The construction of causal networks to estimate coral bleaching intensity Lilian Anne Krug a, b, * , Douglas Francisco Marcolino Gherardi a , José Luís Stech a , Zelinda Margarida Andrade Nery Leão c , Ruy Kenji Papa Kikuchi c , Estevam Rafael Hruschka Junior d , David John Suggett e a National Institute for Space Research, Remote Sensing Division, Avenida dos Astronautas,1758, Zip Code: 12227-010 São José dos Campos, SP, Brazil b University of the Algarve, Centre for Marine and Environmental Research, Campus de Gambelas, Zip Code: 8005-139 Faro, Portugal c Federal University of Bahia, Institute of Geosciences, R. Barão de Jeremoabo s/n, Zip Code: 40170-115 Salvador, BA, Brazil d Federal University of São Carlos, Computer Science Department, Rod. Washington Luis km 235, Zip Code: 13565-905 São Carlos, SP, Brazil e University of Essex, Department of Biological Sciences, Wivenhoe Park, Colchester CO43SQ Essex, United Kingdom article info Article history: Received 30 January 2012 Received in revised form 8 January 2013 Accepted 9 January 2013 Available online 9 February 2013 Keywords: Bayesian network Coral reef Coral bleaching Remote sensing Environmental variability South Atlantic coral reefs abstract Current metrics for predicting bleaching episodes, e.g. NOAA’s Coral Reef Watch Program, do not seem to apply well to Brazil’s marginal reefs located in Bahia state and alternative predictive approaches must be sought for effective long term management. Bleaching occurrences at Abrolhos have been observed since the 1990s but with a much lower frequency/extent than for other reef systems worldwide. We constructed a Bayesian Belief Network (BN) to back-predict the intensity of bleaching events and learn how local and regional scale forcing factors interact to enhance or alleviate coral bleaching specific to Abrolhos. Bleaching intensity data were collected for several reef sites across Bahia state coast (w12  e20  S; 37  e40  W) during the austral summer 1994e2005 and compared to envi- ronmental data: sea surface temperature (SST), diffuse light attenuation coefficient at 490 nm (K 490 ), rain precipitation, wind velocities, and El Niño Southern Oscillation (ENSO) proxies. Conditional in- dependence tests were calculated to produce four specialized BNs, each with specific factors that likely regulate bleaching intensity. All specialized BNs identified that a five-day accumulated SST proxy (SSTAc5d) was the exclusive parent node for coral bleaching producing a total predictive rate of 88% based on SSTAc5d state. When SSTAc5d was simulated as unknown, the Thermal-Eolic Resultant BN kept the total predictive rate of 88%. Our approach has produced initial means to predict beaching intensity at Abrolhos. However, the robustness of the model required for management purposes must be further (and regularly) operationally tested with new in situ and remote sensing data. Ó 2013 Elsevier Ltd. All rights reserved. 1. Introduction Coral bleaching can occur as a response to stressful environ- mental conditions induced by direct local pressures as well as in- direct regional/global climatic change (Carpenter et al., 2008; Suggett and Smith, 2011). Although anomalous surface water heating is acknowledged to be the most important stressor, the interplay of additional environmental variables moderates the net bleaching response (Glynn, 1993; Fitt et al., 2001). Indeed, most observations to date demonstrate that corals are most vulnerable to bleaching during periods of clear sky, calm sea, weak winds and clear water that maximise the amount of heat and light reaching the coral soft tissue (Baker et al., 2008; Brown, 1997; Glynn, 1993). However, such patterns may also be influenced by genetic com- position of the corals, symbiotic algae or hosts (see Baker et al., Abbreviations: AGRRA, Atlantic and Gulf Rapid Reef Assessment; BN, Bayesian networks; BNPC, Belief Network Power Constructor; BSST, best sea surface tem- perature; CI, conditional independence; CPT, conditional probabilities table; K 490 , diffuse light attenuation coefficient at 490 nm; DAG, directed acyclic graph; EM, expectationemaximization; MaxSST, maximum sea surface temperature; jWj, mean surface wind fields; V, meridional wind; MEI, multivariate El Niño index; PCA, principal component analysis; PPT, rain precipitation; SST, sea surface tem- perature; SSTAc5d, sea surface temperature accumulated in five days; U, zonal wind. * Corresponding author. University of the Algarve, Centre for Marine and Envi- ronmental Research, Campus de Gambelas, Zip Code: 8005-139 Faro, Portugal. Tel.: þ351 289 800 900x7372; fax: þ351 289 800 100. E-mail addresses: lakrug@ualg.pt (L.A. Krug), douglas@dsr.inpe.br (D.F.M. Gherardi), stech@dsr.inpe.br (J.L. Stech), zelinda@ufba.br (Z.M.A.N. Leão), kikuchi@ ufba.br (R.K.P. Kikuchi), estevam@dc.ufscar.br (E.R. Hruschka), dsuggett@ essex.ac.uk (D.J. Suggett). Contents lists available at SciVerse ScienceDirect Environmental Modelling & Software journal homepage: www.elsevier.com/locate/envsoft 1364-8152/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.envsoft.2013.01.003 Environmental Modelling & Software 42 (2013) 157e167