Enabling joined-up decision making with geotemporal information

Abstract:

While the use of geospatial data to assist in decision making is becoming increasingly common, the use of geotemporal information: information that can be indexed by geographical space AND time, is much rarer. I will describe our scientific research and software development efforts intended to advance the availability and use of geotemporal information in general. I will show two recent examples of "stacking" geotemporal information to support land use decision making in the Brazilian Amazon and Kenya, involving data-constrained predictive models and empirically derived datasets of road development, deforestation, carbon, agricultural yields, water purification and poverty alleviation services and will show how we use trade-off analyses and constraint reasoning algorithms to explore the costs and benefits of different decisions. For the Brazilian Amazon we explore tradeoffs involved in different deforestation scenarios, while for Kenya we explore the impacts of conserving forest to support international carbon conservation initiatives (REDD+). I will also illustrate the cloud-based software tools we have developed to enable anyone to access geotemporal information, gridded (e.g. climate) or non-gridded (e.g. protected areas), for the past, present or future and incorporate such information into their analyses (e.g. www.fetchclimate.org), including how we train new predictive models to such data using Bayesian techniques: on this latter point I will show how we combine satellite and ground measured data with predictive models to forecast how crops might respond to climate change.