340 likes | 523 Views
Consideration of Temporal and Spatial Dynamics of Vertically Migrating Harmful Algal Blooms in Support of Developing GEO-CAPE Science and Mission Requirements. Steven E. Lohrenz University of Southern Mississippi Gary Kirkpatrick Mote Marine Laboratory Oscar Schofield Rutgers University.
E N D
Consideration of Temporal and Spatial Dynamics of Vertically Migrating Harmful Algal Blooms in Support of Developing GEO-CAPE Science and Mission Requirements Steven E. Lohrenz University of Southern Mississippi Gary Kirkpatrick Mote Marine Laboratory Oscar Schofield Rutgers University
Overview • Introduction • Application of satellite ocean color to HAB detection • Utility of geostationary ocean color for HAB detection • Objectives • Temporal sampling scales • Radiative transfer modeling of simulated Rrs • Surface correlation lengths scales • Approach • Drifter studies and time-series optical profiling and discrete sampling • Radiative transfer modeling • Underway continuous sampling • Results and Discussion • Conclusions and recommendations
Introduction • The Geostationary Coastal and Air Pollution Events (GEO-CAPE) mission will provide high spectral, spatial and temporal resolution imagery in coastal waters of the continental U.S. • Blooms of harmful algae represent a significant and expanding threat to human health and fisheries resources, particularly in coastal waters subject to the influence of anthropogenic eutrophication • In the Gulf of Mexico, the red tide dinoflagellate species Kareniabrevis forms recurrent blooms off the west Florida coast • While satellite ocean color has been useful in aiding ground-based surveys for detection and monitoring of K. brevis blooms, the utility for bloom detection has been limited by relatively coarse scale temporal and spatial resolution
Objectives • A systematic analysis of in situ observations is needed to define temporal and spatial scales required to resolve bloom dynamics and provide diagnostic criteria for discrimination of vertically migrating HAB phenomena • Here we address the following objectives: • Examine temporal variations in upper water column population distributions and associated inherent optical properties (IOPs) to identify critical temporal sampling scales needed to resolve variations • Use the Hydrolightradiative transfer modeling software in conjunction with measured and modeled IOPs to develop simulated datasets of water leaving radiance in relation to bloom dynamics • Examine length scales of bloom features in relation to sensor spatial resolution
Approach • Cruise period: 10/1 – 10/5/2000 • Study area with drift track overlaid on SeaWiFS Ocean Color
Approach • Periodic surface sampling was used to identify an area of high concentrations of K. brevis • Near-surface, Davis/CODE-type drifters were used to track bloom patch • Time-series optical profiling and discrete sampling of population density and size distribution CTD and water sampling ac-9 optical profiler surface drifter
Approach K. brevis PD, PSD K. brevis n‘(l), n(l) MIE Theory • Radiative transfer modeling (Hydrolight) CDOM a440, S Chl a(l), b(l), bb(l) aph* (K. brevis) Hydrolight Hyperspectral Rrs(l)
Approach • Underway surface sampling of chlorophyll fluorescence to resolve surface correlation length scales
Results • Temporal variations in IOPs
Results • Temporal variations in IOPs
Results • Correlation between IOPs and cell densities
Results • Temporal variations in K. brevis population density in upper 2 m • Need for multiple daily images
Results • Simulated Rrs in relation to bloom dynamics using three scenarios • Uniformly mixed water column with varying K. brevis concentrations • Surface layer of varying concentrations • Deep subsurface layer
Results • Used measured IOPs and known optical properties and size distributions of K. brevis(Mahoney, 2003; Craig et al., 2006) as basis for developing scenarios
Results • Comparison of modeled (8 x 105 cells L-1) vs. measured b for upper 3m
Results • Absorption for CDOM and cells estimated using MIE-modeled absorption signature for K. brevis and least squares fit to ac-9 absorption to derive CDOM
Results • These inputs were used to model Rrs for the different scenarios
Results • Comparison to measured Rrs – underestimation of bbp? modeled measured
Results • Third objective was to examine spatial correlation length scales for bloom features
Results • Cruise track overlaid on SeaWiFS pixels for 3 Oct 2000 image
Results • Fluorescence contour for cruise track from same period
Results • We examined correlation length scales of two transects
Conclusions • Temporal variations in HAB bloom migration necessitates acquisition of multiple daily images to resolve and detect • Simulations of Rrs for different HAB vertical distributions showed patterns consistent with prior published work and distinct patterns for different HAB vertical distributions • Correlation length scales for underway were strongly dependent on direction with shortest length scales less than 0.5 km
Effect of Vertical Migration • Reflectance variations Satlantic HyperTSRB Schofield et al., 2006
Correlation Length Scales for Different Biological Properties Results from Mackas et al. (1984) based on Optical Plankton Counter and fluorescence surveys. Phytoplankton biomass scales on the order of 4-7 km, however, sampling resolution was only ~1km!
Horizontal length scales from Autonomous Underwater Vehicle Observations Results from Moline et al. (2005). Data were fit to a Generalized Additive Model and smoothed using a loess smoothing function. Sensors included CTD, optical backscatter (OBS), chlorophyll fluorescence (FL) , and bioluminescence (BL).
Horizontal length scales from Autonomous Underwater Vehicle Observations Moline et al. (2005). Lengths scales based on variogram analyses ranged from the 50-300 m.
Fractal Analysis • Fractal analyses of chlorophyll fluorescence reveal break in scaling at ~100 m (characteristic planktoscale) (Lovejoy et al., 2001) • Variability at all scales (Lovejoy et al., 2000) • Remote sensing algorithms are strongly scale/resolution dependent
Bissett et al.: Approach • Hyperspectral dataset • PHILLS 2 during the 2001 HyCODE LEO-15 • Spectral data at 9 m resolution • Length scales determined by PCA analysis of spectral properties and comparative analysis of relationships of covariance to random noise levels
Bissett et al.: Conclusions • Ground Sample Distance of 50-200 m between 1-10 km of shore • Smaller scales may be needed within 1 km • Offshore there is a difference in optimal GSD depending on whether multispectral or hyperspectral dataset is used • Multispectral suggests 1 km may be adequate • Hyperspectral suggests higher resolution may be necessary (features not apparent in multispectral)