The limits and value of big data: towards a political ecology approach

It's been quite a while since I've posted here. I've been working a bunch with UW-Madison's Center for Culture, History, and Environment on our new site, Edge Effects, including a post on the use of site selection models in conservation, and now as an editor. Check it out!

In fact, I've been thinking more and more about the use of tools like site selection models in the practice of conservation: what they do, the political economy behind their production, and how people use them.

And so I've been looking more and more into how "big data" is coming to bear on the environment. I was scheduled to give the paper below at the recent Dimensions of Political Ecology conference at the University of Kentucky, but my session was already jam-packed! So I'm posting it here instead. I make several arguments in the paper: 1) though much of what we hear about big data comes from the realms of healthcare, academic research, and corporate finance, new kinds of data analytics are indeed coming to bear on familiar territories for political ecologists: conservation, agriculture, resource extraction, and the body; 2) political ecologists are well-equipped to tackle big data; 3) political ecology may in fact offer a unique perspective to critical data studies in general, beyond the realm of conservation and development. These arguments are somewhat superficial here, nor is the paper fully referenced. It is a first draft and I appreciate any feedback you may have:

----

My research focuses on the tools of environmental governance. I’ve looked at ecosystem assessment methodologies that produce viable markets in nature; my dissertation is looking at the political economy of new visualization tools and models, and data collection techniques being brought to bear on coastal land loss and marsh restoration in Louisiana. I’m asking: how are these tools produced and utilized and with what effects in the service of adaptive management and the valuation of ecosystem services. So when I saw this John Deere promotional video, I couldn’t help but be intrigued. [It’s a little slow-moving – you might fast forward to the 3 minute mark].

Promoters of these kinds of “data-driven solutions” to agriculture, business, health, government, and conservation suggest that a frontier of existing and constantly coming online information – big data - and the new tools and techniques associated with compiling and making sense of it - can fundamentally change how individuals and institutions go about decision-making. Political ecology (PE) has always concerned itself with the question of how land managers make decisions, with a critical eye towards contextualizing these decisions within broader political and economic processes and forces. So what if anything does PE have to say about this particular situation?

PE, I argue, has the conceptual tools and interests to engage with these kinds of situations. PEists should be engaging with these kinds of situations – in fact, given its long-standing emphasis on questioning the scalar and political premises of knowledge’s production, circulation, and application it is uniquely positioned to, in ways that would add to the conversation around new forms and objects of data analysis as a whole. Not only should it, but in many ways it must if it want to continue being its critical role as a “hatchet” for what science, state, and capital expect from those working the land.

In this paper, I spell out what big data is, to illustrate briefly three different ways data is being taken up in environmental governance, and ultimately, what PEists can do with big data – what we can ask, what we can show, and how we might practice it ourselves.

"The Body as a Source of Big Data." http://ihealthtran.com/wordpress/2013/03/infographic-friday-the-body-as-a-source-of-big-data/

What is big data?

Even just spelling out what big data is an improvement upon most accounts, which give it an unearned mythical status:

Simply put, because of big data, managers can measure, and hence know, radically more about their businesses, and directly translate that knowledge into improved decision making and performance. Harvard Business Review

It’s crucial yet difficult to pick apart what we mean by big data. It has a certain parallax – that is, it’s an object that looks different when we look at it from different advantages. At one level we might say big data necessarily comprises an object, a method or practice, and a tool – on another level it’s: data, new tools for collecting it, and new tools for analyzing and making sense of it. Let’s look at each in turn.

Object

Colloquially, there’s a sense that big data is simply datasets that Excel can’t deal with. More precisely, this means data where there are more columns than rows, or, as many or more attributes of events and things than those events/things themselves (National Academy of Sciences, 2013). But big data does not just mean big much less numerical data (spreadsheet-ready or not). A more formal definition might include not just voluminous, but: disparate (that is, coming from various sources – different sensors, let’s say, but also different texts), heterogeneous (of different data types – text (e.g. police records) vs. spatial), and uncertain (of varying quality, accuracy, and so on). For some observers and promoters, this translates to: “volume, velocity, and variety.”

"Big Data? Volume, Velocity, Variety." Some also add a fourth v: value (see below!) http://www.wired.com/2013/06/is-big-data-in-the-trough-of-disillusionment/

Method
Whatever the data itself looks like, it won’t speak for itself. There’s a practice to big data, and there are two important moments to consider here.

Data collection
First, big data is marked by the use of new technologies that can record more and more observations, more and more rapidly, and perhaps above all, more and more remotely  - that is, at some (physical, social, computational) distance from where, who, and what is in fact sifting through the data. Some of the foremost big data stories we hear about fall into one – or often more - of three kinds of a remote governance: 1) Enhanced remote sensing (RS).  Political ecologists have regularly both employed and critiqued RS (Turner 2003), but: 1) this is in many ways RS on steroids – one company, Skybox, is interested not simply in capturing land cover, but extremely fine details of land use (such as the number of cars parked at a mall on a given day); 2) one of the interesting new developments we’re seeing is the integration of remotely sensed data and user-provided data. For instance, satellites are capturing spectral data in west Africa, which algorithms then parse into general land cover categories. Managers of a program out of Cornell then provide incentives to local pastoralists to actually go out and confirm, or “ground truth” the cover type. As my friend Patrick Bigger put it, “It's like Uber surge pricing for environmental surveillance.” 

This leads us into the second kind of remote sensing: 2) self-reported data. Data collected unconsciously (or sometimes consciously) – with permission or without it – from location-based apps or perhaps from website activity (your “clickstream”). This kind of data collection ultimately raises important fears about privacy. But those in the “quantified self” movement embrace collecting as much data as they can about themselves, in the hopes of optimizing say their health or even their investment strategies; 3) Finally, an emerging connectivity of devices – from smart phones to servers to satellites - gathering, transmitting, and analyzing data at a distance has some tech leaders envisioning a so-called “internet of things,” an objectification of everything – literally everything, from trees to buildings (see video below) – into governable sources of data. California has set up water level sensors to give them real-time feedback on lake and reservoir levels, and the USFS is experimenting with real-time monitoring of all sorts of different forest measures.

