The new SimWeight empirical transition potential modeling procedure in LCM. SimWeight is based on a modified K-nearest neighbor machine-learning algorithm. It offers the strength of modeling that is normally associated with the Multi-Layer Perceptron neural network without its complexity.
Land Change Modeler Enhancements
- A new REDD tab to support projects aimed at Reducing Emissions from Deforestation and Forest Degradation.
- A pioneering new land cover change modeling procedure, SimWeight, a machine learning procedure that has proven to yield results that rival that of the Multi-Layer Perceptron with minimal (and easily understood) parameters.
- A new land cover preprocessing procedure called Harmonize that coordinates the land cover layers in terms of their spatial characteristics, background masks and categorizations.
- An integrated link to the popular MAXENT procedure for species distribution modeling.
Fig. 2 An example of the Maxent interface in LCM and its output. In this example, the range of Dromiciops gliroides is modeled based on a collection of observation points and a set of environmental variables. This new Maxent option extends the existing group of species distribution modeling tools.
Fig. 3 A Multichannel Singular Spectrum Analysis (MSSA) of monthly anomalies in lower tropospheric temperature using an embedding dimension of 13 months. The embedding dimension establishes a time frame of focus in the analysis of evolving spatio-temporal events, and also indicates that the analysis will be run as an Extended PCA on 13 lagged versions of the same series.
Fig. 4 IDRISI has long had an exceptionally strong suite of neural network classifiers including Multi-Layer Perceptron (MLP), Self-Organizing Feature Map (SOM) and Fuzzy ArtMAP neural networks. With this release we add an experimental Radial Basis Function (RBF) neural network for the classification of remotely sensed imagery.
Fig. 4 The new Chain Cluster procedure has strong similarities to the logic of the Fuzzy ArtMAP neural network procedure, is particularly simple to parameterize when used with standardized image inputs.
Earth Trends Modeler Enhancements
- Principal Components Analysis (PCA) and Empirical Orthogonal Teleconnection (EOT) analysis now offer extended modes where multiple data series can be analyzed simultaneously.
- Multichannel Singular Spectrum Analysis (MSSA) and Multichannel Empirical Teleconnection analysis are now included analyzing patterns that evolve in space and time.
- All Principal Components Analysis procedures now offer both T-mode and S-mode orientations for analysis – the first GIS/Image Processing software system to offer both.
- A new procedure for Canonical Correlation Analysis (CCA).
- An auto-arrange feature whereby IDRISI automatically arranges map elements such as titles, legends, scale bar, insets, etcetera.
- The Composer window has changed; besides a new a new interface design, it is sizeable in order to better handle long file names and compositions with many layers.
- Map windows can now be very simply resized by pulling out or pushing in the lower-right corner.
- With the Selva edition, IDRISI breaks through the Windows 32-bit display architecture, with the ability to now display images much greater that 32,000 rows and columns, depending on your hardware.
New Analytical Modules
- A Radial Basis Function (RBF) neural network classifier has been added to complement the existing suite of neural network classifiers.
- A Chain Clustering procedure has been added to the basic clustering tools available in IDRISI.
- A Durbin-Watson module has been added independently of ETM to map locations where serial correlation is present.
New or Revised Import/Export Modules
- Support for KML files (Keyhole Markup language, used by Google) has now been extended to include the import and export of point, line and polygon files as well as raster images.
- An import routine to convert MODIS tiled imagery to IDRISI raster format. The files are imported and then the tiles are mosaicked, with options to mosaic tiles of different geographic extent.
- The import utility for MODIS Quality Control data has been extended. Quality control maps can now be generated for vegetation indices using MODIS collection 5, land surface temperature collections 4 and 5, and Albedo collection 5.
Every so often we take the opportunity to substantially revise modules that we think could benefit from a different
approach. With the Selva release, we have focused on the core of the distance-based routines – DISTANCE, COST,
VARCOST, DISPERSE and BUFFER. The optimization is substantial, with most routines running considerably faster.