1

Sensing an Experimental Forest

Processing Environments and Distributing Relations

SURROUNDED BY THE SAN BERNARDINO NATIONAL FOREST and situated within the San Jacinto Mountain Range in California, there is one particular patch of woods that is distinct in its ecological processes. This forest is equipped with embedded network sensing that digitally detects and processes environmental phenomena, from microclimates to light patterns, moisture levels and CO₂ respiration in soils, as well as the phenology, or seasonal timings, of bluebirds and auditory signatures of woodpeckers. These multiple modes of experimental forest observation are part of a test site for studying sensors in situ. A “remote sensing lab,” the University of California James Reserve is an ecological study area that has hosted field experiments since 1966. The use of this ecological study area to test electronic sensors developed through the Center for Embedded Networked Sensing (CENS) research project is at once a continuation of experimental ecological practices in this area, as well as a shift in the technologies and practices for studying environmental processes. The question that arises here is: When the ecological experiment changes, how do experiences also change?

This notion of experimenting and experiencing as springing from a shared modality is put forward by Stengers in her discussion of Whitehead, where she uses “a (French-inspired) neologism” that does not draw a “clear distinction between the terms ‘experience’ and ‘experiment’ as there is in English.” This merging of terms is also a critical way for describing the speculative approach of Whitehead, which might be characterized by a crossing-over of experience and experiment, where experimenter and experiment are part of a unified and concrete occasion.1 This point of entry is important for this discussion, as it immediately points toward a consideration of sensors not as instruments sensing something “out there” but rather as devices for making present and interpretable distinct types of ecological processes. These processes are articulated computationally, and they draw together a wide range of experiencing entities that begin to in-form new arrangements of environmental sensing. The becoming environmental of computation extends to the experiencing entities that sense and express ecological processes.

The use of wireless sensor networks to study environmental phenomena is an increasingly prevalent practice. Sensing projects encompass studies of seismic activity, the health of forests, maps of contaminant flow, and the tracking of organisms from dragonflies and turtles to seals and elephants. These projects generate sensor data that are meant to provide greater insights into environmental processes. At a time when ubiquitous computing is extending to multiple aspects of everyday life, where the Internet of Things promises to have your refrigerator communicating with supermarkets, and smart city designs propose to harvest your location data to ensure your roast-chicken dinner is prepared on time, sensing environments for ecological study is just one set of practices within a larger project of programming environments through distributed modes of computation. Sensor networks arranged over static and mobile platforms and widely distributed throughout environments are the common thread throughout these projects, but the deployment of sensors within ecological study sites has been one of the key and ongoing areas for early sensor research and development.

As discussed in the introduction to this study, although a range of research has been conducted on ubiquitous computing and the Internet of Things,2 less has been written in the context of digital media theory or science and technology studies about the ways in which understandings and practices of environmental science have shifted through sensor systems and how these shifts have also had ongoing effects on more “participatory” sensor projects. While sensors and sensor systems were initially developed for use in military contexts, wireless and embedded sensor systems have further developed through ecological study, which has in turn provided an additional basis for deploying sensor systems within social media and citizen-sensing contexts.3 This chapter focuses on the use of sensors for study in environmental science in order to consider how these science-based sensing practices might influence practices in expanded areas such as citizen sensing.

Situated within the context of these ubiquitous computing developments, this chapter specifically focuses on the distinct forms of sensing that concresce in relation to the monitoring of environmental phenomena. One key advantage that sensor systems are meant to provide is the ability to understand the complex interactions and relations within ecosystems in greater detail. Ecological relations are meant to appear in higher resolution because sensors monitor and make available aspects of environmental processes as they unfold over time rather than as more discrete moments—and because more data are available for generating models of complex interactions. This study asks how the ecologies that materialize through more continual sensor observation are not simply the result of increased data output and processing, but might also be understood as generative sensory relations articulated across humans, more-than-humans, environments, and devices. In what ways do distributed sensor technologies contribute to new sensory processes by shifting the relations, entities, occasions, and interpretive registers of sensing? How do the interpretative practices that are individuated in experimental environmental sensing test sites in-form attention to environmental problems? And what are the implications of these experimental environmental sensing arrangements as they migrate into policy and influence participatory sensing processes?

In order to consider these questions, I first give an overview of the increasing use of sensors for monitoring environments and studying environmental change. The generation of more and higher-quality data is seen as critical to developing more advanced insights into how environments are transforming, and so the sense data produced through these projects are often gathered for the purposes of advising science and policy, in addition to testing prototype computational technologies in the field. Environmental monitoring can bring with it a sense of increased responsibility; and the commonly used phrase, “all eyes on earth,” is a way of articulating the watchful concern that sensors embody and operationalize through the continual observation of environmental processes.

But sensors connect up more than just a network of human- or sensor-based eyes. This chapter draws on more-than-human theory to move beyond human-centric interpretations of computational sensing technology and engages with Whitehead’s approach to experience as something that concresces across human and more-than-human subjects. As Whitehead suggests, perceiving subjects are neither exclusively human nor pregiven, but combine as feeling entities through actual occasions.4 In this way, sensors might also be understood not as detecting substantialist external phenomena but as contributing to inventive processes for making interpretive acts of sensation possible—and for articulating environmental change and matters of concern. This is a way of saying that interpretation matters, and that experience to be interpreted concresces across multiple registers and entities. In addition, interpretation is integral to processes whereby things come to take hold as objects of relevance.5

Based on a consideration of the distinct articulations of sense across more-than-human and environmental processes, this chapter moves to focus specifically on the use of embedded networked sensors at the James Reserve ecological study site. Drawing on fieldwork carried out at this site where sensors were tested in situ, as well as a review of scientific papers and online records of sensor data, I discuss new formations of distributed sense that concresce through these experimental forms of environmental sensing. Part of the way in which sensors might be understood as operative within distinct registers of experience is as distributed computational technologies. Sensors are distributed in at least two ways: in terms of their spatial distribution, by monitoring environments in a widespread and localized way; and in terms of the distributions of experience that generate sense data and interpretations.6 If we take seriously Whitehead’s suggestion that sensing entities concresce through experiences (or prehensions) and that they are inseparable from occasions of experience, then how do experimental environmental sensor arrangements mobilize distinct sensing practices that are creative of new environmental abstractions and entities?7

