Animal-borne sensors successfully capture the real-time thermal properties of ocean basins

Climate change is perhaps the most pressing and urgent environmental issue facing the world today. However our ability to predict and quantify the consequences of this change is severely limited by the paucity of in situ oceanographic measurements. Marine animals equipped with sophisticated oceanographic data loggers to study their behavior offer one solution to this problem because marine animals range widely across the world’s ocean basins and visit remote and often inaccessible locations. However, unlike the information being collected from conventional oceanographic sensing equipment, which has been validated, the data collected from instruments deployed on marine animals over long periods has not. This is the first long-term study to validate in situ oceanographic data collected by animal oceanographers. We compared the ocean temperatures collected by leatherback turtles (Dermochelys coriacea) in the Atlantic Ocean with the ARGO network of ocean floats and could find no systematic errors that could be ascribed to sensor instability. Animal-borne sensors allowed water temperature to be monitored across a range of depths, over entire ocean basins, and, importantly, over long periods and so will play a key role in assessing global climate change through improved monitoring of global temperatures. This finding is especially pertinent given recent international calls for the development and implementation of a comprehensive Earth observation system (see http://iwgeo.ssc.nasa.gov/documents.asp?s=review) that includes the use of novel techniques for monitoring and understanding ocean and climate interactions to address strategic environmental and societal needs. Acknowledgments The sea surface temperatures we used were derived from NOAA’s GOES daily SST satellite data. They are processed and made available at http://www.seaturtle.org/maptool/. The source data have a nominal spatial resolution of 6 km and have been validated to be within 0.5°C of actual SST. We wish to acknowledge use of the Maptool program for some analysis and graphics in this paper. We are grateful to the Ministry of Agriculture, Forestry, Land, and Fisheries in Grenada for permission to attach transmitters to turtles on nesting beaches; Ocean Spirits Inc. for logistical help in the field; the late Bob Randig and Moyra Hays for help with the attachment system. Funding was provided by grants to GCH from the Natural Environment Research Council of the UK (NERC). Limnol. Oceanogr.: Methods 3, 2005, 392–398 © 2005, by the American Society of Limnology and Oceanography, Inc. LIMNOLOGY and OCEANOGRAPHY: METHODS the atmosphere will be 30 times greater than that in the ocean. Clearly, small changes to the energy content of the oceans could have considerable effects on global climate and detecting these small changes undoubtedly requires that ocean temperatures be measured accurately and with high precision. Indeed, recognition and an increasing appreciation of the dominant role that the world’s oceans play in climate regulation (e.g., Alexander et al. 2002; Gregg et al. 2003; Sutton and Allen 1997) has led to the need for increasingly finescale oceanographic information. So great is the requirement for detailed climate information that the demand by far outweighs the availability, because of, amongst other reasons, the high cost of the instrumentation required to collect oceanographic information and the relatively sparse coverage by monitoring devices in the world’s oceans (Anonymous 2001). One solution to resolving this paucity of data is to use the fine-scale oceanographic information that is collected when investigating the distribution and pelagic behavior of marine animals (Hooker and Boyd 2003; Lydersen et al. 2004, 2002). The concept of using animals as oceanographic platforms is not new (Boehlert et al. 2001; Weimerskirch et al. 1995), but has only recently become feasible because the technological tools to produce effective monitoring equipment have only recently been developed. These include small, low power microelectronics and computing techniques (Fedak 2004; Fedak et al. 2002; Kooyman 2004). These refinements have created a synergism between the biological studies of marine vertebrates and oceanographic studies (Lydersen et al. 2004) that allow us to explore the links between animal behavior, foraging activity, and oceanographic features, such as frontal systems, local eddies, and thermoclines in real time while the animals are still at sea. It follows that larger marine species in particular (because they are able to carry larger devices) may be used as platforms of opportunity to gather detailed oceanographic information especially, because these animals can collect information from logistically difficult areas, at fine temporal and spatial resolution, and at relatively low cost (Lydersen et al. 2004). Validation of the quality of environmental data collected by “animal oceanographers” is central to this emerging area of ocean monitoring. Some previous work using data loggers deployed on animals that were subsequently recaptured to allow logger removal has addressed this issue of the data reliability (e.g., Boehlert et al. 2001). This approach of using data loggers can only be used with certain animals because of the necessity for recapture, but has the advantage that data loggers can record environmental data at high frequency and then be recalibrated when they are recovered. Relaying environmental data remotely from free-ranging animals that are not recaptured has potentially greater utility in that a wider set of species can be used. However, there are two key obstacles that need to be overcome with this approach. First there is limited bandwidth available within the most widely used satellite system, Service Argos (http://www.cls.fr/manuel/default. htm), so that elegant data compression tactics are needed to recover large amounts of data. Second, if sensors are not recovered, they cannot be recalibrated and hence require long-term stability. Here we tackle these two issues and show that high quality temperature data can be relayed via satellite over long periods from broadly ranging animals. Materials and procedures During May to July 2003 seven free-living leatherback turtles (Dermochelys coriacea) were equipped with state-of-the-art Satellite Relay Data Loggers (SRDLs) to study their behavior and to demonstrate the utility of ocean temperature measurements made in this fashion (see Hays et al. [2004] for attachment protocols). In addition to the primary function of gathering data about turtle behavior, the SRDLs were programmed to measure temperature upcasts on the deepest dive in each 12-h period (provided that the dive reached a depth of at least 25 m). To measure temperature, the SRDL contains a bead-in-glass thermoprobe (G.E. Thermometrics) mounted in the water flow at the front of the speed sensor. These devices are aged by baking at 300°C to accelerate the drift that is inherent in silicon thermistors. Each device is then calibrated at 0°C, 10°C, and 25°C to produce coefficients of the Steinhart-Hart relationship between log (resistance) and temperature (Steinhart and Hart 1968). To ensure that the resistance of the thermistor is faithfully captured by the SRDL before being converted to temperature, a further calibration is performed which uses fixed precision resistors to identify and remove the effect of component variations in each SRDL’s analogue measurement circuitry. The manufacturer’s stated time constant for the thermoprobe (plunge into still water) is 300 ms. Temperature and pressure were sampled at 1 Hz, and the results averaged into 1 dbar bins (1 dbar increase in pressure is equivalent to approximately 1 m of seawater). These raw data were then processed according to the method used for XBT floats: that is, a 5-point median filter was applied to remove outliers, followed by a Hanning filter (Orstom 2000). Twelve depth-temperature points were then selected to approximate the cast by the broken-stick algorithm, and these coordinates were encoded along with a timestamp to fit in a single 31-byte ARGOS message. The resulting data string was stored in a buffer where it was made available for transmission for up to 5 d. The SRDL is highly configurable: in particular, it allows priorities to be assigned to the various data types that it collects, reflected in the volume of each type that it transmits. In this case, temperature casts represented 12.5% of the transmissions made by the SRDL. We excluded data for one of the seven SRDLs because we identified a software error, which resulted in an uplink error when the SRDL transmitted information to ARGOS. Because sampling effects coupled with a change in measurement range were possible in our data set, we restricted the validation to those times (for each of the SRDLs) when a wide range of temperatures was being recorded. McMahon et al. Turtles as climatologists