Time Series Tool Description
The SeaBASS Regional Time Series Tool is designed to provide comparisons over time between averaged level 2 satellite-borne ocean color instrument observations as well as in situ measurements. For a given region, one or more satellite sensors must be selected and then all available satellite and in situ data collected within the bounds of the study site will be plotted for comparison in common figures and frequency distributions. User options are available to select a measurement product of interest (e.g., Rrs, IOPs, Chl, etc.), adjust the data averaging period (weekly, monthly, seasonal), the year range, and other graph options. Each region has a brief description that includes links to extra information about the primary in situ data sources and a map that shows the boundaries used to define each region. Users may also download the data that were used to generate the figures.
To use the tool, users must use the appropriate drop-down menu to pick a region, plus one or more satellite sensors. After those have been selected, the results will be displayed further down the webpage in the form of time series figures and percent frequency distributions. Further options on the page allow the users to adjust the type of data measurements displayed, the averaging time frame, and to control aspects of the X and Y axes. The checkbox option for "Plot nearby wavebands together" allows users to toggle between grouping close-by sensor bandcenters together (only relevant for wavelength-specific products, like Rrs or IOPs) or whether to plot all the wavelengths separately. The results are available for download as ASCII text files containing metadata headers followed by a delimiter-separated data matrix. They are structured in a quasi-SeaBASS file format that can be read by SeaBASS Read Tools (with some extra steps needed for interpreting the date information). Separate download buttons are available to obtain files containing either the averaged data values or the unaveraged data (the latter includes each satellite extract's filtered mean, standard deviation, and total number of valid pixels). All data points are located on separate rows, with rows clustered together according to their source (i.e., in situ and various satellite sensors.) The "data_source" column is useful for indexing because it indicates the data source of each row (e.g., "In situ", "MODIS-Aqua", "MODIS-Terra", "VIIRS-NPP", or "SeaWiFS"). In unaveraged files, the "cruise" column gives additional information about the origin of the data.
Assembly of Time Series Regions
Candidate sites for time series regions were identified based on locations where multiyear records could be obtained for in situ measurements of at least one ocean color product. For each region, geographic coordinate boundaries were defined to create sampling areas for satellite and in situ measurements. Each region's size was picked individually to balance encompassing a maximum amount of sampling data versus trying to maintain relatively homogenous waters. This process was subjective but involved combining knowledge of in situ conditions with examination of seasonal satellite records. The sampling polygon (generally a rectangle) for each region can be viewed by zooming in on the provided map. Due to localized differences, regions ranged greatly in size from small (e.g., approximately 16 km^2 at the near-shore AERONET-OC Lisco site) to much larger regions (e.g., approximately 171,000 km^2 around BATS in the Sargasso Sea). These spatial boundaries are used to compile all available satellite overpasses and all available SeaBASS validation data that have been processed and passed quality controls explained in following sections.
Assembly of In Situ Data
Please refer to the SeaBASS Validation Description for details of how in situ data are assembled for the SeaBASS Validation Search tool. This time series tool relies on the same sources, processing, and quality controls of in situ measurements, i.e., this tool and the Validation Search tool share the same in situ data. Those data include measurements archived in SeaBASS, as well as certain datasets obtained from external sources (for example AERONET-OC).
Assembly of Satellite Data
For each of the time series regions, level 1 satellite data subscriptions were established to obtain ocean color measured within the spatial boundaries of our regions of interest. For each satellite sensor's overpass, only the pixels measured within region boundaries are extracted and processed to level 2 using l2gen software with the settings associated with the latest Ocean Color reprocessing version (R2018.0 at the time of writing). The extracted pixels over a region are evaluted as candidates for inclusion in the time series based on their l2_flags quality product. Any scenes with greater than 80% total pixels flagged as clouds, ice, or high sensor zenith angle were eliminated. Next, for the remaining extracts, all pixels whose l2_flags product contained any of the following flags were masked and excluded due to quality concerns: ATMFAIL, CLDICE, FILTER, HIGLINT, HILT, HISATZEN, LAND, LOWLW, NAVFAIL, NAVWARN, or STRAYLIGHT. For the remaining pixels, filtered means were calculated (i.e., the mean of pixels remaining after eliminating all pixels more than 1.5 standard deviations from the unfiltered mean). Next, a product-specific range filter was applied to remove any values with extremely low or high values (see Table 1), and the remaining values were stored. This process is repeated for multiple satellite sensors; at the time of writing, 4 sensors were used: VIIRS-NPP, MODIS Aqua, MODIS Terra, and SeaWiFS. The system is designed to allow for reprocessing the satellite data if and when needed in the future, such as if another Ocean Color data reprocessing occurs, or if there are changes to the algorithms used for ocean color products.
Table 1. Satellite product-specific range filters to remove any extremes
| Product |
Accepted Range |
Units |
| a |
0 - 5 |
m^-1 |
| adg |
0 - 5 |
m^-1 |
| aph |
0 - 5 |
m^-1 |
| bbp |
0 - 0.3 |
m^-1 |
| AOT |
0 - 5 |
unitless |
| Chl |
0 - 300 |
mg*m^-3 |
| Kd_490 |
0 - 5 |
m^-1 |
| Rrs |
-0.01 - inf |
sr^-1 |
Assembly of Time Series
After in situ and satellite data have been assembled and processed, the results are stored in a relational database and are made available to public users via the web-based time series tool. The data are averaged into common time bins, with users able to choose between weekly, monthly, or seasonal averaging. For weekly averaging, measurements are assigned week numbers between 1 to 52 (with the 52nd bin assigned the "extra" 1-2 final days of each year.) Monthly averaging consists of assigning data to bins 1-12 based on the month number, and seasonal data consisted of assigning data from each 12 month period into bins numbered 1-4, with each bin consisting of data measured during 3 months. The seasonal bin numbers consist of the following months, 1: Dec, Jan, Feb (with Dec from the previous year), 2: (Mar, Apr, May), 3: (Jun, Jul, Aug), 4: (Sep, Oct, Nov). The mean values are calculated and displayed on the time series figures and percent frequency distributions for the user-selected year ranges. Percent frequency distributions were based on all available data measurements within the user-selected time range, not just common bins where all data sources were available.
The time series tool is updated with new data when they become available. Satellite data are added via automation as soon as refined processing versions can be processed, typically on the order of one-to-several weeks delayed from real time. New in situ measurements are also added, though more sporatically, because those measurements are available less regularly and also involve additional review and processing steps.
Notes
Note that while the Time Series Tool shares the same in situ data used by the Validation Search tool, there are some important differences between the two SeaBASS tools. The Time Series Tool displays all available in situ data whether or not there were coincident satellite data (and vice versa). That approach is different than the Validation Search which is designed to create and compare only "match-ups"; the Validation Search restricts results to where in situ and satellite data of the same measurement type were successfully paired within a +-3 hour time window using a satellite spatial area of 5x5 pixels centered over the in situ measurement (Bailey and Werdell, 2006). For most Time Series Tool regions, the satellite measurements are derived from a much larger spatial area, and measurements are averaged across longer time scales. Also, while a full suite of satellite algorithms are calculated for every possible bin in each Time Series Tool region, the in situ measurements available in most regions are limited to one or just a few different types.