Data analysis:
The promise of new sources of information means little, however, without the conceptual and practical apparatuses that allow the data to be understood (as some might put it, for the data to become information or knowledge). I’ll name a couple of the key maneuvers “data scientists” make here. I focus on the practice of analysis, but will note the here: analysis is becoming a commodity (what Wilson (2014) calls “Analytics TM”), bringing with it the promise that the world is “infinitely” analyzable. But making the earth suitable for analysis requires first indexing it. For one company, the goal is: “To index the earth the way Google indexes the Internet,” that is to, from each particular observation of the earth’s surface, discretize certain phenomena and values (e.g. extent of deforestation or, as above, number of cars in a parking lot) and then associate these objects with the observation, allowing for easy aggregation, statistical manipulation, and retrieval. Next, for many, the volume of data suggests that traditional tenets of statistical representation can be set aside. When n=N – when the sample of data is in fact the population, when we are able to collect and index all crime data for a city - we do not need to try to extrapolate, which introduces uncertainty in prediction or in analyzing new information (Mayer-Schönberger and Cukier 2013). Instead, simple yet surprising and powerful correlations and trends –for instance that Wal-Mart’s sales of strawberry pop-tarts skyrocket before hurricanes - are enough, enough for Wal-Mart to keep the shelves stocked, but probably not ask why strawberry pop-tarts go so quickly. This has led some – Chris Anderson, famously - to declare “the end of the theory.” Data need no explanation – the trends speak for themselves, allowing for prediction of future events (such as what you’ll order next on Amazon). Data scientists’ final move has been to develop more and better algorithms by which these correlations and predictions can be made. This involves “machine learning” – or the recursive tweaking of equations to “fit,” or properly explain, the data so that predictions can be made. There are generally two kinds of algorithms at play here, which will be familiar to those working with RS: supervised and unsupervised (It’s also worth noting that even though there are much more advanced algorithms at play (e.g. neural networks) these are the basics, and they have their roots in many of the tools that social scientists might use: PCA, linear regression, and so on.) I’m going to skip over an explanation of the differences between these, but you might get the impression that “unsupervised algorithms” are artificial intelligence come to life. While many big data proponents suggest that the “data speak for themselves” – and that’s what the phrase “machine learning” suggests, that machines alone can discover and interpret patterns - data managers will be the first to note that even unsupervised algorithms require active intervention and subjective choice: clustering – an important unsupervised algorithm involves an initial choice on the part of the analyst as to how many clusters to look for; clusters must also be meaningfully interpreted.

Ayasdi's analysis of key basketball "positions" as illuminated by the visual topology of the data. http://www.ayasdi.com/wp-content/uploads/_downloads/Redefining_Basketball_Through_Topological_Data_Analysis.pdf

Tool
Regardless of whatever fanciful things algorithms can do to “reveal” the pattern of a dataset and make predictions, these results need some translation out of R or whatever analysis software package they’ve been developed in. While proponents suggest data speak for themselves, it might be more accurate to claim that data visualize themselves. That data patterns and results can and must be seen is a tenant of data science, meaning that whatever goes on in the analysis should be translated into some visual medium, all the better to communicate with decision-makers around. And so we see alongside new analytics, new tools: for instance, “dashboards” and “decision support tools” that collate data and results, providing decision-makers “levers” to pull to move forward.

Palantir's tool for "adaptively managing" salinity levels in the San Joaquin Delta. https://www.palantir.com/2012/09/adaptive-management-and-the-analysis-of-californias-water-resources/

What does big data have to do with ecology?

Increasingly, conservationists, other land managers, and powerful actors in environmental governance (financiers, policy-makers, etc.) are adopting these tools. Conservation has a long history of the use of technology for management, all the way from mandating best available pollution control technologies to RS and GIS to big data today. For many conservationists, the claim is explicitly that new technologies “can help save the planet.” This is an important claim to investigate on several accounts: 1) the word help is important, but is always vaguely defined. To what extent is this a proposal that ecological degradation simply needs a technical fix (as opposed to addressing the root social causes)? 2) the claim embeds a particular scalar argument (save the planet, not just certain ecosystems). Political ecology has always been concerned with both of these kinds of questions. I offer three very short vignettes simply to illustrate what happens when big data and conservation meet, and to point to areas of research where political ecologists should be working.

"Finally, Data You Can Actually Use." Climate Corporation. http://www.climatepro2015.com/

Smart farming
Smart farming is everything in the John Deere promo linked to above. Smart farming advocates aim for the integration of field, equipment, and meterological data to support precision planting and optimizing crop yields. The idea is that this data – made accessible through interfaces and dashboard tools – can help farmers limit unnecessary applications of fertilizers (through near real-time, sub-meter access to satellites able to pick up near-infrared light, allowing insight into nutrient deficiencies and growth rates), and plant the right kinds of crops in the right place given current climatological and soil conditions. A number of different firms have set themselves up in this market, including Skybox, which bills their recently launched private satellites as providing “Big data. From space,” and Climate Corporation, which collects public weather data, translating it onto a private platform to sell to farmers (and was recently acquired by Monsanto).

Valuing nature
While smart farming aims to increase the value of agricultural production by conserving inputs, conservationist organizations themselves are aiming to use big data to value ecosystems in their own right. They have set up a couple of cases using social media data to reveal the otherwise unknown or, by default, zero economic value of nature. In California, The Nature Conservancy uses user-generated data from the eBird app to identify where migratory birds are during particularly important intervals and then pays rice farmers in key gaps along the birds’ routes to keep their fields flooded for longer, preserving habitat that can be hard to come by in the context of the state’s long drought. These mini-wetlands are called “pop-up habitats,” and TNC is also talking about it in terms much like smart farming, calling it “precision conservation.” Elsewhere, researchers from the Natural Capital Project – a coalition of academic ecologists and conservation non-profits – have used Flickr data to estimate the recreational values provided by different ecosystems, as a way of making clear to policy-makers the job and revenue creating aspects of nature. Instead of providing surveys to tourists to ask them how much they visited a national park, the researchers used Flickr data as a proxy for visitation, and to guess at how environmental degradation at these parks might change recreational attendance.