As Whitehead suggests, abstractions are not separate from concrete things, but rather influence “the process of concrescence” and provide a “lure for feeling.”8 The concrescences that come together here might be understood not just as scientists-devices-flora-and-fauna but also as relations that individuate and are individuated through data sets and algorithmic processes, across sedimented environmental effects, and through responsive modes of environmental action. The coming together of an experiment presents the possibility for distinct experiences and subjects to concresce. Sensing an experimental forest is not about detecting information “out there” but about “tuning” the subjects and conditions of experience to new registers of becoming. Tuning is a way to describe the co-creation and individuation of agencies within experiments and the complex process of developing facts or matters of concern within such experiments.9 This chapter sets out to provide an understanding of the dynamic, distributed, and multiple modes of computational sensing environments that might also provide insights for more “cosmopolitical” participation, where sensing is a process of multidirectional tuning and experiencing.10 The becoming environmental of sensor-based media is then distributed to include multiple subjects, organisms and technologies, as they process their environments.

Instrumenting the Earth

The use of instrumentation within ecological study, from bird ringing to anemometers, has a longer history than the more recent use of networked sensor systems.11 However, the miniaturization and faster processing speeds of sensors have contributed to their increasing use as instruments within ecological study.12 Sensor systems—composed of relatively small-scale in situ sensors and actuators that are able to collect and transmit data through networked connections, as well as undergo remote reprogramming—have been described as nothing less than another “revolution” comparable to the rise of the Internet.13 These imagined and actual transformations involve extending computational capacities to environments through sensors, where objects and phenomena are transformed into sensor data and made manageable through those same computational architectures.

In related literature, sensor networks have also been described as a revolution in scientific instrumentation, similar to the telescope and microscope, where a new order of insights might be realized. But instead of probing outer or inner space, sensor networks operate as “macroscopes,” which enable a new way “to perceive complex interactions” through the high density and resolution of temporal and spatial monitoring data.14 While issues related to providing a reliable power source, ensuring the robustness of hardware, and maintaining the validity and manageability of large data sets remain, sensor systems present the possibility for understanding environmental processes and relations more thoroughly by providing real-time data that are more detailed than existing modes of data collection, including remotely sensed and manually gathered data that may exist at a much larger scale or more discrete moments in time. The hope is that a background of new and undiscovered relations may be connected up and made evident through these sensing devices.

A wide range and number of projects now employ sensors for environmental monitoring, from bird migration and nesting to the social life of badgers, to water quality monitoring, phenological observations, the acoustic sampling of volcanic eruptions, and the monitoring of microclimates in redwood forests.15 One of the key projects within sensor-systems development—a 2003 study of Leach’s Storm Petrels at Great Duck Island, a wildlife preserve in Maine—established that “habitat and environmental monitoring is a driving application for wireless sensor networks.”16 This sensor project employed static sensor nodes and patches, with “bur-row motes” and “weather motes”—or sensor nodes—to study the underground nesting patterns of migrating birds.

As with many similar and subsequent sensor deployments, this project produced more detailed data on previously unobserved ecological phenomena and relationships while also providing a test-bed for experimenting with the sys-tem architecture of sensor networks. The ecological relationships observed—or sensed—are in many ways coupled with the capacities of sensor networks, which similarly are adapted to and “learn” from the processes under study. The “tuning” of sensor networks may take place not just between scientists and devices but also across devices, code, and ecological processes. In this way, sensors become environmental by tuning in and developing along with the phenomena and organisms under study.

At the same time that sensor observations are intended to provide more detailed accounts of environmental phenomena on the ground, they also contribute to the building up of multiscalar and widely distributed approaches to environmental sensing, including remote sensing by satellites and airborne observations. These data are often generated across scales and derived from diverse modes of sensor input for wider and more detailed views on environmental processes and to study the effects and possible impacts of environmental change.17 Multiple “observatories,” together with long-term ecological research sites (LTERs), and the U.S. National Ecological Observatory Network (NEON), attempt to collect and synthesize sensor data across the United States. While a site-specific sensor project may study the detailed relationship between birds’ nesting behavior in relation to microclimate and multiple other environmental factors, this same study may benefit from climate data resources or may contribute to climate monitoring programs. In other words, the sense data gathered may have the potential to elucidate environmental relations within a particular area of study, as well as across expanded and yet-to-be-gathered data sets—as long as the data to be compared are of compatible formats.

Just as sensing systems are proliferating, numerous attempts are underway to amalgamate and make sense of the many forms of data—a key “cyberinfrastructure” task—since the multiple formats and provenances of data may mean that they are rendered meaningless for ongoing use and study if not consistently handled.18 Sensor-gathered data sets, which are typically “heterogeneous,” are increasingly brought together not only in larger data networks but also in mapping platforms where fine-grain sensor data provide a real-time “ground-truth” to coarser remote-sensing and field-gathered data. From Microsoft’s SenseWeb to the former DIY-sensing platform Cosm, such platforms intend to consolidate environmental sensor inputs.19 The range of possible sensor inputs is illustrated by one Microsoft diagram, “Instrumenting the Earth,” which outlines twenty different modes of sensor input, from snow hydrology and avalanche probes to citizen-supplied observations and weather stations.20 Innumerable potential points and processes in the environment become the basis for sensor input, and it is from these delineated sites of input that newly observed relations might be studied, articulated, or managed.

