Chlorophyll Bloom Dynamics: Datasets

Overall Description:

This archive contains processed satellite chlorophyll-a (CHL) data products, statistical trend analyses, and generalized additive model (GAM) outputs used to investigate interannual variability and long-term trends in summertime phytoplankton blooms in the North Pacific Ocean surrounding the Hawaiian Islands. The study region spans approximately 18°N–35°N and 170°W–130°W. Data are derived from the Copernicus GlobColour merged satellite ocean color product (Level 3, daily, 4 km resolution). The Hawaiian Islands Contiguous Zone (CZ) was masked to remove nearshore coastal effects. All processing was performed in Python and R.

Individual file descriptions:

chla_day_deseason_detrend_20260422.nc NetCDF4 file containing daily satellite chlorophyll-a anomalies with both seasonal and long-term trends removed. The seasonal cycle was removed by subtracting a per-pixel monthly median climatology computed across the full record. The long-term trend was removed by subtracting a 1825-day (5-year) running mean from the seasonal anomaly at each pixel. Variables include: CHL_anom (deseasonalized and detrended chlorophyll anomaly, mg m⁻³), CHL_trend (the extracted long-term trend, mg m⁻³), and CHL_clim (12-month median climatology, mg m⁻³). Spatial coverage: global at source resolution (4 km). Temporal coverage: September 1997 to October 2025, daily. Source data: Copernicus GlobColour Bio-Geo-Chemical L3 daily multi-sensor merged product.

bloom_summary_2p_1mad_20260423.csv CSV file containing summary statistics for 28 annual summertime phytoplankton bloom events (1998–2025). Each row represents one bloom defined by manually identified start and end dates. Columns include: start and end dates of the bloom window; bloom center date, latitude, and longitude estimated via chlorophyll-weighted center of mass (scipy.ndimage.center_of_mass); bloom magnitude defined as the 95th percentile of raw chlorophyll values at anomalously high pixels; maximum bloom spatial extent (km²) on the day of peak area, accounting for latitude-dependent pixel size; date of maximum area; and percent cloud cover on the peak-area day. Bloom pixels were identified where the deseasonalized/detrended chlorophyll anomaly exceeded the per-pixel median plus one scaled median absolute deviation (MAD × 1.4826). The study region spans 18°N–35°N, 170°W–130°W with the Hawaiian Islands Contiguous Zone masked.

chl_gam.Rdata R data file containing a fitted generalized additive model (GAM) of the form CHL ~ SLA + s(time) + s(lon, lat), where CHL is satellite chlorophyll-a concentration, SLA is sea level anomaly, time is day of year, and lon/lat capture spatial structure. The model was fit using the mgcv package in R with a Gamma family and inverse link function on a random subsample of 100,000 observations from the 2018 bloom period. Predictor data were resampled to the SLA grid (lowest resolution) using nearest-neighbor interpolation, and the Hawaiian Islands Contiguous Zone was masked prior to fitting.

sla_gam.Rdata R data file containing a fitted generalized additive model (GAM) of the form SST ~ SLA + s(time) + s(lon, lat), where SST is sea surface temperature, SLA is sea level anomaly, time is day of year, and lon/lat capture spatial structure. The model was fit using the mgcv package in R with a Gaussian family (default) on a random subsample of 100,000 observations from the 2018 bloom period. Predictor data were resampled to the SLA grid using nearest-neighbor interpolation, and the Hawaiian Islands Contiguous Zone was masked prior to fitting.

senslope_nomask_20260328.nc NetCDF4 file containing per-pixel Sen's slope trend estimates for summertime (June–October) chlorophyll-a anomalies. Variables include: slope (Sen's slope in mg m⁻³ per day), intercept, and upper and lower 95% confidence bounds on the slope. Sen's slope was computed using scipy.stats.mstats.theilslopes at each grid cell, with actual time values (days) as the independent variable to ensure correct physical units. Pixels with fewer than 10 valid observations were excluded. Spatial coverage: study region at 4 km resolution. To convert slope to annual units, multiply by 365.25. The significance of each pixel's trend can be assessed using the companion Kendall tau p-values (computed separately).

mag_area_regression_results_20260422.txt Text file containing regression results for bloom magnitude and maximum bloom area as functions of year. Two model types were fit: (1) Gamma generalized linear models (GLM) with log link using statsmodels in Python, and (2) nonparametric Sen's slope with Kendall tau significance test using scipy. The Gamma GLM coefficients are on the log scale; percent change per year can be computed as (exp(slope) − 1) × 100. The file reports coefficients, standard errors, p-values, Kendall tau statistics, and 95% confidence intervals for both bloom magnitude (95th percentile of raw CHL at extreme pixels, mg m⁻³) and maximum bloom area (km²).

chla_day_summer_mean_20260423.nc NetCDF4 file containing the temporal mean of deseasonalized and detrended daily chlorophyll-a anomalies for summer months (June–October) across all years in the record. The variable CHL_anom_mean (mg m⁻³) represents a single 2D map (latitude × longitude) computed as the nanmean across all summer days. This product characterizes the mean spatial pattern of summertime chlorophyll anomalies in the study region. Source data: chla_day_deseason_detrend_20260422.nc, subset to months 6–10.