Hedging bets

Finally, some data analytic firms are pushing the boundaries of new data forms in their use for valuing nature. These firms write algorithms that sift through satellite imagery and other economic data to make sense of and correlate environmental and economic change, then package this information and sell it to hedge funds in need of investment advice. For instance, Skybox says that it can monitor say gold production from mines across the globe, giving investors in real time literally a bird’s eye view of this facet of the economy. Without having to wait for Newmont Mining’s quarterly statement, investors can with the help of Skybox’s algorithms, more or less spy on the mining giant’s operations to get a feel for whether output is on the rise. The same goes for logging in tropical rain forests, oil and gas wells, and, again, crop condition. In the case of forest monitoring, Skybox suggests that not only can this data be useful for investors, but for activists as well. They’d like to sell their analytics to both sides.

Questions to ask and possible perspectives

1. Material effects
These three vignettes, brief as they are, raise a number of questions and suggest several general lines of inquiry. First and foremost we need to better understand what sort of material impacts are detectable and attributable to big data in practice. What difference does big data analysis make in the transformation of certain landscapes, in comparison to earlier modes of governance? Might it increase the speed at which change occurs (a la pop-up habitats) or improve conservation outcomes by illuminating unseen trends? What is the environment effect of, for instance, Palantir’s tool for managing Delta water salinity? Are water levels actually optimized? What and who do the landscapes reflect? What in ecological terms is a “pop-up habitat”? Moreover, we could also better understand and communicate the environmental impact of server farms: the greenhouse gas emissions stemming from work and life now situated in “cloud” space, the SO2 emissions from diesel generators, and the water usage to cool down servers.

"Data is the new oil" http://dismagazine.com/issues/73298/sara-m-watson-metaphors-of-big-data/

2. Data as waste and value
Political ecologists focused on the intersection of cultural and environmental politics have fruitfully been investigating questions about waste and value, noting that capitalism relies on discursive translating bodies and landscapes as alternatively waste and as valuable (Moore 2013; Goldstein 2013; Gidwani and Reddy 2011). What is previously deemed waste(land) can be enclosed and incorporated as value, and valuable labor can be laid to waste (only to be speculated upon as someday bearing value again). In many ways this is exactly how data managers talk about big data - as a kind of hoard, a resource to be mined, something that in a kind of parallax view, splits the difference between waste and value [I'm grateful to Mohammed Rafi Arefin for making this point clear to me]. They believe that there is value in the data waiting to be realized, just as oil is waiting in the ground, ready to be extracted, refined, and transported to realize its value. This may help us understand data privacy issues in a new way - the unsolicited collection and analysis of data may be a sort of enclosure of otherwise waste information. Big data also more directly promises to be (environmental) economists’ holy grail: long have they sought ways to understand how people value things whose value is not revealed by markets. With big data capturing all sorts of “traces” we leave behind as we click around the Internet, the idea is this data can serve as proxies revealing preferences. When there exists a dataset like Flickr, the story goes, we have a much better sense of just how much people value wildlife. 

No matter what, we should not accept some managers’ techno-optimism without reservation. Like hoarders, analysts fear being overwhelmed by their data. The actual work of having to sort through the data mess just as easily gives the analyst a sense of dread as it inspires hopeful visions of data-driven decision-making.  Acknowledging data mining’s limitations, managers vividly describe the problem of having to sort through too much stuff to get to the valuable parts as one of powerlessness. “Without a clear framework for big data governance and use,” one consultant writes, “businesses run the risk of becoming paralyzed under an unorganized jumble of data.” For one manager reflecting on a massive dataset, “the thought of wading through all THAT to find it stops you dead in your tracks.” Another similarly evoked the feeling of being stranded at sea: “There’s a lot of water in the ocean, too, but you can’t drink it.” Continuing the drowning theme, others simply acknowledge that in big data analysis there is a risk that the analyst will be overwhelmed by inconsequential results: "This set of patterns [in a big dataset] often contains lots of uninteresting patterns that risk overwhelming the data miner.” Indeed, data miners are anxious about not finding the true or most interesting pattern, and instead finding only what they want: “Data is so unstructured and there’s so much out there, you are going to find any pattern you want … whether it’s true or fake.” They fear finding themselves in the data. This is analogous to the “ecological anxiety disorder” (Robbins and Moore 2013) ecologists find themselves in where they either see their concepts as too normative or that humans are a negative influence in ecosystems. We should leverage this concern. As political ecologists might ask what climate change adaptation programs in the developing world expect of their target audiences, we can ask, what do new digital technologies for conservation, farming, and environmental governance expect of their users?

Drowning in big data. Intel. http://www.intel.com/content/www/us/en/big-data/big-data-101-animation.html

I think in many cases we’ll find that these tools expect their users to be rational actors who predictably respond when provided a particular set of informational inputs. That is, the conceit of these tools is to ignore the political economic contexts which shape the extent to which users will even be able to acknowledge, decipher and act upon them (as well as the political economic contexts of their making). There are good reasons, political ecologists know, that a US farmer – especially a small-scale one, but we can also think about the average age of farmers, the fact that most work off-farm jobs - will be unable to perform the sort of decision-making John Deere presents. Yet it is not just these farmer-centric limits to analysis that will be key to recognize – we can think also about the political economic structure of data collection and analysis today: its centralization (like GMO seeds, farmers may or may not own their data) and monopolization (Monsanto recently bought out Climate Corporation), and the limits to “data-driven” action these introduce. Data and new tech are not black boxes for farmers just because they are technically overwhelming and once removed from farmers, but because they are socially boxed off as well.