While these sensing projects and networks have been under development within universities and public institutions, technology companies working individually or often in collaboration with universities are also developing multiple sensor network systems for environmental observation. These projects range from Nokia’s “Sensor Planet” to IBM’s “A Smarter Planet,” HP Labs’ “Central Nervous System for the Earth” (CeNSE), and Cisco’s “Planetary Skin” (in collaboration with NASA, the University of Minnesota, Imperial College, and others).21 Governments and their militaries are also investing in the development of sensor networks, with white papers and research issuing from the EU, China, and the United States DARPA, among others.22 Many of these sensing projects raise ethical issues related to surveillance, while still other projects are enabling new forms of resource exploitation. The project of monitoring and managing environmental relationships continues to be a way in which the governmentality—and even environmentality—of sensor systems unfolds, where sensor capacities may point toward particular relations to manage or sustain in distinct ways.23

All together, these environmental sensing systems variously undertake a project of instrumenting or programming the earth.24 Within a sensor-ecology imaginary, the planet might be understood as an entity to be sensed and transformed into data. Improved sensing capabilities are critical to advancing understandings of environmental change while also indicating ways of acting (whether through automated systems or environmental policy) in response to that data. With small-scale, distributed, and pervasive computation embedded in environments, new relationships emerge not just to studying but also to managing environments, since sensor systems computationally describe and capture environmental processes while also providing the promise to “design and control these complex systems.”25

In many ways, the notion here is that increased amounts of environmental data allow for the improved management of environments. Data are descriptive indicators capturing environmental processes. But from a Whitehead-influenced perspective, it could be argued that sense data are less descriptive simply of preexisting conditions and more productive of new environments, entities, and occasions of sense that come to stabilize as environmental conditions of concern. The ways in which phenomena are tuned into as sense data are one part of this operation of the becoming environmental of computational sensors; but the ways in which sensory monitoring gives rise to new formations of sense within and through data, computational networks, humans, more-than-humans, and environments also in-form distinct sensing practices. Since sensor networks offer distinct insights into the complex interactions and processes within environments, then the ways in which these relationships are joined up, articulated, and transformed into new observational capacities also matters.26

Sensing an Experimental Forest

Turning now to a more detailed discussion of one embedded sensor network project, the CENS sensor installations at the James Reserve forest, I consider how the rise of distributed sensing might be looked at more closely in the context of this experimental project and test site. The CENS initiative is one of many sensor developments as discussed previously, and it is a well-known and frequently cited project for sensor research. Established in 2002 as a National Science Foundation Science and Technology Center, the CENS project was a collaboration between several California-based universities. The project, which finished in 2012, focused on four key areas of research: Terrestrial Ecology Observing Systems (TEOS), Contaminant Transport and Management, Aquatic Microbial Observing Systems, and Seismology. A fifth area of research, Participatory Sensing, grew out of the project research into ecology and focused on how sensor applications may be used for citizen engagement in environmental and social issues.27 This discussion focuses on the TEOS sensing deployments, which were primarily situated at the James Reserve (while the other study areas were located in a diverse range of sites). Participatory Sensing is a further project research area that I briefly address in the conclusion to this chapter.

The James Reserve ecological study site is in many ways an environment for developing experimental practices as well as for transporting laboratory techniques into the “wild.” The fieldwork that I conducted at the James Reserve also moved from the laboratory to the field, as I first visited the CENS laboratory at UCLA where most of the sensor prototypes were developed, and then observed the sensors at work in situ at James Reserve. I held informal interviews with researchers involved in the CENS project, mapped the different locations and functions of sensors in the field at James Reserve, and compared the online records of sense data with the sites where sensors were installed. However, this is not a project of “following the scientists,” which is by now a well-established area within science and technology studies.28 Instead, through a discussion of fieldwork conducted at the site, I attempt to understand processes and sites of sensing as they intersect with ecological practice and cultures of computation. Rather than focus exclusively on how ecologists use sensors to obtain scientific meaning or generate data or facts, I concentrate on James Reserve as a particular ecologi-cal research site that concresces through a distribution of sensing processes across organisms, ecological processes, and sensing technologies in the form of computational hardware and software, online interfaces, conservation infrastructures, resident scientists, environmental change, citizen scientists, publics, and visiting researchers. In other words, I attend to the becoming environmental of sensor-based media as a concrescence of these experiencing entities.

The nearly 12-hectare and 1,640-meter-high site is characterized by a complex intersection of ecosystems, “including montane mixed conifer and oak forest, montane chaparral, wet and dry meadows, montane riparian forest, a perennial stream, and an artificial lake.”29 Since James Reserve is located in a relatively remote wilderness setting, it is effectively “off the grid,” and is a study area that generates its own solar power and has its own well for water. In this sensing lab or experimental forest, infrastructures are realigned, not as obvious allocations of roads, electricity, and water, but rather as new arrangements of energy, sensation, and observation.

Sensing in the James Reserve is distributed not only across this experimental site (and at distinct locations for the study of ecological processes) but also across larger sensor networks. Many of the CENS James Reserve sensors measure phenomena over time, which is meant to enable researchers to study sequences of data that are fine-grained and relatively continuous in comparison to more discrete data sets, with data captures taking place in localized settings as frequently as every fifteen minutes or more. Still other sensor test beds are in place that connect up to larger networks, including national observatories such as NEON. Observations are successively gathered and joined up in far-reaching networks, so that sense data becomes an amalgamated and comparative networked infrastructure of ecological observatories for studying environments and environmental change.

CENS sensor systems have been developed and deployed within a larger project that seeks to collect data in order to respond more effectively to environmental challenges. Higher-resolution data promise to create more effective models for predicting and managing environmental events. This “new mechanistic understanding of the environment” involves a near-future commitment to developing a “critical infrastructure resource for society” in the form of detailed environmental monitoring.30 The promise to respond to crises more effectively develops not just through larger data sets but also through more extensive data gathering that is better tuned to detecting anomalies and extreme events, since most ecological data have largely consisted of documenting ecological conditions within a logic of averages and generalities.

