Novel use of image time series to distinguish dryland vegetation responses to wet and dry years

Authors: Myers, Emily; Browning, Dawn; Burkett, Laura; James, Darren; Bestelmeyer, Brandon

Submitted to: Journal of Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/21/2024
Publication Date: 7/8/2024
Citation: Myers, E.B., Browning, D.M., Burkett, L.M., James, D.K., Bestelmeyer, B.T. 2024. Novel use of image time series to distinguish dryland vegetation responses to wet and dry years. Journal of Remote Sensing. 1. Article 0190. https://doi.org/10.34133/remotesensing.0190.
DOI: https://doi.org/10.34133/remotesensing.0190

Interpretive Summary: Producers, land managers, and indigenous communities are confronted with drought, shifts in the timing of rainfall and high temperatures that challenge our ability to produce food and fiber and threaten ecosystem services and livelihoods nationwide. Mitigating the harmful effects of climate change requires land surface monitoring, a prospect that is especially difficult in arid rangelands where landscapes are extensive, remote, and in some cases with modest vegetation cover. We evaluated the potential for time series images from near-surface cameras (PhenoCams) to complement data from the Harmonized Landsat-Sentinel image time series using Ecological Site concepts widely-used by USDA NRCS and Bureau of Land Management (BLM). We found that both HLS and PhenoCam greenness responses (or phenology) were useful for capturing the response of herbaceous vegetation cover in wet years. We found that greenness responses were more difficult to capture during dry years, but that contrasts between PhenoCam and HLS-measured phenology could still be useful in distinguishing between different grass- and shrub-dominated  states. PhenoCams could be an inexpensive tool to expand coverage for landscapes of concern and signal response from drought in ways not easily discerned via satellite imagery. Understanding the strengths and limitations of complementary instrument designs can help land managers, producers who practice precision agriculture, as well as scientists who formulate and test models to improve forecasts based on climate change scenarios.

Technical Abstract: Remote sensing methods are commonly used to assess and monitor ecosystem conditions in drylands, but accurate classification and detection of ecological state change are challenging due to sparse vegetation cover, high spatial heterogeneity, and high interannual variability in production. We evaluated whether phenological metrics are effective for distinguishing dryland ecological states using imagery from near-surface camera (PhenoCam) and satellite (Harmonized Landsat 8 and Sentinel-2, hereafter HLS) sources, and how effectiveness varied across wet and dry rainfall years. We analyzed time series over 92 site-years at a site in southern New Mexico undergoing transitions from grassland to shrubland on different soil types. Rainfall was a driver of phenological response across all ecological states, with wet years correlating with later start of season, later peak, higher peak greenness, and shorter growing season. This rainfall response was strongest in shrub-invaded grasslands on sandy soils. PhenoCam estimated significantly earlier start of season than HLS for shrublands on gravelly soils and earlier end of season than HLS for shrub-invaded grasslands on sandy soils. We propose integrating seasonal metrics from high-frequency PhenoCam time series with satellite assessments to improve monitoring efforts in drylands, use phenological differences across variable rainfall years to measure differences in ecosystem function among states, and use the timing and strength of peak greenness of key plant functional groups (grasses in our study site) as an indicator of ecological state change.