A Legend in the Making: Mapping Mangroves in the Florida Everglades

Category: Land Cover Change & Disturbances
Project Team: Everglades Ecological Forecasting
Team Location: NASA Langley Research Center – Hampton, Virginia

Landsat 5 image highlighting the Florida Everglades mangrove extent using a cloud filtering algorithm and false color bands. Image Credit: Everglades Ecological Forecasting Team

Landsat 5 image highlighting the Florida Everglades mangrove extent using a cloud filtering algorithm and false color bands. Image Credit: Everglades Ecological Forecasting Team

Authors:
Donnie Kirk
Caitlin Toner
Emily Gotschalk
Rachel Cabosky
Brad Gregory
Candace Kendall

Mentors/Advisors:
Dr. Kenton Ross (NASA Langley Research Center)
Dr. Hans-Peter Plag (Old Dominion University, Mitigation and Adaptation Research Institute)
Dr. Marguerite Madden (University of Georgia, Center for Geospatial Research)

Abstract:

Mangroves act as a transition zone between fresh and salt water ecotones by filtering and monitoring salinity levels along the coast of the Florida Everglades. Mangroves offer specialized habitats and provide shoreline stabilization, critical to a region beset by tropical storms. These areas give way to marshlands that depend on the services mangroves provide, as they require larger quantities of freshwater. In an attempt to assist in maintaining the health of the threatened mangrove species, efforts have been made within the park to rebalance the ecosystem. The National Park Service requires a way to track the distribution of marshes and mangroves. The DEVELOP Ecological Forecasting team used Google Earth Engine and satellite imagery from Landsat 5 and 8 with comparison to existing vegetation maps. The team was able to conduct the classification to display mangrove and marsh regions in 1995, 2005, and 2015. After considering several geospatial analysis platforms, Google Earth Engine was selected due to the accessibility of its open source platform. In order to make the process replicable for the Everglades National Park, a comprehensive methodology of classifying mangroves in Google Earth Engine was developed. The process was designed with the intent that the methodology be transferrable to personnel at Everglades National Park. The current extent map, in conjunction with TerrSet and transition maps, supported the creation of the forecasted models.

Previous story | Main Page | Next story

7 Comments

Amber Jones 22-08-2016, 13:53

Great job! Demonstrating Google Earth Engine cloud masking & classifying capabilities makes it much easier for NASA data users!
I also appreciated the long-term on going solution for mangrove extent mapping. How will that tool work for the end-users?

Reply
Amber Jones 22-08-2016, 13:52

Great job! Demonstrating Google Earth Engine cloud masking & classifying capabilities makes it much easier for NASA data users!
I also appreciated the long-term on going solution for mangrove extent mapping. Now that the sample areas have shown an acceptable level of accuracy, is this project continuing over the extent of the Everglades?

Reply
amber jones 22-08-2016, 13:54

Sorry about that! I couldn’t edit or delete this one. When I replayed it, I heard there was a second term project scheduled. Disregard.

Reply
Daryl Ann Winstead (Mekong River Basin Agriculture) 18-08-2016, 17:11

Very interesting project and great video! Why did the project choose the time periods of 1995, 2005, and 2015? Thanks in advance for your response!

Reply
Emily Gotschalk 31-08-2016, 15:00

Thanks Daryl Ann! Sorry for the late response! We chose 1995, 2005, and 2015 as our sample years because we wanted to see trends over a decadal scale. Also, we hope to compare our maps with a mapping project that was conducted in 1994-1995 so we wanted to start within the same year for that comparison.

Reply
Katie Moore 12-08-2016, 14:29

That cloud-removal feature is useful. fmask, right? What is it filling in the missing pixels with? An average of pixels from other imagery for that time of year, or an average of its neighbors in the same image? etc.

Reply
Emily Gotschalk 17-08-2016, 14:43

Hey Katie! Good question! After utilizing the fmask tool on each of the Landsat scenes from the three sample years, we aggregated the entire year’s worth of scenes in order to have a complete cloudless image to represent each year without any holes left by the fmask. A reducer was then applied to the aggregated image to make sure that there were no clouds in any of the remaining pixels that were not already removed by the fmask algorithm. Thus, each pixel in the final image should represent the least cloudy pixel from the whole Landsat collection for that sample year.

Reply

Leave a Reply