However, data expressive of average conditions do not capture the effects that major if singular disruptive events have on environments and rapidly shifting ecological relations and processes.31 CENS and related projects such as NEON are oriented toward the objectives of monitoring changing environmental processes, where an increasing number of disturbance events are contributing to the perceived need to develop different practices and technologies for sensing environments. The expression and agitation of environments (which, as Whitehead suggests, “seep” into all things) also turn up in and transform the sensing practices and technologies that monitor them. Instruments for capturing sense data are here specifically honed toward disturbance, since environmental change becomes more of a matter of concern within ecological study. At the same time, disturbance detection rather than observation of norms begins to influence what counts as relevant sense data.

Machine Ecologies

The sensors at work in the James Reserve within the TEOS group of research projects consisted of everything from soil sensors that detected moisture levels, a Rhizotron installation of tubes that allowed robotic cameras to capture images of root growth and CO₂ sensors at three different soil depths to estimate soil flux, a bird-audio system involving sonic booms triggered by camera activity to capture woodpecker auditory data, weather stations for gauging microclimatic conditions, tree-sap flow sensor systems, nest boxes with cameras and audio installed within bird boxes, pan-tilt-zoom tower cameras on thirty-foot-tall poles, and a Moss Cam web camera. At the time of this fieldwork, there were over 550 connected and untethered sensor nodes, as well as reconfigurable robotic mobile sensors working above and below ground, within waterways and across tree canopies, capturing data on plants, animals, birds, soil, microclimate, and more.32 Sensor observations provided the ability to observe fungal growth patterns, soil CO₂ production, the times at which plants shut down their CO₂ fixing, and all manner of activity that typically takes place outside the scope of direct human observation.33

The initial proposal for this project made a bid to develop “distributed sensor/actuator networks [that] will enable continual spatially-dense observation (and ultimately, manipulation) of biological, environmental, and artificial systems.”34 Midway through the project, many of the initial proposals for comprehensively distributing a large number of small sensors within an area of study shifted to a practice of strategically deploying sensors in precise locations to study specific ecological activities and to develop a hierarchy of sensing platforms that could span from small-scale motes to larger sensors such as imaging robots on cables.35 The sensor practices and arrangements developed in the James Reserve context were specific responses to site conditions and processes, so that phenomena to be observed in-formed which sensors would be used and how. At the same time, the difficulty of creating a pervasive sensor network led to a focus on specific sites of study as a more feasible test of the technology. This points to a key aspect of the sensor systems: they were almost always physically proximate to that which they monitored. Sensors were distributed in the environment, and networks were developed and paired with those environments.36 Sensors in the field at James Reserve were wrapped around tree trunks in a loop of foil and cables; they were interspersed in the ground as arrays at regular intervals; and they were clustered at bird boxes to cross-correlate microclimate in relation to nesting at distinct locations.

The ways in which sensors were paired with environments was not a simple mirroring, however. Sensors proximate to roots and soil, for instance, did not stream all possible data all the time. Instead, sensor motes within a network talked to each other to coordinate data detected, processed, and sent according to distinct algorithms. Part of this configuration had to do with energy efficiency, where motes were triggered to record events only at select times and were turned off during times of inactivity to save energy. Indeed, a key aspect of imagining the possibilities of sensors as environmental systems involved thinking through how it may be possible to realize “pervasive sensing” without “pervasive infrastructure,”37 which primarily meant not requiring a central electrical grid for power. The sensors at James Reserve were in part powered by a solar array that was the primary source of energy to power this elaborate sensing lab, which was supplemented by batteries, including motorcycle batteries, for distinct devices to transmit their sensory data via wireless and networked connections.

Part of the algorithmic processing of sensor data involved setting sensors to pick up, filter, and amalgamate data within established ranges. The processing that sensors undertook was ad hoc and in situ, rather than a continual capturing and streaming of environmental activity. Each mote within a network was already set to detect some things and not others, to make correlations among certain data criteria, and to discard anomalies and redundancies according to predetermined phenomenal ranges. Sensor motes detected events within a specific range, and processed and communicated this data across short distances or hops to other sensors within the network for collection at sensor nodes. Data were typically fused and processed at each individual mote in order to make real-time streaming more efficient and effective.

While sensors were physically proximate to what they sensed, that which was sensed and communicated traveled through channels of algorithmic detection and processing. While sensor applications are intended to record extreme events and anomalies, the algorithms that capture data have a tendency to smooth and fuse data at the source in order to conserve energy and generate manageable quantities of data, which even with these filtering mechanisms can easily run to several million records per year per sensor patch. These syntheses are intended to turn data into “high-level information,” where the multitude of records and raw data transform into something like observations or experience.38 This transformation required “data reduction” in the form of “in-network processing” that aggregated similar data and filtered redundant data.39 As CENS researchers Jeremy Elson and Deborah Estrin write,

In this way, processes of filtering, aggregating, and selecting have already been put in place to turn sense data into relevant information. At the same time, these filters may not always capture intended phenomena. A researcher walking through the James Reserve forest might create noise that is picked up on sonic booms, which through algorithmic parsing activates cameras to record activity. In this field of environmental sensing, researchers might fall within the data event-space of motion detection, but inaudible birds traveling in a different column of air might not be detected.

Processes of producing data are also processes of making sense: the experiment is generative of modes of experience. These processes include how sensors are developed in the lab, tested in the field by technologists and scientists, merged with historic ecological study practices, and read across new data sets, while also producing distinct insights into ecological relationships by connecting up multiple experiencing subjects. The architectures and algorithmic processes for relating sense data are a critical part of how sensor systems operate. They articulate how sense data will come together into arrangements indicative of environmental and planetary processes.

