Monitoring ecosystem services with Essential Ecosystem Service Variables (Schwantes, Firkowski et al., in review)
Monitoring ecosystem services is required to assess the changing state of human-nature interactions. To standardize the monitoring of multiple facets of ecosystem services, the Group on Earth Observations Biodiversity Observation Network (GEO BON) recently proposed the Essential Ecosystem Service Variables (EESVs), which are organized into six classes: Ecological Supply, Use, Anthropogenic Contribution, Demand, Instrumental Value, and Relational Value. We discuss challenges and opportunities for applying the EESV framework and demonstrate its potential for linking data to monitoring and decision support. Furthermore, we show how the EESVs could encourage comparisons and interconnections among ecosystem service indicators adopted by the monitoring frameworks of different policy agendas and conventions.
Bayesian Belief Networks reveal how an underlying driver shapes interactions among ecosystem services (Schwantes et al., in prep)
We use Bayesian Belief Networks (BBNs) to determine if there is a true interaction between ecosystem services or if ecosystem services are responding to the same underlying shared driver. By including ecosystem services and drivers of change in Bayesian Belief Networks, we can better understand relationships between ecosystem services (e.g., trade-offs and synergies) and their drivers. We find key differences between BBNs with and without the driver. Not accounting for the driver within the BBN results in the detection of spurious trade-offs, failure to detect trade-offs and synergies, and overestimation of trade-offs and synergies.
A comparison of approaches to quantify carbon for ecosystem service assessments through time (Schwantes et al., submitted)
We review the strengths, limitations, and best practices of assessing carbon storage & sequestration over time for four approaches:
land cover maps with reference carbon values by cover type,
statistical & machine learning models linking field measurements to remotely sensed data, and
mass-balance models representing key carbon pools and flows between them
Monitoring supply, demand, and use for the service of air quality regulation in Canadian cities (New project)
We are estimating air quality regulation, an ecosystem service provided by urban forests, over time in Canadian cities. Monitoring multiple facets of a service (e.g., ecological supply, demand, use) is needed to support decision-making and to assess whether the demand for clean air is being met. The demand for air quality regulation will likely increase as wildfires become more common and as cities continue to grow, requiring policies that support both pollution reduction and urban forest expansion.
Previous Research Projects
Forecasting species & community shifts with climate change Tool development project
R Shiny App to share research findings: a part of PBGJAM
We forecasted species reorganization under climate change using generalized joint attribute models, GJAM. We incorporated abundance estimates across taxa, including birds, small mammals, beetles, ticks, and vascular plants, from the National Ecological Observatory Network (NEON), as well as remotely sensed data including satellite-based imagery and LiDAR for improved habitat and vegetation structure characterization. I created an R shiny app to visualize model outputs: PBGJAM and I also developed a python package (geedataextract) using Google Earth Engine’s python API, which allows for efficient pre-processing and spatial/temporal averaging of remotely sensed imagery, climate, elevation, and soil datasets.
Monitoring drought-induced tree mortality In 2011, Texas experienced the most severe one-year drought since record keeping began, killing millions of trees across the state. We developed two remote sensing approaches to quantify canopy loss from the 2011 drought. Firstly, in central Texas, we conducted a time-series analysis to identify areas with persistent canopy loss, using Landtrendr. Within these areas, we estimated percent tree canopy loss using a zero-or-one inflated beta regression model (Schwantes, et al., RSE, 2016). Secondly, to create canopy loss maps across the entire state of Texas, we used Random Forest, a machine learning algorithm, to identify drought-impacted areas with greater than 25% canopy loss. We scaled from field estimates to 1-m orthophotos to 30-m Landsat imagery (Schwantes et al., GCB, 2017). Time-series, geospatial, machine-learning, and statistical analysis were all conducted using ArcGIS, Python, R, ENVI, and IDL.
Identifying climatic thresholds associated with canopy loss and forecasting how likely these thresholds would be crossed in the future Focusing on a 100-km by 1,000-km transect spanning the State's five-fold east–west precipitation gradient (1,500 to 300 mm), we compared spatially explicit 2011 climatic anomalies to our canopy loss maps. Much of the canopy loss occurred in areas that passed specific climatic thresholds in temperature and vapor pressure deficit anomalies, percent deviation in precipitation, and 2011 difference between precipitation and potential evapotranspiration. Future climate data under the representative concentration pathway 8.5 trajectory project that average values will surpass the 2011 VPD anomaly during the 2070–2099 period and the temperature anomaly during the 2040–2099 period (Schwantes et al., GCB, 2017).
Accounting for landscape heterogeneity to improve spatial predictions of tree vulnerability following drought We employed a non-linear stochastic model of soil moisture dynamics across a watershed, accounting for spatial differences in aspect, topography, and soils. Using MATLAB, we modeled water stress for Juniperus ashei, a dominant tree/shrub that experienced significant mortality during the record 2011 drought. We then projected future dynamic water stress through the 21st century using future climate projections from 10 global climate models under two scenarios, and comparing models with and without landscape diversity. Favorable microsites or refugia may exist across a landscape where trees can persist; however, if future droughts are too severe, the buffering capacity of a heterogenous landscape could be overwhelmed. When projecting future water stress, we can account for these potential refugia by incorporating landscape diversity into models (Schwantes, et al. New Phytologist, 2018).