By Scott Stewart, NBT Solutions, scott.stewart [at] nbtsolutions.com
Marine spatial planning (MSP) is a critical management process that has proven to be a successful approach to managing the human-to-human and human-to-ecosystem conflicts that are increasingly emerging in our oceans and lakes. In its report titled Marine Spatial Planning – A Step-by-Step Approach Toward Ecosystem-Based Management, the Intergovernmental Oceanographic Commission (IOC) defines MSP as a “public process of analyzing and allocating the spatial and temporal distribution of human activities in marine areas to achieve ecological, economic and social objectives that are usually specified through a political process.” For the most part, the process is intended to identify and resolve conflicts among humans who use these resources, and conflicts between human uses and the marine environment itself.
MSP aims to maintain an objective process in which the values of all stakeholders are considered. Much of the work involves analyzing data, including the compilation, management, and sharing of spatial, temporal and descriptive data sets—primarily geospatial data that can be analyzed using GIS technologies.
The geospatial data sets that support the MSP process can be divided into several thematic categories or layers, including jurisdictional boundaries, federal regulations (for example, the Marine Plastic Pollution Research and Control Act), navigation and marine infrastructure, geology and seafloor, habitat and biodiversity, and human use. An MSP planner then attempts to identify and reconcile areas where there are visible conflicts among uses.
Many of the geospatial data sets used in the MSP process already exist or can be compiled with minimal effort and coordination. However, the lack of data representing human use or the socioeconomic dimension of coastal and marine resources is well-documented and accepted by MSP planners. Called the “missing layer,” the human dimension of the marine environment has been used only sparingly in the MSP process, and even less in the GIS-based decision support systems on which the MSP process relies. In its previously referenced MSP report, the IOC states: “The human dimensions of the marine environment are widely recognized as important to include and integrate into decision making. However, there are few layers of socioeconomic information that one might combine with the biophysical in, for example, spatial suitability analyses for the establishment of a marine protected area."
One possible way to include this “missing layer” is to mine social networks for human-use activities—Twitter, for example. Launched in 2006, Twitter allows subscribers to generate, read and send 140-character posts known as “tweets” that contain a variety of information and content that depends on the subscriber’s interests, hobbies, profession, culture and various demographics. In 2011, there were an estimated 300 million Twitter subscribers that generated more than 200 million tweets per day. With the enormous amounts of information being generated as daily tweets, one might expect that some of these posts contain information on human uses and experiences of the coastal and marine environment. Boaters may tweet about the wonderful and fun day they are having on the water, while a fisherman may take a picture of her latest catch and post it as a tweet. The key, then, is to be able to parse these tweets and derive human use from them—no easy task.
To accomplish these goals, NBT Solutions leveraged its TweetSpot software package framework to collect Twitter feeds from these “citizen sensors” using the Twitter application programming interface. Twitter feeds are then categorized, semantically annotated and formatted using a combination of vocabularies and ontologies (domains). The goal was to have the resultant platform be flexible enough to respond to changes in vocabularies and ontologies in order to help MSP planners quickly compile human-use indicators to include as part of GIS-based decision making.
A simple categorical ontology mechanism can be used to map high-level MSP human uses to specific words in tweets. For example, “ski,” “boat” and “fishing” could all be mapped to the category of “recreational boating.” This mapping was designed to be flexible enough to add or remove words quickly from categories through an administrative interface. Words could be a part of one or many human uses (for example, “fish” could be part of both the “recreational boating” and the “recreational fishing” categories). Results are then placed on a map as pin points, with a statistical breakdown of all human activities in the area of concern.
The results show promise, albeit with some limitations, and with quite a few ideas for improvement. For example, to improve results matching, NBT plans to include semantic and natural language parsing to increase the utility of the match beyond a simple keyword match. As it stands, there will be a large degree of false positives from ambiguous word usage (“shooting fish in a barrel” could be mapped to a fishing activity, for example). Also, map visualization could be improved by using convex hull and heat mapping methodologies to show “clouds” of activity or areas of use, instead of merely displaying points on the map.
Even with limitations, the promise of mining Twitter and other social feeds for human use using locational (and also temporal and seasonal) data embedded in social posts is evident, and worthy of future research and investigation. Using these and other tools, the “missing layer” in MSP can begin to be shaded in.
A demo of the application is available at http://nefms.nbtsolutions.net/tweetspot/web/.
Scott Stewart is the chief technology officer for NBT Solutions LLC. He is based in Williamsburg, Virginia, US. Scott has 20 years of experience in a wide variety of fields, including database design and implementation, data quality and assurance, scientific data processing, environmental monitoring, cryptography, and web services architectures in both the government and private sectors.