Inevitably, the focus on gathering massive amounts of sense data raises issues related to data ontologies. Sensor networks provide the basis for monitoring and acting upon environments, and yet the data and connections made across sensors are selectively captured and joined up, and are also subject to failure and incompatibility of data.41 Different data standards, classification techniques, and dispersed practices in-form the content and processing of dataspaces.42 Databases and dataspaces are more than collections of objectively observable facts—they are embedded within and performed through infrastructures of science, governance, and public outreach. On the one hand, there are issues related to how an entity becomes data, as Wolff-Michael Roth and G. Michael Bowen have discussed in relation to the digitization of lizards.43 On the other hand, there are questions about what constitutes data (a lizard may seem to be a clear artifact of digitization; but when its habits and habitat become part of the sensed data, where does the organism begin and the environment leave off?). Data ontologies in-form which data are collected, but they also in-form possibilities of sense by giving rise to new actual entities and occasions for articulating and experiencing relevant sense data.

System as Sensor and Proxy Sensing

In order to create a more effective parsing of environmental phenomena, sensors are not just used as individual devices that simply generate discrete sense data at the James Reserve. Instead, multiple sensors and sense criteria within a sensor network are often also brought together to form a composite picture of a distinct environment under study. Chemical analysis of pollution may provide readings on contaminant concentration levels, but additional sensors may also work out the direction and speed of contaminant travel, as well as the size of an affected area, by cross-correlating multiple sensor data. In this process of data fusion, the “system is the sensor.”44 Sensors working together within a network establish a computational pattern of correspondences, where the physical sighting, sensor type, coding, and correlating of data coalesce into an environment of sensor data that in-form observations. When the “system is the sensor” and the network operates as a sort of distributed instrument,45 it might be possible to create models and forecasts of ecological processes and, through these sensor systems, act upon environments.

Sensor systems may also act as proxies for the environments they sense. Sensors as proxies are not standing in for a more-real version of environments, but rather are sensory operations that mobilize environments in distinct ways. Sensor networks perform—and so transform—environmental systems. Data may be correlated across sensor types, or sensors may trigger other sensors to capture phenomena, or trigger actuators to collect samples for later study.46 Inferences can be made about phenomena through sensors and actuators, and sensors can be arranged through flexible, multiscalar platforms that investigate particular sensing relationships.

As a CENS “Distributed Sensing Systems” white paper notes, “Embedded sensing can involve a mix of observations with inherently different characteristics. For instance, it is common for systems to include multiple sensors, each with a different form of sensory perception or modality.”47 This is the case in James Reserve, where seemingly traditional image and audio technologies provide a new way to “sense” phenomena in the absence of direct biological sensors. While the majority of sensors now available are capable of detecting physical and chemical attributes, devices such as cameras become newly deployed as biological sensors in the absence of direct biological sensing capabilities, where physical and chemical sensors algorithmically set to filter for event detection can automatically trigger cameras to record biological events.48 Imager and audio modes of sensing are activated within a computational network that mobilizes these forms of sensing as distinct and often proxy operations. The possibility to articulate relationships and interactions within environments to a higher fidelity is something that is meant to be generated through sensor applications that join up environments across sensor system hardware, software, databases, and cyberinfrastructures, as well as distinct sites and the more-than-human processes.

Proxy modes of sensing do not just extend to sensors triggering other sensors or actuators to perform sensing operations but also include proxies that become apparent vis-à-vis more-than-human processes. A not-uncommon technique within environmental study, where climate change in deep time may be studied through ice cores as proxies for past climate events, proxies within sensor-based environmental monitoring are mobilized to infer and detect traces of ecological processes. In the James Reserve, for instance, phenology is a central area of study. In order to capture seasonal relationships, organisms may be observed for the ways in which they “process” environments.

The perceptive capacities of Violet-Green Swallows and Western Bluebirds, in addition to Star Moss and other organisms, are placed under observation through webcams and Cyclops networked image sensors, which capture images and data related to these organisms often at least every fifteen minutes per day, if not more frequently.49 The bird cams and Moss Cam, or web camera specifically monitoring the growth of Star Moss, generate a store of image data that can be compared to microlocal temperature and related data, as well as data captured throughout the James Reserve site. The birds’ choice of a nesting location, or the failure to raise chicks due to absence of food or low temperatures, can be captured in this context where the birds’ activities are made available as a sort of proxy sensor of phenological processes. Birds may provide key environmental sense data through computational networks that make sensible these registers of more-than-human experience. What is clear is that sensors do not just capture data, they shift the processes of sense across these multiple registers, so that more-than-human perceptive processes concresce in newly relevant arrangements.

Similarly, the Moss Cam generates images and daily records that contribute to a picture of seasonal patterns and “event effects.” These effects might include lack of moisture in the summer, which contributes to mosses “burning through” their CO₂ reserves—in other words, higher temperatures can correlate to an increased release of CO₂ by mosses, as they consume stored energy and move toward states of dehydration and dormancy. Here, what counts as “sensing” is not a simple matter of observing mosses through a web camera over time, but instead involves observing how the moss is a sensor, or a biomonitor that is itself detecting and responding to changes in the environment.50 The mosses’ morphological changes to local conditions are an expression of an ecological relationship that is further entangled in the complex shifts of climate change. In this respect, the mosses may be expressing sensory responses to human-altered worlds, yet to understand more fully what those alterations involve, it is necessary to observe sensing organisms in order to register the effects of increasing carbon and temperatures. The delay and resonance within these environments is not as immediate as a typical sensory example might assume. Yet in this study, the ways in which sensing organisms “take account” of environments multiply, where the sensory input and means of detection are distributed and computational. The becoming environmental of computational media then further takes place through organisms and their processing of environments.

In a sensor-based study of phenology, sense operations are distributed and collaborative. In these forms of collaborative sense, sensors experience and provide proxy experiences across a sensing system that generates distinct occasions of sense. But these collaborative qualities of sense concresce not through researchers primarily but through the dynamic responses of organisms to environments and the sensors that collect data in relation to which algorithms query, filter, and record these changes. The more dynamic sensory modalities that concresce in this relationship are examples of inventive ecological experiences and subject-superjects, as discussed earlier. The timings at which plants leaf out, for instance, might even begin to disrupt and alter scientific models that expect seasonal timings to unfold at times established through prior empirical study. In these encounters and formations of sensory practice across organisms, ontologically prior categories of sense become more mutable and ontogenetic, where more-than-human modalities of sense indicate the shifting encounters of sense in which we are engaged. Sensor systems mobilize multilocated and multispecies processes of sensing, which in part enable the development of distinct capacities to sense change, where the scope of computational sensing and proxy sensing expands to include more-than-technological perceptual processes.

