A case study in the Western Indian Ocean Modelling ecological susceptibility of coral reefs to environmental stress using remote sensing, GIS and in situ observations: A case study in the Western Indian Ocean Joseph Maina1 Valentijn Venus2 Ecological Modelling, in Review 1 Mombasa, Kenya 2.ITC, Enschede, The Netherlands
Most diverse marine ecosystems Coral Reef Ecosystems Most diverse marine ecosystems Economic value Geophysical value
Problems Decline in coral cover Ecological shift Loss of live livelihood Ecological shift Source: Gardner et al., 2003
Climate change and coral bleaching Climate models forecast: SST increased by 1oC for last 100 yrs Current increase 1-2 oC per century Corals near their thermal threshold Increased frequency and intensity of coral bleaching
Case study: Western Indian Ocean
Main objectives Relative importance of environmental variables -spatial pattern of coral bleaching Identify specific areas likely to be resilient Suitability of low-moderate spatial resolution remote sensors
Methods: research approach 2 3 1 4 5
Methods: satellite data Data Product Satellite/Sensor Spatial Resolution Time Scale Sea surface Temperature (oC) NOAA AVHRR ~4 km Monthly; 1985-2005 Chlorophyll a (mg/l) SeaWiFS ~9 km Monthly; 1997-2005 PAR (Einstein/m2/day) Ocean current (m/s) OSCAR: TOPEX/Pseidon;JASON;QuikSCAT 1o x 1o Monthly; 1992-2005 Wind speed (m/s) SSM/I (Special Sensor Microwave/Imager) 0.25o x 0.25o Weekly; 1997 to 2005 UV irradiance (Milliwatts/m2/nm ) TOMS Daily; 1996 to 2005 **28 Derived variables: long term and short term ≈ 5000 images
Satellite-in situ comparison Unpublished in situ data by Dr.Tim McClanahan, WCS
Methods: bleaching observation data 33405 colonies sampled from 66 reefs (WCS) 216 bleaching occurrence & severity point data (www.reefbase.org)
Statistical Analysis: selected Results Bleaching as a function of environmental variables Short term conditions Historical conditions R Square F Ratio Prob > F AIC 0.62 11.8 <.0001 329 Variable t Ratio SST anomaly 3.7 13.9 0.001 Wind speeds anomaly 3.6 13.3 SST Hotspot 3.3 11.2 Currents anomaly -3.1 9.6 0.003 UV radiation -2.8 8 0.006 2.8 7.8 0.007 Surface currents 1.9 3.8 0.057 PAR anomaly -1.8 3.2 0.078 R Square F Ratio Prob > F AIC 0.56 18.9 <.0001 350 ` Variable t Ratio Meridional currents -5.17 26.8 UV radiation -4.42 19.6 Wind speed -3.7 13.7 SST CV 2.46 6 0.02 SST hotspot 2.25 5.1 0.03
Reef base data: Mean against observed bleaching
Modeling Susceptibility – concept High Low Resistance + Tolerance Recovery = Resilience Adopted from Obura 2005
Methods: Long term conditions
Methods: Fuzzy logic functions
Normalized parameters using fuzzy logic Methods: Modeling Susceptibility Normalized parameters using fuzzy logic Susceptibility from Wind velocity
Integration of parameters – model 1 Spatial Principal Component Analysis Selected PC’s I II III IV V VI VII Contribution ratio (%) 56.5 13.4 8.4 7.7 4.2 3.7 2.7 Cumulative contribution (%) 69.9 78.3 85.9 90.1 93.8 96.49
Integration of parameters: model 2 Number of layers Pixels within a each layer
Cosine amplitude – pair wise relation strength Integration of parameters (2) Cosine amplitude – pair wise relation strength P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 Eigen vectors/Scores Max SST 1 0.098 SST 0.92 0.092 UV 0.89 0.87 0.091 Chlorophyll 0.51 0.38 0.43 0.08 CV 0.9 0.91 0.59 0.097 Bleaching model 0.21 0.01 0.61 0.11 0.055 Wind speed 0.79 0.83 0.2 PAR 0.7 0.37 0.78 0.02 0.66 0.084 Zonal currents 0.64 0.65 0.69 0.34 0.03 0.63 0.074 SST Hotspot 0.48 0.42 0.39 0.41 0.28 0.081 Meridional currents 0.49 0.55 0.35 0.57 0.24 0.5 0.44 0.073 Slope 0.54 0.62 0.47 0.082
Results: Susceptibility Models Kappa statistic = 0.7
Evaluating SM: Mortality from 1998 ENSO Adj R2 = 0.22 P = 0.03 Adj R2 = 0.17 P = 0.06 unpublished data mortality data by Mebrahtu Ateweberhan, PhD
Results (2): management implications More than half IUCN category I& II Marine Protected Areas located in moderate to high
Key Findings: summary Long term and short term environmental conditions predicted coral bleaching Good correlation between susceptibility and mortality More than half IUCN no take zones located in moderate-highly susceptible areas Moderate resolution data suitable for meso-scale studies
RS data/model limitations Uncertainties: spatial and temporal boundaries Assumes strong connectivity – interpolation of data to coastal areas Bulky data - processing time Delivery formats - (AMIS, ASI?) Uncertainty: expert knowledge & ecological data
Recommendations Long time series data Moderate to high resolution data for local scale studies – hierarchical modeling (AMIS, ASI) Simplify data access methods/conventional formats (AMIS, ASI) Closed area management should review status of MPA’s
Thank you ‘All Models Are Wrong’ Acknowledgements: EU Erasmus Mundus program Consortium Directors: Prof’s: Peter Atkinson, Peter Pilesjo, Katarzyna Dabrowska, and Andrew Skidmore Mr. Valentijn Venus, ITC, The Netherlands Dr. Chris Marnnaettes, ITC Dr. Colette Robertson, NOCS, Southampton, UK Mr. Bas Beistos, ITC Mr. Aditya Singh, UoF, USA Dr. Tim McClanahan, WCS, NY, USA Dr. Jay Herman, NASA, USA Mr. John Gunn, Earth and Space Research, USA Mr. Ruben van Hooidonk, Purdue University, USA Dr. Mebrahtu Ateweberhan, GEF-World bank project, Mombasa, Kenya Dr. Ruby Moothien-Pillay, MOI, Mauritius Dr. Graham Quartley, NOCS, Southampton, UK Dr. Valborg Byfield, NOCS, Southampton, UK