Location: Southwest Watershed Research CenterTitle: Considerations for achieving cross-platform point cloud data fusion across different dryland ecosystem structural states
|SWETNAM, T.L. - University Of Arizona|
|GILLAN, J. - University Of Arizona|
|SANKEY, T.T. - Northern Arizona University|
|MCCLARAN, M.P. - University Of Arizona|
|Heilman, Philip - Phil|
|MCVAY, J. - Northern Arizona University|
Submitted to: Frontiers in Plant Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/4/2017
Publication Date: 1/10/2018
Citation: Swetnam, T., Gillan, J., Sankey, T., McClaran, M., Nichols, M.H., Heilman, P., McVay, J. 2018. Considerations for achieving cross-platform point cloud data fusion across different dryland ecosystem structural states. Frontiers in Plant Science. 8:2144. https://doi.org/10.3389/fpls.2017.02144.
Interpretive Summary: Vegetation responds rapidly to seasonal precipitation in dryland ecosystems. New technologies and methods for collecting very high resolution data are being tested to quantify both rapid vegetation growth and disturbances such as fire and herbivory. Three dimensional data were collected in 2015 on grassland and shrubland sites within the Walnut Gulch Experimental Watershed east of Tombstone, Arizona using a small unmanned aerial system, a manned aircraft, and a ground-based terrestrial lidar scanner. In 2016, similar data were collected on the Santa Rita Experimental Range south of Tucson, Arizona at savannah and woodland sites. Accurate bare earth elevation models are critical for making measurements of plant structure. In the grassland, shrubland, woodland and savannah sites we found aerial lidar methods to be more accurate at sensing bare ground beneath vegetation than either terrestrial lidar or unmanned aerial systems, but less precise in sensing the structure of herbaceous and woody plants. However, the cost and scheduling of manned aircraft capable of lidar may preclude high frequency data collection.
Technical Abstract: Dryland ecosystems undergo long periods of senescence punctuated by rapid growth following seasonal precipitation events. Remote sensing of vegetation dynamics which capture new growth as well as herbivory and disturbance require both high spatial and temporal resolution data acquired by various optical sensors. Challenges to collecting useful information emerge across sensors and platforms at different scales in space and time. In the present study, we identify strengths and weaknesses for several modalities of terrestrial and aerial remote sensing, including small unmanned aerial systems (sUAS) and manned aircraft. Following the North American monsoon in late summer 2015 we collected (1) structure from motion multi-view stereo (sfm-mvs) imagery and (2) lidar from two sUAS, (3) terrestrial lidar, and (4) manned aircraft lidar over a grassland and a shrub site on the Walnut Gulch Experimental Watershed east of Tombstone, Arizona (30.74° N, -110.05° W). In spring and summer 2016 we collected sUAS sfm-mvs imagery and terrestrial lidar in mesquite woodland and savannah sites on the Santa Rita Experimental Range south of Tucson, Arizona (31.80° N, -110.84° W), where manned aerial lidar was available from 2011. In the grassland, shrubland, woodland and savannah sites we found aerial lidar (both sUAS and manned aircraft) to be more accurate at sensing bare ground beneath vegetation than either terrestrial lidar or sUAS sfm-mvs, but less precise in sensing the structure of herbaceous and woody plants. The cost and scheduling of any aircraft capable of lidar precluded higher frequency data collection, whereas the sfm-mvs by sUAS was not limiting. Despite the utility of sfm-mvs for monitoring vegetation phenology and structure at sub-centimeter resolution (exceeding 35,000 points per meter square), sfm-mvs was a poor identifier of bare earth beneath herbaceous plants during the peak growth period. Sfm-mvs bare earth models were satisfactory during the leaf-off spring period coinciding with minimum herbaceous cover. Computational costs of sfm-mvs became a limiting factor at progressively larger spatial and temporal extents. We recommend future dryland ecosystem remote sensing studies ensure the most accurate bare earth elevation model possible prior to making plant structural measurements.