Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Wetland Landscape Spatio-Temporal Degradation Dynamics Using the New Google Earth Engine Cloud-Based Platform: Opportunities for Non-Specialists in Remote Sensing

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  Transactions of the ASABE. 59(5): 1331-1342. (doi: 10.13031/trans.59.11608) @2016
Authors:   Alice Alonso, Rafael Muñoz-Carpena, Robert E. Kennedy, Carolina Murcia
Keywords:   Data, Google Earth Engine, Mapping, NDVI, Remote sensing, Time series, Wetland.

Abstract. The complex nature of coupled human-natural systems often hinders the identification of forces and mechanisms causing observed environmental changes. The analysis of long-term time series can allow better understanding of those interactions and hence inform more adapted restoration and management programs. However, long-term time series of ground-measured vegetation variables are often not readily available due to the tediousness of the work and the financial and time investment required, especially for large-scale wetland systems. Remote sensing can help overcome this issue by providing more than 40 years of Earth cover images, but until recently the processing and analysis of these images was restricted to experts in remote sensing. The new Google Earth Engine (GEE) opens the remote sensing information mine to engineers or scientists without advanced knowledge in the field, including easy access to petabytes of publicly available remote sensing data and their spatial analysis in the cloud. To illustrate its capabilities, we used GEE to generate time series of the normalized difference vegetation index (NDVI) for the human-impacted Tempisque watershed with its severely degraded downstream Palo Verde wetland in northwest Costa Rica. We detail the processing and analysis steps to facilitate replication to any other case study. After defining the boundary of a study area, any user can generate a list and a video of the Landsat and MODIS images available for the area, NDVI maps, and a time series of the NDVI values spatially aggregated over the study area or a set of previously delineated polygons. One of the challenges we address is discriminating between the multitude of image collections available in the GEE catalog for vegetation mapping. We evaluate and compare the results obtained from five selected Landsat and MODIS image collections and identify the collections that give the best quality results for our case study. We conclude that MODIS is more appropriate for this tropical region because of the higher temporal resolution and hence higher probability to record cloud-free images. NDVI time series from Landsat images demonstrate a significant number of missing values. Landsat maps of NDVI suggest that the entire watershed and protected wetland witnessed an overall increase in vegetation greenness and hence cover since 1986, matching the abandonment of cattle ranching and the known degradation of the wetland by cattail invasion. Within-season vegetation variability can be tracked using MODIS images despite their coarser spatial resolution, showing high variability linked to precipitation patterns in this region. This GEE application illustrates new opportunities for biosystems engineers and other scientists to integrate historical vegetation and land cover data into comprehensive datasets to understand human impacts on ecosystems, and provides the guidance to do so.

(Download PDF)    (Export to EndNotes)