3. Visualizing like a state

More specifically, with the increasing emphasis on data-driven solutions to everything, we should expect to see “analysis paralysis”, or situations in which too many options and too much information are presented to decision-makers for them to be able to take any coherent action. This may be particularly true for policy-makers, the kind targeted by tools like Palantir’s salinity monitor:

…our analysis tools “can give policy makers maximum insight into the relationships between the variables that affect the Delta’s health and allow them to make decisions that appropriately weigh the interests of all parties involved.” Palantir

What exactly is “maximum insight”? What does it mean for tools to objectively weigh interests? There are many reasons that the kind decision-making subjects envisioned here doesn’t and won’t actually exist. Many actually existing policy-makers do not even know what kinds of tools are technologically possible – they don’t know what to ask for from programmers. The ones I’ve been talking to tell me things like “someone told me I wanted a dashboard” – in other words, they do not actually know they need the kind of easy to use tools Palantir is offering them, but are instead convinced by the discourse that tells them they do need such tools. And even if decision-makers do seek out dashboards, what they’re often asking for is the “easy button,” the button that gives them an answer spelling out what to do. Yet many programmers will reply that that’s something they can’t give them. It’s not their place to give the one answer, but to provide only maximum insight and present all interests, letting the decision-maker do the rest and pinning accountability on them. Or, merely “suggesting” changing the type of seed to plant, like the tool in the John Deere video. In other words, tools lure decision-makers with the promise of easier, more defensible basis for their choices, while at the same time deflecting as much of that responsibility back towards them.

4. Ideologies of Nature

Finally - and I’m grateful to Dan Cockayne for pushing me on this – we should ask how big data and analysis rely upon and change ideologies of nature. As Neal Smith (1984) argued, not only does capital physically produce nature, but it produces ideologies of nature, bourgeois notions of capital, race, gender and so on as natural. How does TNC’s creation of “pop-up habitats” change how we think about nature? In what ways might we start to think about environmental protection in terms of “surge-demand” and “precision” and understand ecosystems as “real-time” or “pop-up” events? What about the reverse- when instead of talking about nature through data, we start to conceptualize data using our existing languages for talking about nature? What is the effect of describing computation as happening in an “environment,” or more perniciously, naming data as a resource – an object like oil, coal, or gas to be mined and extracted?

Conclusions
What can PE in particular say to the larger conversation surrounding big data?  Below, I name three points political ecology can make in this discussion, three points that summarize the questions and perspectives sketched above. I end with an argument for a political ecological practice of big data.

I believe political ecology is equipped not only to adequately analyze the kinds of situations I sketched above, but that through these sorts of “knowledge politics” critiques it can usefully inform critical data studies. Much in the way of reflexivity on big data concerns its implications for privacy, its heightening of the digital divide, and its use as an academic research tool. Much as political ecologists know the problematic epistemologies of remote sensing, these reports usefully point out that social media data in particular cannot provide the kinds of information or insights many analysts want it to – much data isn’t geotagged, and even the images or tweets that are do not necessarily indicate that the person can be adequately related to that location (boyd and Crawford 2012; Crampton et al. 2013; Wilson 2014). But we should not stop at these critiques of big data as our own method for understanding the world (nor stop at crucial questions of what work big data does as a governance “meme” – Graham and Shelton 2013)  – we must extend them to the particular situations in which big data becomes an applied analysis tool regardless of its foundational epistemological flaws. In terms political ecologists are familiar with: we can illustrate not just data’s moment of production, but its moments of circulation and application. Indeed, Crampton et al. (2013) call for a kind of contextualization or “situating” of big data, as research tool if not means of decision-making. This kind of intimate grounding through case studies is what political ecologists excel at.

We must articulate that big data is produced, and we must show how and why. Not content to merely illustrate that new technologies constrain or afford certain actions by governance actors or land managers, we should provide “chains of explanation” beyond our research contexts back to larger forces at play. We can provide crucial inroads into questions like: what’s the political economy that allows a firm like Skybox to index the earth? Who’s producing these tools? How do private firms rely on and generate value out of public data? How does big data gain “value” for firms or farms? What is it that allows data name value in the world, as TNC and the Natural Capital Project believe it does?

Likewise, big data is something which is practiced with people, in spite of the mystification of algorithms as wholly autonomous entities. Political ecology has a long-standing concern with understanding the concrete, embedded decision-making and social situatedness of land managers – as opposed to discursively idealized subjects (like rational actors) or, to use one of Piers Blaikie’s images, bureaucrats in airplanes (designing a population stabilization program based on abstract concepts and figures the like of which, Blaikie wryly notes, are not in the minds of two lovers as they lay down to bed. This concern should lead us to consider data managers as very much the kind of land managers Blaikie started from and focused on in his Political Economy of Soil Erosion: “actual people making decisions on how to use land.” I mean this in two ways: 1) those making decisions that have concrete bearing on the treatment of particular ecologies, as when TNC’s data analysts, with the help of their algorithms, decide where to extend migratory bird habitat; 2) I also mean it in the sense that those we traditionally consider land managers – farmers, pastoralists, peasants – are fast become data managers in and of themselves, be it as “smart” farmers or as pastoralists incentivize to go out and ground truth RS imagery.

Finally, we must understand the actual effects and outcomes of new techniques of data management. While it is crucial to recognize how data potentially connects domains in novel ways – i.e. when Skybox syncs ecological data with financial data – we should try not to reproduce discourse that “algorithms will rule our world.” If we do, we miss key resistances and ambiguities. We should remember that Skybox supports activists who feel that the technology can offer more visibility to illegal extraction activities be it in the Amazonian or Appalachia. We should be reminded to look for these fractures and to try to widen them, to realize that we can engage in a “politics of measure” that either questions the very measurability of things (Mann 2007; Robertson and Wainwright 2014) or asserts the strategic utility of doing so (Wyly 2007; Cooper 2014).

Which leads to the last big question to pose here: do we as political ecologists employ big data ourselves in support of our “hatchet and seed” mission? Is data just the object of our critique, along the lines I have laid out here? These are questions human geographers are asking themselves right now (Graham and Shelton 2013, special issue), some seeing it as an opportunity to rethink empirical social science in a more interdisciplinary way (Ruppert 2013), with others more cautious, suggesting big data are the epiphenomena of the real issue, and the real need is to question the datafied production of knowledge, to question the idea that big data is truth that simply needs mining (Wilson 2014). I think political ecologists have a lot to gain by reflexively engaging with the analysis of massive data sets, extending PE’s tradition of sitting critically if sometimes awkwardly (Walker 2005) between political critique and ecological fieldwork, cognizant that its own tools of research are often those it seeks to criticize (a la Lave et al. 2013; Turner 2003).

