Calibration of ac-9's should be according to Wetlabs protocols. Absorption values of pure water should be according to Pope and Frye (1998). AC-9 users should be careful to check the stability of interference filters annually. Finally, when reporting absorption values, one should be sure to designate whether values are for total, particulate, or dissolved fractions. Many of the above comments apply to attenuation. In situ backscattering measurements can now be made with the commercially-available Hydroscat instrument. Regular calibration of the instrument in the field and factory are critical to the proper interpretation of the results. Moreover, "sanity checks" of the data can be done if reflectance and absorption data are available, using the reflectance model of Gordon et al (1983; Barnard and Zaneveld, pers comm.). Other issues that Hydroscat users should be aware of are: 1) data can be affected by instrument orientation (horizontal vs. vertical mounting), especially when the instrument is close to the sea floor, and 2) bubbles can affect any comparison between in situ Hydroscat measurements and shipboard scattering measurements.
Remote Sensing Reflectance (Mueller)
A meeting was convened at Old Dominion University earlier this year for compiling data on RRS+ and RRS-, with the goal of comparing such data and defining protocols. Data are still being collected but is clear that reliable reflectance measurements in the red are rare. As with the above discussion on backscattering, microbubbles near the surface can severely impact surface reflectance measurements. In order to minimize the variance of F/Q, summer match-ups with the sun at peak elevation are ideal. Closure in reflectance measurements can be checked using the reflectance relationship of Gordon et al. (1988) 2 R/Q= S li (bb/(a+bb))i i=1 where Q is the variable distribution function (often set equal to p), and l1 = 0.0949 and l2 = 0.0794. The impact of wind speed on surface reflectance will be addressed by Jim Mueller using tank experiments. A meeting will be held at the Santa Fe ASLO meeting in February, 1999 to discuss the status of the reflectance data set.
Particulate Absorption (Mitchell)
A "round robin" experiment was conducted at Scripps Institute of Oceanography. The goal of this workshop was compare the estimates of b, the path length elongation factor. It was observed that estimates were good to 15-20% for cells>1 µm diameter. Interestingly, Prochlorococcus had b values significantly different from other algal species used in this workshop. This is likely due to its small size. Vicarious checks are recommended for resulting cholorophyll-specific absorption by comparing the 683nm specific absorption with maximum a*683 for chlorophyll a & c containing cells (max a*683 ~ 0.025 m2 mg chl-1) or for chlorophyll b containing cells ( max a*683 ~0.03 m2 mg chl -1) Another "sanity check" is to use a and k values to estimate the mean cosine (mu bar = a/k). It was pointed out that the protocols need to address b . Comparisons of ac-9 vs. filter pad absorption should be done in eutrophic waters where the ac-9 is not limited by path length. Absorption of colored dissolved material requires a preliminary filtration process to remove particulate material. Glass fiber filters can contain significant quantities of CDOM, and must be either pre-combusted at 5000 C (as suggested by Bricaud) or treated with 10% HCl. In either case, pure water should be run through a filter prior to sample filtration. Just as glass fiber filters can leach CDOM, they can also absorb it. This issue makes use of GFF filters for dissolved absorption measurements questionable, and it is preferable to use pre-rinsed polycarbonate filters.
Pigments (Clark)
There was considerable discussion on HPLC vs. fluorometric chlorophyll a determinations. Some of the important issues are that ratios of HPLC chl a: fluorometric chlorophyll can range from to 0.6. Another important issue is that HPLC pigments are difficult to do consistently due to the considerable expertise required. Some other questions raised were 1) what combination of HPLC derived pigments are needed to compare to fluorometrically-derived chlorophyll? 2) what protocols must be used to process HPLC measurements? Several recommendations were made concerning the measurement of pigments. 1) If the SIMBIOS team deems that HPLC pigments are a priority, then funds need to be made available for a laboratory to run such samples for all SIMBIOS investigators. The sample burden for such a laboratory could be simply surface samples from clear; SeaWiFS overpasses rather than entire water column profiles from all stations. 2) A second recommendation is to write out protocols for processing HPLC pigments. International JGOFS protocols exist for flourometric chlorophyll measurements and are sufficient for SIMBIOS. 3) Water soluble pigments should be included in the measurement of accessory pigments. 4)Fluorometric chlorophylls are mandatory for all SIMBIOS measurements while HPLC measurements should be done when possible. 5) Investigators need to be wary of limitations of both techniques.
In situ match-ups (Schieber)
Discussion centered around how to improve match-ups. It was suggested that viewing geometry be also taken in consideration in comparing satellite and ship derived Lw's. Moreover, recording all pertinent information for each of 9 pixels centered over the ship station is essential, in order to adequately describe the statistics. It was also pointed out that there are strong non-linearities involved in taking regional averages, which bias the statistics to underestimate chlorophyll a. Lastly, in order to adequately describe match-up statistics, it is essential that all data are available to all SIMBIOS investigators.
Minimizing error due to residual atmospheric noise (Frouin)
It was pointed out that significant errors in chlorophyll prediction can result from using algorithms based on R443/R550 or R490/R550. The NDPI algorithm ([R443-R555]/R490) has less error from residual atmospheric noise, especially at low chlorophyll concentrations.
Atmospheric Measurements, Algorithms & Analyses
Summarized by John N. Porter
Many suggestions and lively discussions took place during this session. In general the discussion/suggestions can be grouped into how the atmospheric validation measurements could be improved and secondly on how the atmospheric correction could be improved in the SeaWifs algorithm. Although many suggestions were made, it was realized that the data sets collected over the past year are just now being processed. This means we do not yet have an overall assessment of the SeaWifs atmospheric correction. In particular we need to better understand under what conditions the SeaWifs approach might works well and under what conditions it works poorly.