In an account of ubiquitous computing as distributed cognition, Hayles suggests that distributed computation could operate as machines for aiding, and so enhancing, human perception.51 Here, however, computational devices are not augmenting human perception as such, and humans are not even the central perceptual processors toward which distributed sensation and computation might be directed. More-than-human proxy sensing points to the ways in which sensor technologies, instead of providing supersensing or cognizing capabilities to supplement human modalities, filter, connect up, and in-form environmental relations in distinct ways, and so change what modes of sense humans may even experience. New ecological arrangements of subjects—and superjects—concresce through these sensory processes.

Environmental monitoring through sensor networks is a practice of making—and not just capturing—environments as process. Sensor networks are tuned to distributions of relations. They tune into discrete sense criteria and amalgamate these across sensor networks and through proxy modes of sensing to make particular environmental relations more evident and sensible. Environmental monitoring through sensory networks mobilizes and concretizes environments in distinct ways by localizing computational processes of sensing within environments and across more-than-human experiences while also articulating those relations through algorithmic processes for parsing data. As these processes inevitably compose the possibility of sensing environments in particular ways, they also in-form which participants and participatory modes of sensing register in the perceptive processes of sensor technologies. Such sensing practices, moreover, are replete with political effects. Within the context of sensor networks, the sensory arrangements that are identified within data may become the basis for identifying and protecting matters of concern. Yet they might also overlook those “non-sensuous” background events that may still generate new sensing arrangements but which are not interpretable within present modes of sense data.52

Distributing Sense

As discussed in the introduction to this study, the initial developments of ubiquitous computing are often attributed to Mark Weiser’s 1991 suggestion for computation to move from desktops to the environment, so that computational processes would become a more integrated and invisible part of everyday life.53 Yet another possible reference point could be Alan Turing’s 1948 ruminations on how to build “intelligent machinery” with sensing capacities on par with humans. Turing reviews the options for such a project, first considering how to atomize every part of the human sensing ensemble and replace it with equivalent machinery. Emulating human vision, speech, hearing, and mobility, such a contraption “would include television cameras, microphones, loudspeakers, wheels and ‘handling servo-mechanisms’ as well as some sort of ‘electronic brain.’”54 This project would inevitably be “of immense size,” Turing notes, “even if the ‘brain’ part were stationary and controlled the body from a distance.” But data would not enter the thinking machine through its remaining static, and so “in order that the machine should have a chance of finding things out for itself it should be allowed to roam the countryside.” But in such a scenario “the danger to the ordinary citizen would be serious.” Add to this the hazards of such a machine taking up all of the usual activities of human interest, and this contraption would be altogether unwieldy. Turing’s more practical recommendation is to behead the body, to work with the brain as the critical site of processing, and later attend to the sensory apparatus as a secondary concern.55

Even if Turing’s proposal does consolidate the “thinking machine” into a central and seemingly Cartesian apparatus, his thought experiment on the sensing body in pieces and distributed throughout the countryside remains a potent figure for ubiquitous computing. What is striking about Turing’s example is the way in which the thinking machine, even when distributed, would emulate the human body, which serves as a template for understanding how sensory data would be captured and centrally computed. While computational sensing technology can now be understood as more than a double of or prosthesis for human sensing, Turing’s figure of the body in pieces raises questions about how particular distributions of sense might reconfigure environments and processes of sensation.

Could such distributions of sense point toward modes of sensation where computation reassembles not as a singular sensing subject but rather as a processual and multilocated experience comprised of numerous sensing entities? How are sensing practices individuated, and how do they concresce, across potential sensor networks? In this way, sensing also assembles not as a mental or cognitive operation but as an environmental and relational articulation across multiple bodies and sites of sensing.56 Within Turing’s example of the sensing body in pieces, this could mean that we attend not to how the body might reassemble toward human perception and functionality but rather to how the “countryside” and the many inhabitants, processes, and processors of this distributed and distributive milieu begin to rework how the thinking-sensing machine captures, configures, and acts upon its inputs.

Perception in the World

Turing’s sensing apparatus points to the distributed processes that make sensing possible, even if the sites of sensation do not return to a coherent human processor. Indeed, as Whitehead suggests, perception might be understood to be in the world and distributed through more-than-human processes—it is not the special preserve of a human decoding subject. Instead, multiple participants express and unfold a distinct experience of the world, independently but contemporaneously within an immanent series of events.57 At the same time, the excitations of environments are fused to all modes of “matter,” where “the environment with its peculiarities seeps into the group-agitations which we term matter, and the group-agitations extend their character to the environment.”58 “There are numberless living things,” Whitehead writes, that “show every sign of taking account of their environment.”59 This taking account of environments is a way of capturing what is relevant, and—through being affected—of transforming environments and relations.

Sense data might be seen as a concrescence of multiple ways of taking account of environments, whether through researchers or devices or environmental events. But these data are necessarily articulations of the ways in which environments are gathered and expressed through varying subjects—here, with subjects understood in the broadest possible way. Sensing systems generate and concresce distinct articulations of environmental relations within and through data and across sensing “subjects/superjects.” Rather than take on a Kantian view of how “the world emerges from the subject,” Whitehead, with his “philosophy of organism,” seeks to understand how “the subject emerges from the world,” thereby constituting a “superject,” or a subject that is always contingent upon actual occasions and experience.60 As Shaviro notes in relation to Whitehead:

Sensor technologies are constitutive of sense—they too “experience” the world and generate perceptive capacities.62 Sensors that map in real time a greater density of ecological relations might generate a processual approach to environments by focusing on interactions and even multiple modes of perception. At the same time, to identify a phenomenon as constituting sense data is to make a commitment to distinct “forms of process,” so that environmental processes are selected and concretized in those forms. The process of selecting sense data involves capturing a moment in time, an “instant,” that is re-sutured with other data to form a pattern of ecological processes. While approximating a more process-based and even real-time monitoring of environments, sensors are also productive of practices of selecting and interrelating discrete observations in order to arrive at an understanding of ecological processes. The selection of temperature, vibration, light levels, humidity, and other measurements across primarily physical (although to some extent chemical and biological) criteria in-forms the instants that are sensed, the forms that are documented, and the processes that might be reconfigured.

