Google Summer of Code Robert Scheller 2015 Ideas
The Dynamic Ecosystems and Landscape Lab at PSU is a foci of development for the LANDIS-II forest simulation model. LANDIS-II simulates forest succession, disturbance (including fire, wind, harvesting, insects), climate change, and seed dispersal across large (typically 10,000 - 20,000,000 ha) landscapes. LANDIS-II tracks the spatial distribution of discrete tree and shrub species and has flexible spatial and temporal resolutions. Learn more about our research and see our publications list. LANDIS-II advances forest modeling in many respects. Most significantly, LANDIS-II, 1) is completely open-source at all levels with extensive documentation, 2) has a large library of ecological processes to choose from, including many explicitly designed to track landscape carbon dynamics, 3) has flexible time steps for every ecological process, 4) uses an advanced architecture that allows rapid model development and easy distribution and installation of model components. We are also very proud of our large and active community of users and developers. LANDIS-II manages and executes discrete extensions (modules or plug-ins). Each ecological process is programmed as an independent extension that interacts with the landscape through an explicit interface with the core LANDIS-II program. The user specifies which extensions will be used to best simulate the forest dynamics of interest. LANDIS-II allows scientists to easily develop and share their own extensions. These proposals address various substantial needs of the LANDIS-II community.
LANDIS-II currently runs 90% of the time on local desktops. This has served us well for the past 10 years but people are attempting to use increasingly complex and large landscapes that tax even the best workstation. Therefore we need a centralized 'cloud-deployment' that would allow users to upload input data and run their simulations and receive output via the Google, Microsoft, or Amazon cloud platforms. This would also free people from having to do unique installations to their local Linux cluster or similar. Consideration would need to be given to how users are billed by the service.
We recently developed a metadata library for LANDIS-II that generates metadata xml for all model outputs. The design is flexible and accommodates csv delimited data (e.g., number of fires per time step) and maps (raster data in a variety of formats). However, the current design is limited by fixed columns lists, unlabeled nominal data, and poor integration across extensions. It has been implemented for a limited number of model extensions. The next version of the metadata library should substantially improve performance and portability and should be propagated to all extensions and to the core. LANDIS-II is written in C#.
We now have an output metadata library that is capable of generating xml metadata for all model extension outputs. The input is, however, limited to text files. Input metadata would allow the option to read in data via xml. Doing so would in turn open up many possibilities for a dynamic input interface, implemented either as a stand-alone application or as a web interface that would generate necessary inputs that a user could copy to the computer for execution. This metadata plus interface would free scientists from implementing complex input operations and would enable the creation of a dynamic interface that responded to rapid changes to required inputs. LANDIS-II is written in C#. The input interface could be written in any Windows-compatible framework.