Issue link: https://read.dmtmag.com/i/97084
It Takes a Process to Monitor Gradual Change DYNAMIC DECISION POINTS W hen discussing temporal databases, people often think about events. These are discrete changes that happen at specific points in time (e.g., an installation or an incident) or are at least recorded at specific points in time (e.g., inspection, sampling or survey). Asset databases are built on discrete changes; equipment is installed, replaced or upgraded on a certain date. Aerial photos are taken annually or biannually and have a specific capture date that drives metadata (e.g., leaf on or leaf BY ERIK SHEPARD off). Social media, the disruptive force behind crowdsourcing, is generated with a timestamp reflecting the author's state of mind at that point in time, but also reflecting the thoughts and norms of the larger social consciousness at that time. It Still Makes a Sound Erik Shepard is director and principal of Waterbridge GeoDesign Inc.; e-mail: erik@ waterbridge.biz. 30 But the worldview that "all changes are discrete and created or observed at specific times" reflects a bias toward anthropocentrism. If a tree falls in the woods and no one is there to hear it, does it still make a sound? The truth is that change often is gradual and based on processes only partially seen or understood. Tectonic shifts happen gradually over centuries or millennia. Infiltration of pollutants into groundwater happens over years. Climate change happens over time—perhaps decades for anthropogenic change or millennia for naturogenic change such as an interglacial period. Component inputs to this gradual change may be understood, but not the complete processes. Such gradual changes are studied by building models of the process, to the best of our ability. Geospatial tools are key to these types of models, because changes happen in place as well as time and often are affected by changes nearby. Antarctic glaciers undergo faster erosion at the sea boundary, because of the sea's saltiness. Random inputs are modeled with stochastic processes. Samples and surveys, taken at discrete points in time, provide input to develop continuous time models. These models, in turn, become inputs into activities performed by geospatial systems. Model-driven risk G E O W O R L D / N O V E M B E R 2 O 1 2 The vision of dynamic GIS and decision support is driven by the ability to describe gradual changes that occur through ongoing processes. is used to perform planned maintenance activities. Model-driven inputs are used to inform sustainable design. Real-time inputs from sensor platforms are used to build dynamic systems that respond to change, which update and calibrate the models. Incorporating models as output and input helps realize the vision of dynamic GIS. GIS is the technology for interfacing business with the world, which is ever-changing. Real-World Calibration Of course, all models need calibration, which requires real-world inputs. For gradual processes, infrequent surveys look more like discrete events than slowly changing shifts. One way to address this deficiency is by collecting data more frequently; instead of a year, collection in near real time. The best way to currently do this is through wireless sensor networks, which work together and can be deployed using several different strategies. Sensors can collect data at fixed time intervals (e.g., once per second) or can be programmed to respond to changes in the environment. For example, a sensor measuring temperature may record data each time the temperature changes by five degrees. Technically speaking, this is still discrete time, but the closer it gets to continuous recording, the closer it gets to being able to measure gradually changing values. Using data from these sensors as model input, there are duel benefits: a model of gradual change as well as continuous calibration as additional data are collected. For a model of a single variable, this is nice to have. But real systems in the real world are complex, comprised of many variables—often interdependent, co-varying attributes. The ability to use sensor data for each of these variables fine tunes models using near-real-time inputs. Calibrating these models via infrequently sampled data would make fine tuning difficult and real-time fine tuning impossible. The vision of dynamic GIS and decision support is driven by the ability to describe gradual changes that occur through ongoing processes as well as discrete changes driven by events. Models that describe processes—and dynamic, near-real-time inputs from field-deployed sensors—will become more critical as the challenges from our increasingly resource-constrained world are faced. Government Special Issue