Salt Cedar Encroachment along the Lower Gila River
Using high resolution image data and LiDAR to classify vegetation on BLM lands along the Lower Gila River.
Objectives
- Goal: Identify location of species along the river to inform BLM management activities.Improve remediation efforts in tamarisk-invaded riparian ecosystems.
- Map current species composition and distribution.
- Inform resource needs to impact restoration and future control efforts.
Study Area
- Study area encompasses 26 miles along the Gila River.
- Located east of Yuma between Dateland and Sentinel, Arizona.
The stretch of the Lower Gila River we examine in this study is located in southwest Arizona just east of Yuma along Interstate 8.
Data
- BLM acquired LiDAR on March 3, 2016.
- 15cm 4-band orthophotos captured on May 22, 2016.
- Drone flights and field work conducted September 30, 2016, April 10, 2017, February 9, 2018 and May 8, 2018.
Workflow
This diagram shows the inputs, methods and outputs of our workflow to identify vegetation species along the Lower Gila River.
- Using the 15cm 4-band orthophotos and NDVI a CART classification was performed to identify vegetation, litter and ground.
- Removed litter and ground pixels to create segments of vegetation.
When we removed the pixels classified as litter or ground only vegetation pixels remained. We grouped coinciding vegetation pixels to create vegetation segments.
- Geotagged Photos converted into points for validation and training purposes.
- Identified Salt Cedar, Mesquite and Creosote segments.
- Some segments included groups of Salt Cedar or mixtures of Salt Cedar and Mesquite.
We used Geotagged photos and Drone imagery to identify the species of segments within the area of interest.
Example of species types identified via Drone imagery.
- Various statistics were calculated for each vegetation segment including:
This table represents the statistics used to classify the species of each vegetation segment. Those statistics highlighted in yellow were utilized by the CART algorithm.
- Due to the inability to individually separate each canopy we created group classes.
- Joined Salt Cedar = Group Stands of only Salt Cedar.
- Mixed = Stands of Salt Cedar and Mesquite.
- Individual Classes = Salt Cedar Type 1, Salt Cedar Type 2 Mesquite and Creosote.
We were able to identify five different vegetation classes based on segment composition and field data.
Results of CART classification on segmented vegetation. Each segment is identified by their dominant species.
- •Current overall accuracy is 81%
•If solo Salt Cedar classes are combined overall accuracy is 86%
•If all Salt Cedar classes are combined overall accuracy is 94%
The segmented CART classification achieved an overall accuracy of 81%.
Possible Die-off?
- Comparison of NAIP data from 2007 and 2016 seem to indicate a decline in live vegetation.
Areas in red indicate loss of vegetation darker shades indicate greater loss. Areas in green indicate increased vegetation with darker shades indicating greater increases.
2007 false color 1m NAIP image pictured on the left. 2016 False color 15cm orthoimage pictured on the right. Clearly, the vegetation is much denser in the 2007 scene then in the 2016 scene.
MODIS NDVI data seems to back up the premise of a decline in live vegetation along the Lower Gila River.
Conclusions
- Salt Cedar dominates the Lower Gila River.
- Current methods can identify areas of living vegetation and litter.
- Hyperspectral data would do a better job of discerning species types.
- Limited number of Mesquites in the area make them hard to classify.
- Weak Salt Cedars are confused with Creosote.