The basis for developing “facts” within the sensing experiment then directly pertains to the forms and processes of experience that are generated and connected up across sensing subjects.63 The concrescence of data also requires subjects that can prehend and experience the data. Subjects may be attuned or resistant to receiving data based on prior or concrescent experiences. But the means of gathering data might also contribute to the possibilities for processing and integrating data. In this way, sense data as experienced by subjects may be generative of superjects where the experiences and perceptions generated are in turn formative of the subjects that experience. This runs counter to the notion that a founding subject is the entity that experiences. If, as Whitehead suggests, subjects are always superjects, then subjects are always necessarily distributed and concrescent in relation to actual occasions.64 Subjects, whether stones or sensors or humans, become environmental in this way since they are involved in feeling and concrescing actual worlds.

Approaches to media and sensation often focus on the ways in which technologies train or otherwise attune the human senses within a mediatory or prosthetic relation. But the interactions and processes of sense are arguably not fixed within sensory organs or technologies through which mediations are typically understood to occur. In this way, sensation is not primarily an inquiry into relations between human subjects as they perceive more-than-human objects. Instead, the sensory relations within which sensors are mobilized give rise to a more ontogenetic understanding of perception, where sense and expressions of perception are articulated processually and across multiple sites and subjects of inventive sensation.65 In this way, new perceptual engagements are distributed across sensing capacities and engagements (perhaps similar to what Luciana Parisi has called “technoecologies of sensation”), which give rise to distinct sensory processes, informational-material arrangements, and ethico-aesthetic possibilities.66

Such a condition resonates with what Patricia Clough refers to as the importance of focusing on “an empiricism of sensation” rather than “an empiricism of the senses.”67 Technologies, including sensor systems, can be understood as generative ontologies that in-form the experience and conditions that make sensation possible and changeable. Rather than studying “the senses” as given, it may be more relevant to study experience and how distinct types of sensation become possible, and to consider further what modes of participation and relation these processes of sensation facilitate or limit. To bring this analysis back to sensor technologies, sense data are not simply items to be read and gathered as machinic observations of environments that scientists process. Instead, sense data are indications of a process of becoming sensible, where environments, humans, and more-than-humans are individuated as perceiving and perceivable entities.

Collaborative Sensing

The modes of sensing that concresce within the context of ecological sensor applications might, as discussed earlier in this chapter, begin to be described as collaborative sensing practices taking place across multiple subjects and through distinct processes of experience. These modes of sensing could further be referred to as types of “intimate sensing,” as Stefan Helmreich has suggested in relation to fieldwork undertaken with oceanographers who employ a complex array of sensing technologies in their research. Sensing, in this account, is comprised of a research “ecosystem,” and involves much more than a device focused on an object of study, since bodies enter into a circuit of sensation with instrumentation technologies. As Helmreich writes, “These scientists see themselves as involved not so much in remote sensing as in intimate sensing.” Multiple forms of sensing are articulated across different technologies—and so with researchers involved in studying ocean ecologies: “The mediations are multiple and so are the selves.”68

Influenced by Charles Goodwin’s discussion of how forms of “collaborative seeing” are produced within the space of a scientific vessel,69 Helmreich develops an analysis of the sensing processes that become concretized within these body-environment-technology relationships, where new registers of feeling might sediment through repeated engagement with these devices. The multiple selves that Helmreich discusses most frequently refer back to scientists and crew members on ocean-sensing expeditions, but by extending this approach through a Whitehead-oriented understanding of experience it is possible to include even more expanded collaborative formations of sense. The experiences provided by and through more-than-human processes, as well as the processes that unfold within sense data, in-form a more environmental approach to what might constitute “collaborative” modes of sensing.

Within the area of more-than-human theory, sensation is increasingly understood as distributed in and through more-than-humans in the form of organisms and technologies, together with their environments. At times influenced by Foucault’s well-known “death-of-man” statement, media scholars as far-ranging as Friedrich Kittler and Katherine Hayles have in different ways undertaken analyses of media that dispense with an assumed human subject as the principal site of meaning-making in order to recast the relations that concresce in and through media technologies.70 As Hayles suggests, environmental modes of computation—RFID in her analysis—raise questions about the effects of “creating an animate environment with agential and communicative powers.” Such technologies allow us to move toward “a more processual, relational and accurate view of embodied human action in complex environments.”71 Not just sensing but also what counts as the “human” shifts in these scenarios, since computational technologies typically now operate within parallel processes and signal toward a multiplication rather than a centering of subjects.

The subjects that might be discussed as parallel, multiple, or collaborative within environmental sensing extend not just to entities multiplied through more-than-human technologies but also to the incorporation of more-than-human flora and fauna. More-than-human theories of subjects—or ecological approaches to subjects—are becoming increasingly well established not just in media theory but also in philosophy and feminist studies, particularly as articulated in the work of Braidotti, who develops these notions through the work of Deleuze and Guattari (with an emphasis on the notions of ecology developed by Guattari). Braidotti suggests that we begin to work with an “environmentally bound subject” that is also “a collective entity” because “an embodied entity feeds upon, incorporates and transform its (natural, social, human, or technological) environment constantly.”72 In this account, bodies and subjects are even understood as collective information machines of sorts. For Braidotti, “techno-bodies” may be understood as “sensors,” or “integrated sites of information networks; vectors of multiple information systems.”73

Such an ecological approach to subjects resonates with Whitehead’s discussion of subjects/superjects, where bodies-as-sensors are expressive and productive of environments. The sensing that takes places is a practice of processing and transforming. If human bodies are sensors, then by extension so too are the multiple more-than-humans that take in, express, and transform environments. As the preceding discussion of the James Reserve suggests, it is relevant to bring these multiple formations of experience to play across human and more-than-human subjects into an examination of the specific distribution of environmental sensor networks in this ecological study site and to consider how sensors are expressive of environments, what new environments and subjects concresce as experiencing entities, and how the sensing experiment might make these experiences possible.