In short, I’m reminded again of Piers Blaikie, when in the 5^th chapter of PESE,he sketches out what a grounded study of peasant political economy in the context of soil loss would look like. Literally sketches out what he calls “a schematic and heuristic device which suggests the way in which these complex relationships (between people, and between people and environment) can be handled.” And what does it look like? 

Blaikie's heuristic for understand land use decisions and soil erosion. 

What else but a massive, complex, messy matrix of variables and objects of study (in this case, households? A big dataset. What is his critical relationship to this heuristic? He describes it as a watch – a watch keeps time, but it does not give its user any clue on how to use time. In the same way, the heuristic and the data it organizes are meaningful only in the context of political economic theory – it “maps” political economy onto the case study and “attempts to calibrate part of it precisely.” In other words, political ecologists might have our big data cake and eat it too. I’m not entirely sure what this engagement would look like concretely – the GLOBE project at the University of Maryland-Baltimore County is a likely candidate - but it seems to me that’s where we ought to be headed.

Can technology save the planet?

A provocation from WWF's chief scientist John Hoekstra that's exactly where I end up in my new post over at Edge Effects. It's a fun intro to quantum computing that backs into a discussion of the assessment and geodesign software tools that conservationists are deploying around the world to better measure restoration interventions, track environmental change, and fight back against environmental crimes like illegal logging. Ultimately, I'm less sure than Hoekstra that the answer to his question is a resounding yes.

The post comes right on the heels of a few interesting stories out the past few days. First, yesterday the Natural Capital Project has launched a MOOC, where you can learn about their toolset. I've written speculatively about some of those tools here before, but it's great to have the chance to go behind the scenes. Second, Hoekstra held a Twitter-mediated conversation last week during SXSWEco, discussing the potential for drones, big data analytics, and other emerging technologies to, well, save the planet. I'm not even convinced yet that what we're seeing conservation right now qualifies as big data - the term seems loosely applied - but Hoekstra led an important conversation about how to do big data in conservation while recognizing issues of security, digital divides, and privacy.

So check out the post, and be sure to bookmark or follow Edge Effects while you're at it. It's an amazing new site run by grad students affiliated with the University of Wisconsin-Madison's Center for Culture, History, and Environment. 

Ecosystem services: some important stories from 2013

I've assembled a non-exhaustive, non-representative sample of stories in the ecosystem services world (broadly defined) from this year that promise to be important in 2014. Here they are - what are yours?

2013 was a year chock full of hotspots of ecosystem services projects and controversy - like the debates in the UK over the country's new habitat mitigation market - but among them, Louisiana stands out. Dubbed "the Himalayas of ecosystem services," there's been more than enough to report on there. There's the very beginnings of RESTORE Act implementation, for starters. The Act will take all the cash BP gets fined in its civil trial and put it towards comprehensive wetland restoration and sediment diversion projects across the Gulf. It's a windfall for the region, and state agencies and conservationists there want to spend the money wisely, knowing what they get for their investment. They've written a raft of plans on how to proceed, and ES feature prominently as the objects of concern and the measures ($ and otherwise) of success. We'll see more projects coming online in 2014 and begin to see their effectiveness.

Speaking of BP's ongoing civil trial, there've been lawsuits left and right in Louisiana this year that revolve around what's the best way to do coastal restoration and who's to blame for the mess of wetland loss. As arguments came to their final stage in BP's ongoing civil trial, the southeastern Louisiana levee board that was created after Katrina to deal with systemic wetland loss in the area drew on some arcane French-era law on levees to launch a multi-billion dollar lawsuit against oil/gas companies for the part their canals have played in destroying wetlands. That drew the outrage of the state's Coastal Protection and Restoration Authority, who says, no, the Army Corps of Engineers and their levees on the Mississippi are to blame. Gov. Jindal had John Barry - the levee board member who advocated for the lawsuit - sacked while CPRA went ahead with its own lawsuit against the corps. The different lawsuits are not just indicative of differing opinions of who's to blame - the corps or the resource extraction industry - but of what's the best way to do restoration: fill in old oil/gas canals, or breach levees to divert sediment to form new land?

If billion dollar plans and lawsuits weren't enough, New Orleans was named one of the Rockefeller Foundation's 100 resilient cities. NOLA will get a "Chief Resilience Officer" funded by Rockefeller and the city will also be the test site for some new software made by the same company that makes data mining tools for the CIA that will help the new CRO figure out what investments in resilience will be most likely to payoff.

In fact, this year we learned that about half of all federal spending that could be defined as related to ES is on tools for mapping, monitoring, and modelling ES. In the Gulf (and for several other places around the world), The Nature Conservancy and partners have put together a slick interactive tool that lets users visualize different investment options for restoration. ES monitoring is moving to automation at the same time that folks are figuring out how to build new maps and models. The Forest Service runs several experimental "smart forests" that collect lots of data on many different environmental indicators, and they (and many other resource agencies) are also (infamously) exploring the use of drone technology to manage forest fires. There's a growing number of tools for measuring and managing ES, and these tools have become fundamental to the ES paradigm (see a great special issue on them in the new journal Ecosystem Services here). Watch for new efforts at big data analysis and ES in the coming year.

2013 saw yet more institutions organizing business and government around seeing environmental degradation as a matter of nature's benefits not having an economic value. That's not to say these new fora and panels actually did anything about the very issues on which they pontificated. I'm thinking here about November's first World Forum on Natural Capital, which was essentially more a feel-good pep talk for corporate leaders and less a hashing out of actionable tasks. It didn't go uncontested and in 2014 we should expect to see the same sort of opposition that we've see for carbon as business leaders aim to price any and all other ES. In December, the new Intergovernmental Panel on Biodiversity and Ecosystem Services convened in Turkey to finalize their first work plan. It's been years in the making and we'll see in 2014 how it starts to get implemented.

The story that most fell under the radar this year was the White House's executive order on climate change adaptation and resilience. This year, about 30 federal agencies developed their first-ever set of plans for how they intend to respond to climate change in their operations and outreach. The EO goes a step further and calls on all agencies to revamp their programs to make it easier to fund projects that are meant to support resilience, for agencies like Interior to manage their lands for resilience, for agencies to develop data and tools for recognizing resilience, and for agencies to plan for climate change risk. All these have the potential to be driving significant work in the coming year and beyond.