Suggestions for better validation measurements
Several comparisons between satellite and ground based optical depth (865nm) measurements were shown during the SIMBIOS meeting. It appeared that for certain sites the agreement was better than at other sites. During the meeting, it was suggested that the best validation measurements appeared to come from ship observations and that this might be due to land adjacency effects in the satellite image. At this point this hypothesis has not been tested. It would be good to look into this further to see if land adjacency is creating a problem in the validation effort. . The Cimel at the Dry Tortugas for example showed to poorest comparisons with the SeaWifs optical depths. It was suggested that the Dry Tortugas ground system be compared by sending a MicroTops handheld sun photometer to the site for simultaneous measurements. This would eliminate questions about the quality of the measurement at this important site. Another Cimel coastal site was Bahrain, which had fairly good agreement between SeaWifs and ground based optical depths (875nm). This good agreement is rather surprising as dust is likely to be poorly modeled by the SeaWifs aerosol models. The Hawaii HOT site had generally good agreement but certain days were significantly off for no apparent reason. Overall, it was decided that this error in these comparisons could be due several possibilities including incorrect aerosol models, poor quality surface validation measurements, improper correction of the foam in the satellite algorithm, land adjacency effects, and satellite calibration problems.
The aerosol model problem is an obvious one. In order to see under which aerosol types the SeaWifs algorithm might fail, it was suggested that surface measurements should provide additional information which could detect the presence of dust (non-spherical particles) or highly absorbing aerosol (anthropogenic soot). M. Wang suggested polarization measurement capability be added to some of the Cimel instruments. The retrieval approach could follow a paper from Howard Gordon et al.. Unfortunately some questions are still present about the durability of polarization filters (B. Holben). It was also suggested that sky radiances from the Cimel network could be used to look for variations in sky radiances. Currently the sky radiances from the Cimel instruments are not being used. In the future, some effort should be made to see if the presence of non-spherical particles can be detected in sky radiances. The data sets for this approach are already available through the Cimel network. An agreed upon standard algorithm would be needed.
Aerosol absorption is another area which is poorly derived from surface or satellite measurements. It was suggested that aerosol properties derived from the TOMS satellite could be combined with SeaWifs satellite images to derive information on the aerosol absorption. Irregardless of how well the aerosol single scatter albedo can be derived, the TOMS data would allow for the detection of regions with significant absorbing aerosol which is more than we now know. It was also suggested that making ground based optical depth measurements at high air mass would allow for a TOMS type analysis to derive the aerosol absorption optical depth. Both the TOMS and surface approach are new and will require further effort.
Several members also suggested that air mass trajectories could be used to identify which type of aerosol were present. This might help in choosing appropriate aerosol models. While this approach offers promise, it was felt that this approach was more suitable for case studies at this point.
It was also realized that the best possibility for validating atmospheric corrections was through aircraft measurements but that these required a major investment in time and money. Therefore SIMBIOS should take advantage of several upcoming major experiments. These include the INDOEX (spring 1999) experiment in the Indian Ocean (Piotr Flatau is a SIMBIOS funded participant). This experiment will include a suite of ship and aircraft measurements which will be ideally suited for ocean color validation. The ACE-Asia experiment off Asia (~2000-2001) also offers another opportunity for ocean color validation. Also several smaller ongoing experiments in Alaska (K. Stamnes) and Hawaii (J. Porter) offer the opportunuty for ocean color validation from aircraft platforms.
The handheld MicroTops sunphotometers are currently the main instrument being used for optical depth measurements on cruises. Currently no standard protocol exist for the use and processing of these measurements. J. Porter and M. Miller have agreed to write up a report on their experience with these instruments. Contributions from others who have had experience with these instruments are welcome. There was also discussion about other ship based sun photometer instruments by M. Miller (ship based shadowband radiometer), J. Porter (gimballed shadowband radiometer). These instruments offer autonomous operation but have not yet matured enough for broad use. R. Fruin has developed and comercialized a unique instrument (SIMBAD) which can measure aerosol optical depths and surface reflectance. The SIMBIOS instrument pool also has several other instruments (Prede, Cimel). At some point a comparison between the various types of instruments should be carried out and written up.
Suggestions for SeaWifs algorithm
The suggestions for improvements on the atmospheric correction focused on the SeaWifs satellite. R. Fruin suggested the spectral dependence of white caps should be included in the model. He argued that his data as well as a body of more recent data show similar effects. During M. Wang's talk he mentioned that a new aerosol model has been included in the SeaWifs algorithm so that water leaving radiances were not negative after subtracting aerosol radiances. J. Porter suggested that the new aerosol model should be tried which he argues are more realistic. Several newer aerosol models are available with larger coarse mode aerosol than the Shettle and Fenn aerosol model currently used.
Other Suggestions
- It was suggested that SeaDAS training classes be continued. Many of the investigators have not yet taken advantage of this opportunity
- It was suggested that a set of "problem images" be saved with matchup data sets. This would allow for different test to see where the problems might exist. It was suggested that the TOMS data set be included in these problem cases when possible.
- It was suggested that funded SIMBIOS investigators be allowed to run the QC algorithm. This will allow them to test the model in detail. There was also some discussion of which products the users might want. From our meeting, the ability to determine which aerosol model was being chosen and the water leaving radiance at 670nm were desired items.
- K. Stamnes asked for suggestions on the inherent optical properties of the ocean and atmosphere which he could use in his modeling studies (phase function, single scatter albedo, etc.). He did not receive a clear answer as there was not agreement on this matter.