Inventing Experience

From an experimental forest, this analysis of environmental sensing turns back to Turing’s countryside—that apparently static backdrop through which sensing was to take place. While Turing imagines a distributed sensing entity processing its bucolic surroundings, in this analysis of test sensors installed in a forest setting it becomes clear that the surroundings to be sensed are in flux and yet formative to establishing conditions and practices of sense. Through this reading, Turing’s distributed computer becomes a superject, integrated with and formative of the environments and experiences it would decode. It becomes environmental in that it is an entity that generates the formation for further subject-superject experiences. This approach, as discussed throughout this chapter, provides a way of taking account of the abstractions and entities that lure feeling and settle into forms of environmental engagement.

The environment or milieu as differently understood by writers from Whitehead to von Uexküll, Canguilhem, and Foucault, has been discussed as everything from the conditions of possibility to a zone of transformation and necessary extension within and through which experience is possible.74 Within the work of von Uexküll, the now well-cited example of the tick that is provoked to act in relation to certain environmental cues is referenced to signal the ways in which sensation is tied to environments and to suggest the species-specific coupling between these.75 Sensing beyond the human subject can be figured through more-than-human agencies that unfold within environments. But if we take the provocations of Whitehead seriously, then the milieu is not just a site where sensing joins up. Instead, it is also a transformative and immanent process where modes, capacities, and distributions of sense concresce through the experiences of multiple subjects.

Any given milieu or subject/superject is expressive not of scripted coupling as the work of von Uexküll might suggest, but of creativity, as demonstrated in the work of Whitehead and Simondon.76 If inventiveness is a necessary part of perceptive processes, then the environment-as-agitation necessitates a more ontogenetic, collaborative, and extensive understanding of sensing. In this way, perception might also move beyond the notion of hybridities or even mediations of sense and instead focus on the sensing conditions and entities that concresce, as well as that which environmental perceptive processes make possible, and how inventive processes might further generate new forms of collective potential.

The complex interactions that are the focus of study for environmental sen-sor systems are transformed through the perceptive processes that these systems generate. The ecological relations that are to be discovered and studied are bound up with the detection of patterns within sense data. Sensor hardware and software do not simply gather sense data in the world, but are part of the process of perceptual possibility, both as more-than-human registers of perception and through making distinct relations sensible as subjects of ecological concern.

The possibility to relate and to make aspects of relations evident is an important aspect of sensor systems, with political and practical consequences. Sensation might be understood as distributed and automated on one level, yet on another level such automation in relation to environmental processes involves not only running scripted functions but also addressing the open and indeterminate aspects of sensors in relation to environmental processes. This is one way of saying that, whatever the computational program, sensors never operate strictly within a “coded” space, but by virtue of drawing together expanded perceptive processes they inevitably make way for a generative technics of environments.

There are political implications to the implementing of sensor processes: relations are not simply discovered in the world, rather they are individuated through these distinct computational sensing processes. These processes further orient environmental practices and politics, where increased data and improved awareness of ecological relationships are expected to translate into an improved ability to manage environments and potentially prevent the spread of environmental damage. These crucial relationships concresce not just through practices of data collection and monitoring, as well as sharing data within larger networks, but also through drawing inferences across data sets that illuminate key ecological relationships that are to become the basis of concern or protection. As Whitehead suggests, that which counts as a form or datum is what endures within a “process of composition,” which is expressive of “historic character.”77 What counts as empirical requires acts of “interpretation” but also describes a concrescence that continues to have the force of natural fact. Drawing on Locke, Whitehead notes, “The problem of perception and the problem of power are one and the same, at least so far as perception is reduced to mere prehension of actual entities.”78

While Whitehead’s analysis works across philosophic and cosmological registers, and does not directly address sociopolitical analysis of environments, his work does point toward potential translations to be made across experiencing subjects to political possibilities. As Shaviro suggests, following on Whitehead, experience is a site of potential: “It is only after the subject has constructed or synthesized itself out of its feelings, out of its encounters with the world, that it can go on to understand that world—or to change it.”79 In other words, as Whitehead notes, “How the past perishes is how the future becomes.”80 That which is sustained and that which concresces as a register of novelty are processes whereby experience may give rise to new experiences, interpretative practices, and matters of concern. In a different way, Foucault indicates through his discussions on the milieu that sensory arrangements articulate distributions of power, and involve making ongoing commitments to relations and ways of life.81 Sensory processes that occur across subjects are expressive of material–political relations and possibilities for participation.

Environmental monitoring through sensor networks is a technoscientific practice that pertains not just to the study of ecological relations but also to newer modes of participatory sensing and citizen-science activity that rely on the use of the sensing capacities on mobile phones and low-cost sensors to track and gather data from environments. While citizen-sensing applications have developed to move these scientific applications into the hands of the general public,82 even more questions arise as to how or whether sense data makes an effective traversal from data to action. The implications for sensory practices that are articulated within an environmental monitoring context then have relevance for thinking through the processual, relational, and heterogeneous aspects of sensing. Given that the CENS research has moved “out of the woods” to citizen-sensing applications, while at the same time a whole host of participatory applications such as forest monitoring platforms are materializing to protect forests for conservation, how do forests, “citizens,” more-than-humans, and sensor technologies converge to invent new forms of politics that are attentive to present matters of concern and those that are yet to come? In the next chapter, I consider this question in relation to a seemingly more prosaic “sensor” and the sensing practices it operationalizes: the webcam.