The story that wasn't was the US Supreme Court's ruling that appears to constrain regulators' flexibility in determining appropriate compensation for wetland and stream impacts under the Clean Water Act. It's not yet clear whether it'll actually turn out to be problematic. Meanwhile, EPA and ACOE are finally getting around to clarifying what wetlands and streams are within their ambit, a move that environmentalists have long fought for in the legislative sphere. As the draft guidance currently stands, it could bring in millions of dollars more in compensation work yearly because it expands what counts as a water of the US.

The single best piece out there this year on ES was Paul Voosen's history of ES as told through Gretchen Daily, Peter Kareiva, and Michael Soule. He does a brillant job showing how even if it looks like it from 30,000 feet not every conservationist is on board with the project of valuing nature, and he ties this in with an on the ground look at ES "modelling sausage." If you haven't read it yet, go do it now. The runner-up is SciAm's recent piece characterizing the paradigms and debates in wetland restoration today, with a major focus on differing opinions on how to do work in the Gulf.

So what did I miss?

Time to CHAT? Mapping "Regulatory Resistance" in the West

"Mining companies like to say, 'The gold is where the gold is, that's where we need to go,'" said Chet Van Dellen, GIS coordinator for Nevada's Department of Wildlife. "We like to say the animals are where the animals are." New high-tech maps detail wildlife habitat in West, Scott Sonner, 12/13/13

Late last week a coalition of western governors released a new tool meant to help gold miners, transportation designers, energy companies - just about anybody with a natural resource impact - to plan development projects. CHAT, the Crucial Habitat Assessment Tool, going to be one big map for the West, and although it's not entirely filled out yet, the idea is to show those gold miners, hey, here's where our important habitats are. It pays to be clear: the maps are not, as the AP's headline suggests, simply mapping wildlife habitat in greater detail. The tool's resolution is somewhat impressive - down to the square mile - but what it's really doing is visualizing the spaces where project managers can expect to run into problems getting their permits. Some habitats will not be as crucial or as much of a priority as others. The difference may be subtle, but on it turns the role mapping plays in setting the public agenda in environmental governance today.

Here's how CHAT works. Each state has gathered a bunch of data and assigned weights to different kinds of habitats, on a scale from 1-6 (most to least important). The weights are based on information like the condition of habitat as well as economic significance. Each state has its own process, and very often, it's got its very own personal CHAT tool. You should expect no less from the West, and this brand of formal coordination, was likely what got every single western state on board. What CHAT isn't is a project to get all states on board to a similar standard for evaluating habitat significance. It's just meant to project (in the mapping sense) the standards each one already has. Take a look at some of the screenshots of the map if you haven't already, because you can see differences in regulatory regimes on the map.

A CHAT map. From: http://trib.com/news/state-and-regional/the-big-picture-western-governors-unveiling-high-tech-satellite-wildlife/article_7b95cf0e-496e-51df-b324-2a5f929a2232.html

Where you'd expect some important habitats to cross-cut state boundaries, like in Yellowstone, we see that they cut off at Montana, either because the state hasn't gotten around to doing it's categorization yet or because that habitat simply isn't as important to Montana as it is to Wyoming. CHAT is meant to show all western states so that if you're a pipeliner you can see what sort of regulatory resistance you're going to run into across your entire project. Or if you're a gold miner, you can easily see whether it'll be easier to do a project in Utah or Arizona.

It may have been five years in the making, but it's roots go back way further. It wouldn't be much of a stretch to start at the Articles of Confederation to get a sense of what kind of coordination this represents: federalism. Not only does each state gets to develop and share its own particular habitat standards, the map is a way for states to show federal authorities that, hey, we've got everything under control here, much as they are doing with candidate species rulings. More concretely, though, we only have to go back to the mid 90s to understand why we have CHAT now. Federal listing of endangered species like the northern spotted owl generated what boil down to two calls, two sides of the same coin really: state-led environmental policy, and economics-sensitive environmental policy. It'd be no understatement to say that most environmental politics in the West for the past 20 years has been an outgrowth, good or bad, to the issues raised at that time. Utah's and Oregon's governors, on separate sides of the aisle, have developed a set of principles they dubbed, "Enlibra" that they've promoted in the WGA. Enlibra is a new regulatory regime whose ambit is reconciling economic growth and environmental protection, and we've gotten ecosystem services markets and community forestry alike, to name a few examples, out of it. As a prioritization tool rather than a data display tool, CHAT is straight out of the Enlibra playbook.

But here's what it all comes back to: I can't help but feeling that CHAT is like showing your opponent your hand in a game of cards. Of course, it's not like the Nevada Department of Wildlife or some other agency couldn't say, "psych!" and go back on their promise of little regulatory resistance: the map isn't immutable. That also means there's no reason they couldn't go back on their promise of heavy regulatory resistance. The map is a curious legal entity. There's no mandate for all western states to make it: it doesn't have to exist or be used. But it sort of justifies its own existence. All I mean is that by putting the map - described as a "pro-development tool" by the Nevada Department of Wildlife - out there into the world, it's going to be hard to take it back. Developers, regulators, and even the Center for Biological Diversity like it, and that gives it a ton of legitimacy that goes beyond its ambiguous legal status. 

All the cards are on the table now in the West. It's not clear yet whether that's a good thing. It'll probably make regulators' lives easier, for one. There's also certainly a power in being the one to set the terms of engagement. Either way, maps like CHAT are going to play an important role in the making of the relationship between states, nature, and capital in the near term. Just take a look at the interactive maps the Coastal Resilience Network has set up that allows users to choose how important different economic and ecological variables are to determining great places to do restoration. It's not a regulatory map (yet), but you can imagine some of the opportunities that it would afford regulators. It'd make it easier for them to say, for instance, hey, we made the map based on how users (citizens?), not us, weighted restoration priorities. It's not our fault...Stay tuned for more.

