Fig 1.
Map of the study region including major cities, rivers, the masking of the reef matrix and the geographic boundaries of the five subregions for this study.
Light blue shading indicates the reef lagoon area that was the focus of this study. Black circles indicate the locations at which validation samples were taken.
Fig 2.
Time series of monthly averaged MCIPI for the five subregions (ordered from north to south).
Vertical dashed lines represent 6-month periods for all plots.
Table 1.
Annual distribution of Trichodesmium field records noted as “present” with matching MERIS images (N = 25): columns 1–2: number of recorded field observations for each month; col. 3: corresponding frequency of these observations; cols. 4–5: MERIS retrieval statistics showing whether recorded field observations were retrieved by a MERIS image or not in the matching image pool.
Table 2.
Statistics for the MERIS imagery dataset used in this study.
Number of images for each subregion with MCIPI > 10 for the five regions and percentage these images account for as a function of the total number of images.
Fig 3.
Decadal seasonal amplitudes of MCIPI (black) and valid satellite observations (gray) for the five subregions.
The sum of all the seasonal amplitudes over the course of a year for each subregion is equal to 12. The strength of the seasonal signal explains the differences in bar sizes.
Fig 4.
MERIS scenes featuring the largest surface expressions (in size and number of MCIPI) occurring in the lagoonal waters of each subregion between April 2002 and April 2012.
The reef matrix and the land are masked in gray.
Fig 5.
Monthly climatology of surface bloom occurrences (%MCIPI) for FIT over the period April 2002-April 2012.
Table 3.
Monthly and yearly time-series statistics:
Maximum MCIPI computed for each subregion when monthly and yearly aggregations were considered; dates are indicated between brackets. Decadal MCIPI (last column) and associated percentage of the total (all regions considered) were based on monthly aggregations over the 10-year period.
Fig 6.
Beta Generalized Additive Model (GAM) outputs modeling the decadal trends in Trichodesmium blooms in each subregion.
The y-axis is the proportion of lagoonal waters affected by bloom events in each subregion over the course of the 10-year period. The units are bloom pixels per total (cloud-free) pixels. Confidence intervals for each model are shown as gray bands, and the statistical significance of the trends (p-value) is shown.
Table 4.
Average and maximum wind speed (in m.s-1) with corresponding surface aggregation sizes (same day; based on daily data).
Table 5.
Average and maximum wind speed (in m.s-1) two days before MERIS scenes of surface aggregation (based on daily data).