Over 700 million hectares of tropical forest have been destroyed or degraded, and restoring forest cover to some of these areas could yield enormous benefits for carbon sequestration and biodiversity conservation. However, despite ambitious plans to reforest millions of hectares of degraded land, ecological restoration at landscape scales remains expensive and unpredictable. My research aims to quantify how and why restoration outcomes vary across landscapes. To accomplish this, I use spatial models to link large-scale patterns of forest cover change with the ecological processes that underlie these patterns. I strive for results with practical relevance for land management, for example, discriminating between sites where natural regeneration will be sufficient to restore tree cover in a reasonable time vs. sites where more expensive tree-planting will be required. Studying restoration has also provided me with novel opportunities to test fundamental ecological questions, such as how spatial patterns structure population dynamics and whether stochastic or deterministic events are more important for community assembly.
A theoretical framework for restoration ecology
Theoretical ecology has long focused on how spatial dynamics impact ecological outcomes, yet mathematical models that lead to general insights for restoration are rare. I am developing a theoretical framework to explain variable restoration outcomes in empirical studies. My recent work towards this goal includes a mathematical model that evaluates how reforestation rates depend on landscape features beyond the scale of most field studies. This model can be solved to understand how time to tree canopy closure, a key benchmark for forest restoration, depends on seed rain, an important demographic rate that is difficult to measure in the field. I have used the model to demonstrate how landscape features that lead to differences in seed rain can alter the successional trajectories of otherwise identical sites. In addition to general theoretical insights, the model provides a uniting framework for explaining patterns in real data across multiple spatial scales. You can explore the basic model using the sliders below:
Ecological field research on landscape-scale restoration
I am testing the theoretical framework for ecological restoration with field data from multiple sites. The overall objective is to understand how plant growth, survival and recruitment interact with landscape features to drive succession. One challenge for understanding how a grass-dominated patch becomes a secondary forest is measuring a rare event: tree seedling recruitment during early succession. In Los Tuxtlas, Mexico, in collaboration with researchers at the University of Morelos, I am studying tree demography in experimental reforestation plots, where >5000 individual tree seedlings have been marked and measured during the first eight years of succession. This work reveals how tree community dynamics during early succession set the course for secondary forest structure and species composition.
A related challenge is explaining why some founder plants successfully establish new populations while others fail. In South Carolina, I am working in restored fragments of longleaf pine habitat to ask whether landscape features can predict establishment success of reintroduced plant species. These field data are part of a landscape experiment that manipulates connectivity, fragment shape and distance to edge in replicate fragments that include transplants of native species. I am developing demographic models that use these experimental landscape features to predict when a transplanted individual will successfully establish a new population. A particularly rewarding aspect of my involvement with the experiments in Los Tuxtlas and South Carolina has been collaborating with the projects’ original principal investigators to design long-term monitoring schemes for plant demography.
While experiments can disentangle impacts of landscape features on restoration, I am also working with observational data to quantify natural regeneration rates. In southwestern Panama, in collaboration with the Smithsonian Tropical Research Institute, I am analyzing a unique dataset of secondary forest dynamics plots stratified across a range of environmental conditions. These observational data are powerful because the sampling design supports extrapolation to larger spatial scales. Fitting my theoretical model to these spatial data has enabled me to identify demographic barriers to reforestation at different sites—a first step towards spatially-targeted interventions for forest landscape restoration.
Remote sensing to extend demographic insights to landscape scales
Field data can provide insight into demographic rates that underlie reforestation, but plots where individual plants are tagged and monitored are limited to small scales. In contrast, satellite and airborne remote sensing can capture images over much larger areas, from regions to continents. I am using these remotely sensed data to quantify tree demography at landscape scales. I have recently developed a method to predict tree canopy cover and tree height from the open access, multi-decadal and globally available Landsat satellite archive. These predictions represent a significant advance over discrete “non-forest” and “forest” land cover categories that are difficult to relate to continuous changes in tree cover during secondary succession. More generally, the satellite-derived predictions provide spatially-explicit time series of forest structure with unprecedented spatial and temporal extent. I have also recently demonstrated that a single image of high resolution hyperspectral data can predict tree growth rates. This discovery suggests that aerial remote sensing data can detect tree canopy traits related to foliar nutrients, disease and overall tree performance. Ultimately, the ability to detect spatial differences in tree demographic rates with remotely sensed data could enable fine-scale spatial interventions (“precision forestry”) and contribute to sustainable decision-making in forested ecosystems.
Interdisciplinary research for sustainability science
I believe that to address environmental challenges, including large-scale restoration, incorporating human behavior and socioeconomic context is essential. As a Postdoctoral Fellow in the NSF SEES program, I am collaborating with social scientists to study how farmer decision-making influences reforestation rates. A key goal of this research is to directly link social and ecological dynamics, rather than study each separately. For example, to understand why forest regrowth is occurring on cattle pastures in southwestern Panama, I developed a board game (“Cattle Royale”) that represents a model for this coupled natural and human system. By playing this game with Panamanian farmers, I was able to solicit detailed feedback on my model from the stakeholders responsible for land use decisions.
Caughlin T.T., S.W. Rifai, S.J. Graves, G.P. Asner, S.A. Bohlman. Landsat-LiDAR integration reveals widespread forest regrowth in a tropical agricultural landscape. Remote Sensing in Ecology and Conservation 2:190-203.
Caughlin T.T., S.J. Graves, G.P. Asner, M. van Breugel, J.S. Hal, R.E. Martin, M.S. Ashton, S.A. Bohlman. A single hyperspectral aerial image can accurately predict growth rates of tropical tree species in single-species stands. Ecological Applications 26:2367-2373.
Caughlin T.T., S. Elliott, J.W. Lichstein. When does seed limitation matter for scaling up reforestation from patches to landscapes? Ecological Applications 26:2437-2448.