Optimal natures

Recently, the Natural Capital Project released its new tool for watershed-based ecosystem services decision-making, the Resource Investment Optimization System, or RIOS (spanish for rivers). It builds on InVEST, NCP's tool for mapping and valuing all sorts of services. Where InVEST could tell you for instance where to invest in a watershed to achieve the best water quality gains (efficiency), RIOS is geared to help you decide between different sets of investment (optimization).

RIOS joins a fast-growing cadre of other ecosystem services decision-making software tools. A short list includes:

Social Values for Ecosystem Services (SOLVES) - the USGS's tool of choice
Integrated Water Resources planning suite  - led by the Army Corps of Engineers
Simple and Effective Resource for Valuing Ecosystem Services (SERVES) - from Earth Economics
i-Tree - USFS built this one
ARtificial Intelligence for Ecosystem Services (ARIES)

These models literally instantiate ecosystem services as a framework by providing the means for framing services - ES is a framework for understanding tradeoffs in managing nature and here are the algorithms for modeling them. One of the key points the tools have in common is that they are spatially-explicit; what might distinguish them is whether they aim to inform either investment or policy decisions. Or, since ecosystem service policy tends toward treating nature as always already an investment (or lack thereof), the distinction is probably: what kind of investment (public or private)?

These tools parallel a number of data analytics firms working with so-called Big Data on the environment. Many, like Cloudera and Ayasdi work with oil and gas companies to visualize optimize the use of their drilling equipment, in the name of preventing future environmental catastrophes. Others, like Remsoft's suite of tools aim to improve forestry practices by incorporating extensive data on tree health, location, etc. - Google and Microsoft are working on similar software for "seeing the trees and the forest."

In short, the stated goal of these models is to "optimize" environmental management, which, for many of them, also means optimizing business practice. Is there a difference between optimal and efficient? For some, maybe not. But Remsoft's tools, they claim, allow you to "understand and manage the supply-demand balance, identify current and future supply chain bottlenecks, manage production and delivery capacity, forecast costs and revenues, and generate plans that stay within budget." Clearly something more than the sense of efficiency as input/output is going on here. Indeed, optimization, in the language of mathematics and computer programming, means to choose the best from among several alternatives given a particular criteria. Yes, the criterion for Remsoft might be $, but that may or may not be the case for USFS's community forestry tool, i-Tree.

Where does all this talk of optimization come from? That's hard to say, and 600 pg. tomes have been written about it. But there is a curious perpendicular conversation happening in the weird realm of biology, computer programming, and artificial intelligence themselves meet: where NCP, Remsoft, and others want to optimize nature, these researchers think nature optimizes. They "use and abuse" evolutionary concepts (note: optimization is not necessarily about selection pressure) as metaphor for informing tech design, their goals ranging from the everyday to the lethal. Researchers have found that ants respond to disaster and disruption - to their environment - in ways that may inform optimal transmission of information over internet protocols. The US military has enrolled apiologists to use bee swarms as an analogue for drone maneuvering. The goal, of course, being to optimize surveillance and kill rates. What brings together the "optimize nature" modelers and the "nature optimizes" researchers and designers is the idea that the environment serves as a model for our treatment of it.

This is not to get us lost in the thickets of environmental philosophy or social theory. The question is: on the ground, what is lost and gained by thinking in terms of optimizing ecosystem services? Who stands to win and lose? These models are meant to inform land use decisions, and in doing so, they help to bring about the optimized world they only purport to represent. If you model it, they will come. In this performance, the way the models are programmed matters. And what differences are there between the flavor of optimization led by the conservationists using NCP and the timber managers using Remsoft's Spatial Optimizer? One has to inform policy, the other business - can optimization serve as an adequate guiding concept for both?

Code/Nature

I've posted the slides from a presentation I gave at the Dimensions of Political Ecology conference here at the University of Kentucky recently. My argument is pretty straightforward: to have an ecosystem service, like wetland water storage and delay, you have to be able to show where it that service exists in the landscape and software tools like Excel, ArcGIS, and online mapping utilities are really fundamental to that calculation. It's kind of like the old thought experiment - if a tree falls in the forest and no one's around to hear it, does it make a sound? Does a wetland provide a flood mitigation service if it is in the middle of nowhere? I don't mean to get all philosophical on you, but the basic point is that ecosystem services - as valuable benefits of nature to society - might not exist as such if environmental agencies and others weren't able to map where they exist and who they benefit. So, the talk is a modest call to pay attention to regulatory, entrepreneurial, and conservationist exercises in mapping services, like InVEST from the Natural Capital Project. I've pasted the text of the talk below; each paragraph corresponds with one slide.

DOPE 2013 presentation_Eric Nost_Code/Nature from ericnost

1. Ok, so I promise you that this picture of people idling in line to get tickets at the Portland airport is relevant to what I really want to talk to you about today: market valuation of ecosystem services. I’m going to show that airport terminals, in fact share a lot in common with the restoration sites through which Oregon conservationists and entrepreneurs value ecosystem services for market. I make two calls in my talk: 1, for ecosystem services researchers to pay attention to spatially explicit ECS valuation, and to ask for whom such valuations work. 2, to call upon political ecologists to keep paying attention to spatial visualization techniques, but to also pay attention to other technologies through which spaces – like airport terminals, and restoration sites - are made and valued.

2. Welcome to the Half Mile Lane site in exurban Portland, Oregon. It provides a number of ECS. The wetland you see stores and delays water, which mitigates flood impacts for downstream homes. The stream, which you can’t really see, provides habitat for salmon that migrate into the foothills of the Coast Range. A couple of years ago, state environmental agencies and conservationists undertook ecological restoration on the site, turning old farmland and a straightened ditch into a productive wetland and stream.

3.That these services exist as services is spatially dependent, or contextual. I’ll give you three quick quotes to show how. As you see here, the international think-tank for ecosystem services accounting, TEEB, note that you have to have a specific site to have a service. The work of the wetland at HML to store and delay water matters only because there are homes in the 100 year floodplain downstream of the site that benefit.

4. Long-time ECS researcher Gretchen Daily concurs. She calls for focusing on the right places in the landscape. HML’s position, for instance, allows it to slow down and cycle the increased runoff from logging operations.

