Feeding 9 Billion: Innovations in Agricultural Modeling
A new class of computational methods has been developed to construct distributions of where such monitored organisms are most likely to be found in space and time using this data, and are much more accurate than previous methods when dealing with large sets of data.
Previous methods, called Kernel Methods, were based on associating a parametric distribution, such as a normal distribution, with each location point. The new methods, referred to as LoCoH (local convex hull) methods, are essentially non-parametric kernel methods where the kernel associated with each data point is constructed directly from that point and a given number of its nearest neighbors. These methods have application to all types of ecological and biological resource management problems, but will prove especially useful in evaluating the spatial needs of threatened species and designing parks to conserve them.
This study will be published on February 14, 2007 in PLoS ONE, the international, peer-reviewed, open-access, online publication from the Public Library of Science (PLoS)
