Dear MGET users,

Recently I discovered an important bug in MGET’s tools that access AVISO data. If you use MGET to access AVISO data I strongly urge you to upgrade to MGET 0.8a28, released on the MGET website this morning.

The bug affects all tools that access AVISO Delayed Time (DT) datasets. Near Real Time (NRT) datasets are not affected. Tools that access both kinds of data work properly with NRT data but not with DT data, as described below.

The problem relates to AVISO’s publication of experimental DT data that increases the temporal resolution of the dataset from 1 image per 7 days (which has been the resolution of the DT datasets for many years) to a resolution of 1 image per day. AVISO announced this experimental data in April 2011 and released it on their FTP server in alternative subdirectories named “daily_exp”. Any user applications that accessed AVISO data through the FTP server would continue to access the traditional 7-day images unless those applications were specifically changed to access the experimental data in the daily_exp subdirectories. AVISO stated “production is experimental and temporary but may be continued if you show us its value for your works” and gave further instructions for providing feedback.

In addition to the FTP server, AVISO also makes their datasets available via the OPeNDAP protocol. This protocol provides an advantage over FTP in that it allows data to be downloaded for user-specified bounding box. For example, if you are only interested in data for the north Pacific ocean, you can download just that data using a spatial bounding box; with FTP you must download global files and then extract the spatial subset afterwards. The MGET tools access AVISO data with OPeNDAP instead of FTP, to maximize download efficiency.

Unfortunately, sometime between April and October—I am not sure when—AVISO decided to roll out the experimental 1-day data to their OPeNDAP server without announcing that they were doing so. They did not create an alternative location for the experimental data. They just switched the main OPeNDAP URLs to distribute the experimental data instead of the operational 7-day data.

The MGET tools in 0.8a27 and earlier assumed that all AVISO DT datasets use a 7-day time increment. As a consequence, they assumed that each block of 52 images represented one year of data, when they actually represent only 52 days in the new experimental datasets. If you requested that the tools download the first year of data available in a AVISO DT dataset, the tools would erroneously download the first 52 days of experimental data and assign weekly dates to those images.

The result is that any AVISO DT data accessed between April 2011 and now with MGET 0.8a27 or earlier is completely wrong. Any scientific results based on that data are probably invalid. As I have not been actively using the AVISO data recently, I only found out about this last week, after an MGET user noticed that the data downloaded through MGET did not seem to match data available on a NOAA website.

Starting with MGET 0.8a28, the tools will automatically detect the time increment of any dataset accessed on the AVISO OPeNDAP server. For now, that means most DT datasets will have 1-day increments. If AVISO ultimately abandons the experimental 1-day data, they will presumably switch the OPeNDAP server back to serving the 7-day data. If so, the tools will automatically handle the change. If you use AVISO data, I urge you to upgrade to 0.8a28 immediately.

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

Best regards,

Jason