5.Finally, lest you think this focus on landscapes is the domain solely of pundits like TEEB and Daily, consider what the Oregon DSL has to say. One phrase we often hear in the ECS world – we did in the TEEB quote - is “value of nature”. What does value mean? For DSL, it means the opportunity to provide an ecological function. Crucially, this opportunity is location-based.

6. Now, DSL oversees wetland and stream ECS markets in OR. The way these markets work is entrepreneurs restore ECS on sites like HML and sell the ecological benefit they create, as a credit commodity, to housing developers, DOTs, and others that are paving over wetlands and streams in different parts of the watershed. In fact, HML here is one such mitigation “bank” of restoration credits. There many different kinds of actors in the market. You’ve got state agencies like DSL with statutory obligations and ecological inclinations, but also NGO groups with conservation missions, and of course entrepreneurs looking to do banking for profit. This raises a key question: to what extent do market-makers account for context, or value? Or for them, is a service just a service, no matter where it’s provided? How do market actors decide where it is most ecologically valuable to do restoration?

7. What I want to show in the rest of the talk are three things:
1. Digital tools like Excel and GIS allow the OR market to account for context
2. However, these tools and the algorithms that underwrite them are not mirrors of nature. Rather, tools reflect the interests of market actors
3. In Oregon, state agencies and conservationists may have the upper hand in defining and accounting for ECS values.
I make these arguments by outlining three moments in which value is accounted for in OR’s markets. I end by putting out a couple of calls for future research.

8.I’ll tell you first about the assessment moment of ecosystem services valuation in Oregon’s wetland and stream markets. Entrepreneurs hire consultants to do a key part of the work of the market: assess restoration success. In assessment, consultants utilize Excel spreadsheet-based calculators of ecological process. One of these calculators is called the Oregon Rapid Wetlands Assessment Protocol, or ORWAP. There are ones for salmon habitat, water temperature, and other services, but they’re all conceptually very similar, so I’ll focus on ORWAP. Most of them were in fact written by the same person, under contract from DSL and US EPA. He’s been developing these assessments for about 30 years now, which is when he first made a split in assessing ecological process or function, and value.

9.Consultants score functions in ORWAP through a series of multiple choice questions about things like seasonal surface water extent.

10. Consultants also do work in the office, employing several online mapping tools for an assessment of value. Here’s one called Oregon Explorer. Hydric soils are the orange/yellow, but we also see the 100 year floodplain downstream of the site. OR Explorer knows, too, about rare species on the site. It’s bringing a lot of data from beyond the boundaries of the site together, and showing it to the user in one frame. The user can thus answer questions in ORWAP about landscape context by using OE to, for instance, draw a 2 mile radius circle around the site to see how many other similar habitats the site is connected to in the area.

11. What offsite stressors, and risks consultants find in their assessment, regulators can consider in approving or denying a banker’s plan. For instance, regulators often focus on reed canary grass, an invasive species that can spread rapidly on a restoration site from without and foil the project. They question whether a site and its landscape surroundings will, in the end, prove valuable if there is too much RCG around. Theoretically environmental agencies can in this moment deny a banker’s proposal to work on a certain piece of ground that is particularly susceptible to weeds. In reality, however, they are more likely to just modify the banker’s site selection, perhaps by asking them to put more money into a long-term management.

12. Finally, there is a market moment to value’s measure. What conservationists want to see happen in the market is that when a banker brings a site to the market, to get their credits to sell, the amount they get depends in large part on the location of their project. They would get the full amount if they were in what’s called a priority area and less if they were not.

13. These areas are an aggregation of habitat sites mapped by state environmental agencies, and put together by TNC. To be clear: this isn’t how the market currently works, but regulators do use GIS to look at whether bankers are siting in priority areas, and conservationists are pushing for this trading ratio protocol to be adopted.

14. The problem is that if a banker had to do work in a priority area lest they not get as many credits as expected, that could be at least a short-term constraint to the market, especially if land prices in priority areas were higher. In general, through all these moments, how state agencies, with help from conservationists, want to assess value via site selection, will constrain entrepreneurs. At the same time, agencies and conservationists are not now fully determining site selection. Rather, the maps they make give them something to point to and say, bankers should go here rather than there. Or, in other cases, spreadsheets like ORWAP in tandem with maps let them say, look there’s a quarry upstream, don’t go there.

15. The state and conservationists will want these gestures to be assertive. What underwrites their ability to point to a map in the first place? In large part, code. I want to turn quickly to work in information technology studies as a way of understanding valuation in OR. Recently, geographers Rob Kitchin and Martin Dodge wrote a book called Code/Space in which they argued that spaces are increasingly constituted by computer code. An airport terminal is only the kind of space it is if the 0s and 1s that run the check-in stations work; when they crash, the space turns from a hub of international commerce into a den of frustration. In the same way, a restoration site or classes of ecosystems cannot be priority spaces if the software that codes them that way does not work the way it is supposed to. In OR, the ability of ArcGIS to display and combine layers is crucial because layers allow for combining different ecological interests into priorities. Code underwrites how web mapping utilities aggregate different data and draw circles around it as well, allowing foroffsite visualization and valorization. And ORWAP can’t generate a value score without Excel’s ability to run calculations across so many different variables. The state/conservationists’ position is code-dependent.

16.So to wrap-up: Yes, markets in ecosystem services restoration, at least in OR, do have a spatial calculus of value, as TEBB and Gretchen Daily hope for. State agencies and conservationists work to make the calculation their own, and deploy it to their own ends. But it remains unclear how successful they can be. Their ability to write the code and utilizing the tools with which they see value in space will be crucial to their future market-making work.

17. And valuation tools will be worthwhile pay attention in other markets as well, as decision-makers continue to call for the price valuation and marketization of ECS. So ECS researchers should continue looking at the work of spatially explicit valuation, but ask, as Norgaard did, for whom does ECS governance work for? Conservationists? Regulators? Bankers? … Landowners? PEists are indeed well equipped to talk about winners and losers. But this is also a call to PEists to keep looking at code. We’ve looked at spatial visualization technologies before, but need to continue, and to look at code not just in GIS but in Excel and other programs. We can do this in partnership with scholars of the geoweb.