Location: Range Management ResearchTitle: Watching grass grow: Successes and limitations of image-based methods for monitoring grassland phenology
Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 8/15/2017
Publication Date: 8/21/2017
Citation: Browning, D.M. 2017. Watching grass grow: Successes and limitations of image-based methods for monitoring grassland phenology [abstract]. 12th International Congress of Ecology INTECOL 2017, August 21-25, 2017, Beijing, China.
Technical Abstract: Seasonal changes in aboveground primary production (i.e. phenology) are influenced by environmental conditions with implications for land-atmosphere interactions, carbon cycling, and agricultural production. Monitoring phenology and quantifying seasonal patterns across spatially extensive grasslands and savannas require cost-effective, consistent, and accurate approaches for detecting changes in the growing season. Remotely sensed imagery offers capabilities to quantify growing season metrics via vegetation indices that have been linked to biomass and forage and net primary production. However, in many water-limited ecosystems contributions of exposed soil due to modest vegetation cover and susceptibility of vegetation to drought challenge these biophysical retrievals using moderate- and coarse-resolution satellite imagery. This challenge emphasizes the importance of verifying seasonal metrics derived from remotely sensed imagery (e.g., start, end, and length of growing season and timing of peak greenness) to identify limitations and uncertainty of said metrics used to forecast plant responses to climate and estimate biomass accrual and crop yields. We collected and evaluated data at multiple spatial and temporal scales from 2012 through 2016 to quantify the degree of correspondence between field and remotely sensed growing season metrics. This study was conducted at an ungrazed desert grassland site in southern New Mexico. Weekly field observations of plant phenology and canopy greenness were compared with metrics derived from MODIS satellite imagery and near-surface digital cameras (hereafter “phenocams”). MODIS 250-m NDVI vegetation index data are acquired every 16 days while color images are acquired daily from phenocams mounted at a height of ca 6 meters providing data at finer spatial and temporal scales that those from satellite. In addition, phenocams offer the opportunity to identify individual plants in the image to extract species-specific patterns in greenness. Growing season metrics from both types of imagery were derived using Timesat software (v3.2). Phenocam greenness curves for individual species indicated rapid canopy development for the widespread deciduous mesquite shrub (Prosopis glandulosa) over 14 to 16 d in the spring (April to May) whereas peaks in greenness for the dominant perennial black grama grass (Bouteloua eriopoda) occur with warm season rainfall in July. Weekly field estimates of canopy development were significantly correlated with greenness index values from phenocams for mesquite (r = 0.735) and black grama (r = 0.609). Start and end dates for growing season derived from MODIS NDVI corresponded well with phenocam greenness representing the landscape (R2 = 0.691) with better fit for end of season. These data demonstrate the utility of phenocams for reliably depicting species-specific patterns in phenology in this arid grassland although uncertainty with start of season for black grama was less than that for mesquite. The ability to distinguish contributions to landscape greenness between C3 shrubs and C4 grasses can dramatically enhance landscape monitoring efforts over spatially extensive grasslands and savannas where field monitoring is not feasible large areas. Future research will include additional grassland sites using image data from the Phenocam Network and online analytical tools that are in development to bolster opportunities to bridge satellite, camera, and field data to quantify and predict ecosystem productivity.