Dear MGET users,

We are happy to announce the release of MGET 0.8a57 (download here). Highlights include:

·         Support for ArcGIS 10.3. Upgrade instructions here. Note that ArcGIS Pro is not supported, only the traditional ArcGIS Desktop applications (ArcMap, ArcCatalog).

·         New tools for the HYCOM + NCODA Global 1/12° Reanalysis (GLBu0.08). This is a global 4D physical ocean model running from late 1992 to the present day, spanning 80 S to 80 N, in a equirectangular (a.k.a. “geographic”) coordinate system at 1/12° resolution, 40 depth levels, and 1 day and 3 hour timesteps. The model includes water temperature, salinity, sea surface height, and currents. Many MGET users requested support for these data and we are happy to provide it. Note that the GLBu0.08 “reanalysis” only spans 1992-2012. For 2013 and later, MGET seamlessly concatenates the GLBu0.08 “analysis” datasets (expt_90.9, expt_91.0, expt_91.1) to the end of the reanalysis to extend it to the present day. Also, HYCOM does not offer forecasts with GLBu0.08 in a way that we could easily ingest with MGET. To get forecasts with MGET, you must still use the MGET tools designed for GLBa0.08.

·         New tools for finding fronts using the Canny edge detection algorithm. We developed these as an alternative to the Cayula-Cornillon front detection tools which are a popular part of MGET but are challenging to use due to their large number of parameters. The Canny algorithm only has four parameters and they are easy to manipulate. We also noticed that the Canny appears to be less sensitive to spatial resolution than the Cayula-Cornillon algorithm, which appears to require tweaking of some parameters based on resolution to get the best results.

·         Support for the tw (Tweedie) and nb (negative binomial) distributions to the Fit GAM tool. Another popular feature of MGET are the statistical modeling tools. The Tweedie and negative binomial distributions are useful in spatial ecology problems where the data contain a lot of zeros. MGET relies on the R mgcv package to fit GAMs. Last year, Simon Wood enhanced mgcv so that it will automatically estimate the statistical parameters of these two distributions as part of the model fitting process, making them very easy to use. MGET now exposes this capability.

·         Support for R 3.1.3 and numpy 1.9.2. MGET now supports the latest versions of R and numpy.

If you have any questions, feel free to contact me directly.

All the best,

Jason Roberts ()

You are receiving this message because you have subscribed to the mget-announce or mget-help email lists. Click those links to unsubscribe.