I’ve been working on a time-lapse visualization (one above, more below) of coastal Louisiana land loss for a few weeks now, and I think I’m finally at a point where I can share the few I like most.
Trying to animate the amount of land loss since the first well was permitted in the Terrebonne-Lafourche area in December 1916 can be a visual disaster: too many data points, too cluttered, progression too fast/low, too blocky or incorrectly paced, resolution not high enough, or resolution too high for it render properly, not the right frame rate, etc. etc..
Believe it or not, creating the videos took more time than data processing, though it was less hands-on.
The biggest obstacle in creating these videos was not simply showing the changes – rather, most of the difficulty came from processing the actual data to be able to animate the changes with a fine temporal resolution.
The visualization uses four main data layers for this project:
- Land-Water changes for the past 100 years
- Land-Water environment as it is today (or 2015)
- Oil and gas sites with date information
- Oil, gas and sulphur canal locations and construction dates
USGS published a widely-used report in 2011 showing land loss in coastal Louisiana from 1932 through 2010, with varying breaks in temporal resolution. But the gaps in time are inconsistent and, in the earlier records, unacceptably large.
Though the data ranges from 1932-2010, the breaks render the data useless for a large time-lapse project such as this. Only 17 data sets were used to compare total land and coverage between images, which were taken:
- 1932, 1956, 1973, 1975, 1977, 1985, 1988, 1999, 2002, 2004, 2006, 2008, 2009, 2010
Since the oil and gas wells data come in as day, month, year, the USGS information cannot simply be set behind it and represent the peaks and valleys of land loss for a period of, say, 1956-1973, which includes both the period of time in which land loss as well as oil and gas development peaked.
To remedy the data inconsistencies, I used a canals shapefile that I manually drew based on coastal bathymetry and elevation, hydrological flows, satellite and aerial imagery, and data mining of production and construction records. I connected the oil and gas well data with the canals data – both as a QA/QC measure and as a tool to classify each data set into month and year from the wells file (day, month, year or unknown).
The canals and well data was then paired with the land loss data – the canals themselves already visible in the USGS layer, I could attribute it with a month and year based on the paired production and canals data. For example, the 2011 USGS data would show a linear feature eroding between 1956-1973, but since I know when the well and canal construction commenced, I can now classify those specific raster cells into the new date format (e.g. January 1967).
Using land loss erosion and loss rates from both USGS and the Louisiana Coastal Protection and Restoration Authority (CPRA, where I used to work, remember?), I could pair each individual cell within a distance of a known land loss date, expanding outward with a logarithmic buffer in the spatial model,
This analysis was used for land loss that has been previously classified as caused or primarily caused by oil and gas development in large sections (USGS, NASA, NOAA), but either no canals data was available or the gaps in time were too large and needed to be broken down. In addition to the above processes, I added about 50 more data points across varying time periods using Landsat data so the breaks were more consistent.
The current land-water make-up was the easy part – Classify a 2015 high-resolution ortho mosaic and up-scale to a 30 meter resolution to better match the previous data, and use as the base map.
Of course not all land loss in south Louisiana can be attributed to oil and gas production alone – hence the need for an analyses that considered the other processes and factors that affect coastal land loss. Only those pixels of land loss which overlap canals/well positions and those areas which have been previously studied and classified as loss due to oil and gas production.
I’m in the development of another that shows land loss with both oil/gas locations and hurricane tracks, although a combination that includes storm surge would be more representative of the damage by each. Of course that map would look messy and cluttered, so I split them into a series of videos, maps and graphics that considers each variable and its respective role in creating the coastal crisis today.
Additional data layers were used for clarification and validation, including USGS land loss attribution layers, an assortment of data from NOAA, Scripps and NASA, and other publically available data layers.
In addition to the video at the top of this post, I worked on several other renderings, toying with the different visual and audio elements to see which one viewers would respond to most.
Which one is your favorite? Watch and vote below!