At the broadest scales there is a need to be able to observe coral bleaching over very wide areas, in all localities across the planet. The only effective means of “seeing” reefs at these broad scales is using remote sensing with satellite platforms or aerial surveys. Bleached corals can hardly be mistaken in the field, but this also translates to a very distinctive spectral signature that may be visible from remote platforms (Holden and LeDrew 1998; Call et al. 2003). In reality, the practical challenges of remote detection of coral bleaching remain considerable. Coral reefs present highly heterogeneous substrates – even up close, most are a complex patchwork of coral, algae, sponges and other surface cover. All but the highest resolution remote sensing platforms are sampling areas (pixels) thatinclude a very broad mix of reflecting surfaces. This challenge is further compounded by the influence of differing depths of water column – in clear waters it is possible for the differentiation of marine features to 20 m, but it is a challenge to differentiatemany features beyond that. And with most reef corals growing on sloping substrates, considerable variation in depth and reflectance can occur, even within the space of individual pixels. 
 The most widely used remote platforms for general reef mapping are those that allow coverage of relatively large areas relatively cheaply, typically Landsat an SPOT, although the higher resolution of IKONOS clearly enables more accurate featur assessment and classification (Andréfouët et al. 2003; Mumby et al. 2004).
Coral bleaching can be a short-lived phenomenon; and its spatial appearance can vary considerably. Some bleaching events are comprehensive and tightly synchronised (i.e., most species fully bleached at the same time) and these are likely to be easier to detect, particularly in areas of high coral cover. Where only some colonies are bleached, or where the loss of colour in bleached colonies is only partial, detection becomes increasingly challenging.High spatial resolution is undoubtedly the most critical factor in helping to disaggregate
the complex patchwork of substrate which typifies most coral reefs. Further improvement in bleaching detection ability can be achieved with finer spectral resolution (i.e., more and narrower widths of wavelength bands in the sensor); and with improved radiometric resolution (increasing the possible number of grey levels in the image). Temporal return is also critical: corals can shift from bleached to recovered in just a few weeks and dead corals become overgrown with algae in even shorter time-frames. Differential susceptibilities by different species or in different depths means that the “peak” of a bleaching event may only last a few days, although more typically it will last 2–4 weeks. With regular cloud-cover in the tropics a return of, say, 2 weeks for a sensor may be insufficient to capture a bleaching event.The challenge of using satellite platforms, even high-resolution systems, is thus considerable. Andréfouët et al. (2002) tried to assess the optimum resolution for remote sensing of bleached corals using aerial photographs taken during the 1998
bleaching event and subjecting these to interpretation at varying resolutions. The authors noted a rapid tailing off of detection ability with resolutions increasing from 10 cm. As a crude guide, resolutions closest to that of the mean colony size will be most accurate, but resolution up to 1 m may still give some ability both to detect bleaching and to estimate variance between locations. At their study sites on the Great Barrier
 Reef, Andréfouët et al. (2002) also compared satellite-derived images taken before, during and after the 1998 bleaching using 20 m and 10 m resolution imagery, but showed a complete inability to detect even the fairly major bleaching of 1998.Others had more success with these sensors. Philipson and Lindell (2003) showed at least basic detection of a very large-scale bleaching event in Belize using the 24 m resolution IRS LISS-III platform and suggested that much better detection should be possible with, for example IKONOS. Elvidge et al. (2004) showed very good detection with IKONOS imagery on the Great Barrier Reef, but point to the need for a pre-bleaching reference image. Yamano and Tamura (2004) and Graham et al. (2006) were also able to show detection of severe bleaching at Ishigaki Island in Japan, but only in shallow, coral-rich areas. The overall conclusion from these efforts is that bleaching detection at regional to global scales is still not possible. Improvements in availability of high-resolution imagery, notably reduction in the cost of imagery, may help but there are calls for new sensors to specifically target and improve spectral resolution for those wavelengths that might be able to differentiate marine features (Philipson and Lindell 2003).
The finer resolution of aerial photography offers much improved performance compared with current satellite platforms, but at considerably increased cost and with considerable challenges in surveying more remote reef systems. 
Some of the most extensive aerial survey work has been conducted by Ray Berkelmans in overflights over large parts (almost 25%) of the Great Barrier Reef during the bleaching events of 1998 and 2002 (Berkelmans and Oliver 1999; Berkelmans et al. 2004). This work was undertaken from a fixed-wing aircraft at a height of 160 m. These surveys covered over 2000 km of coast and provided critical information on two very large bleaching events that could not have been gathered either from satellite or field-based observations.
These same studies also showed that the timing of such surveys is critical. The best predictor of maximum bleaching levels was found to be the maximum sea surface temperature (SST) over a 3-day period, rather than estimates of longer periods of perhaps less extreme temperature anomalies (Berkelmans et al. 2004). Of course this may not be the case everywhere, but it presents further challenges to those seeking to accurately measure maximal levels of bleaching.
Remote Sensing of Indicators of Bleaching Likelihood
Proxy measures of likelihood of bleaching have been available for some years using very low resolution data on SST. Using night-time only SST records at 50 km resolution, the National Oceanic and Atmospheric Administration (NOAA)’s National Environmental Satellite, Data, and Information Service (NESDIS), with its Oceanic and Atmospheric Research offices, developed a number of tools to predict bleaching likelihood (Strong et al. 1997; Strong et al. 2004; NOAA/NESDIS 2006; Chap. 4). Using data generated from an Advanced Very High Resolution Radiometer (AVHRR), SST measures are gathered twice weekly in near-real-time. These have been useto generate, inter alia, a measure of coral bleaching HotSpots (which are simply areas where the SST is at least 1°C above the mean maximum summertime temperature) and degree heating weeks (DHW), which is an index that summarises both longevity of anomaly and strength (size of temperature deviation).It must be recognised that SST and DHW can only provide approximate pointers to conditions conducive to coral bleaching. Even so, both have shown themselves to be valuable predictors, particularly for the more extreme events. Numerous studies have shown their general validity (Sheppard 1999; Spencer et al. 2000) while a number of more specific experiments have been undertaken to improve and refine recording (see, e.g., http://coralreefwatch.noaa.gov/satellite/publications.html). McClanahan et al. (2007) compared the NOAA data with SST data available from the Joint Commission for Ocean and Marine Meteorology (JCOMM) and found the latter, which is satellite-derived but with corrections based on buoy and ship-based observations, to have a slightly better predictive capacity. As our knowledge of the thresholds for bleaching in different areas improves, the predictive capacity of such
measures may be further improved (see also Chap. 4). 
Various modifiers appear to strongly influence the role of SST in determining bleaching likelihood even at broad scales. Wooldridge and Done (2004) were able to further refine the predictive capacity of such approaches using a Bayesian belief network approach and found that bleaching susceptibility of the Great Barrier Reef in 2002 could be best predicted with a combination of: “site’s heat stress in 2002
(remotely sensed), acclimatization temperatures (remote sensed), the ease with which it could be cooled by tidal mixing (modeled), and type of coral community present”. McClanahan et al. (2007) also show a significant influence of past bleaching impact, reducing the bleaching likelihood in some areas.During an actual warming event, the occurrence or degree of bleaching is further influenced by a range of other factors such as solar insolation and sea state, while different corals typically show different susceptibilities to bleaching (Marshall and Baird 2000; McWilliams et al. 2005). The important role of solar insolation at broad scales was further corroborated by observations of heavy and near continuous cloudcover during critical periods of high temperature, which may have prevented bleaching in the Society Islands (Mumby et al. 2001) and in Mauritius (Turner 1999). Even with refined models, summary data at 50 km resolution is clearly insufficient
to show the fine-scale patterns of variation which certainly occur. Berkelmans et al. (2004) observed changes at a scale of tens of kilometres, indicating local-scale variance, perhaps linked to oceanographic or weather patterns (e.g., upwelling or persistent cloud-cover adjacent to islands). Proximity to land will have further
influence on water temperatures and bleaching likelihood – through direct shading by high terrestrial features, through runoff and through influences on water flows – but these will not be picked up in very low resolution SST data. Others have noted even finer-scale variance in bleaching tolerance or survivorship linked to shading, aspect, or water flows (Spencer et al. 2000; West and Salm 2003). At very broad scales, SST data provide a useful indicator of bleaching likelihood, and it seems that various refinements could further improve such models. At the same time, however, the finer-scale variance in bleaching events needs to be documented
and understood. Such patchiness may be of critical importance in both recovery and adaptation (Grimsditch and Salm 2006).
Summary of Remote Sensing Tools
– There are considerable challenges to using satellite sensors. Generally, very fine resolution is critical, and clearly detection will be better in areas where coral cover is high. There are, as yet, no broadly accepted tools. 
– Aerial photography is highly successful not only in detection, but in quantifying and mapping bleaching impacts.
– Perhaps the most reliable tool is the use of temperature anomalies with AVHRR data. The coral bleaching HotSpots and DHW statistics provide powerful measures that also have the advantage of being consistent and comparable between regions and over time.
– These same data are also used in post hoc studies as a proxy measure of presumed impacts.