North American vegetation model data for land-use planning in a changing climate:
Dates
Publication Date
2012-01-01
Start Date
2086
End Date
2095
Citation
Gerald E. Rehfeldt, Nicholas L. Crookston, Cuauhtémoc Sáenz-Romero, and Elizabeth M. Campbell, 20120101, North American vegetation model data for land-use planning in a changing climate:: , http://charcoal.cnre.vt.edu/climate/.
Summary
Data points intensively sampling 46 North American biomes were used to predict the geographic distribution of biomes from climate variables using the Random Forests classification tree. Techniques were incorporated to accommodate a large number of classes and to predict the future occurrence of climates beyond the contemporary climatic range of the biomes. Errors of prediction from the statistical model averaged 3.7%, but for individual biomes, ranged from 0% to 21.5%. In validating the ability of the model to identify climates without analogs, 78% of 1528 locations outside North America and 81% of land area of the Caribbean Islands were predicted to have no analogs among the 46 biomes. Biome climates were projected into the future [...]
Summary
Data points intensively sampling 46 North American biomes were used to predict the geographic distribution of biomes from climate variables using the Random Forests classification tree. Techniques were incorporated to accommodate a large number of classes and to predict the future occurrence of climates beyond the contemporary climatic range of the biomes. Errors of prediction from the statistical model averaged 3.7%, but for individual biomes, ranged from 0% to 21.5%. In validating the ability of the model to identify climates without analogs, 78% of 1528 locations outside North America and 81% of land area of the Caribbean Islands were predicted to have no analogs among the 46 biomes. Biome climates were projected into the future according to low and high greenhouse gas emission scenarios of three General Circulation Models (cgcm, hadley, and consensus) for three periods, the decades surrounding 2030, 2060, and 2090. Prominent in the projections were (1) expansion of climates suitable for the tropical dry deciduous forests of Mexico, (2) expansion of climates typifying desertscrub biomes of western USA and northern Mexico, (3) stability of climates typifying the evergreen–deciduous forests of eastern USA, and (4) northward expansion of climates suited to temperate forests, Great Plains grasslands, and montane forests to the detriment of taiga and tundra climates. Maps indicating either poor agreement among projections or climates without contemporary analogs identify geographic areas where land management programs would be most equivocal. Concentrating efforts and resources where projections are more certain can assure land managers a greater likelihood of success.
Because General Circulation Models (GCM) are key to understanding future climates, the tool we envision must take into account the variability in GCM output resulting from different model formulations and emissions scenarios. However, to land managers, variation represents uncertainty, and uncertainty often leads to inaction. Our approach was to emphasize similarities in responses projected from the disparate formulations rather than dwell on the differences. The tool must also contain provisions for identifying future climates with no contemporary analog because new climates may best be suited for species assemblages that do not exist today. To be useful, the scope of a vegetation model should be neither too broad to constrain local interpretation nor so parochial as to limit management options. It is mandatory, moreover, that this tool incorporates powerful statistical techniques that retain the robustness and flexibility necessary for anticipating and accommodating actual changes.
Our analysis uses the biotic communities of Brown (1994), mapped and digitized by Brown et al. (1998). This classification system meshes well with our goals: It is based on distributions of flora and fauna without reliance on physiography, the coverage includes all of North America, and altitudinal zonation of vegetation is an integral part of the system. For simplicity, we use the term "biome" to reference the biotic communities. Previous analyses used this classification system for modeling climatic control of biome distributions in western USA. These analyses produced errors of fit (~10%) that largely resulted from an imperfect alignment between digitized polygons in the classification system and the digital elevation model used to generate point estimates of biome climates. Misalignment had greatest impact at borders of polygons, along shore lines, and on mountain peaks, and therefore, was of greatest source of error with small, irregularly shaped polygons and for biomes occurring in altitudinal sequence. In this analysis, we considered biomes from throughout North America and used supplemental information to alleviate adverse effects of the misaligned data files.
We built on the statistical modeling of Rehfeldt et al. (2006) to predict contemporary realized climatic niches of North American biomes. Ecological niche models have received considerable criticism mostly because of their inability to represent migration or colonization potentials, competitive interactions, and effects of enhanced CO2 on water-use efficiency. The first of these is not an issue. Niche modelers themselves make no claims of projecting species distributions, but instead emphasize that their models project suitable climates. Likewise, competitive interactions that limit species distributions can be defined climatically and are invalidated only when future climates have no contemporary analog. By predicting and projecting the occurrence of novel climates, statistical models can negate this second point of contention. However, it is true that correlative models are not yet capable of accounting for physiological impacts of enhanced CO2 on gas exchange and productivity. Such effects, however, are well characterized in only a few species, and impacts on species distributions are unknown. More research is required before effects of enhanced CO2 can be incorporated with credibility into vegetation models, whether correlative or mechanistic.
Land managers require decision-support tools suitable for dealing with oncoming climate-mediated ecosystem changes. Progress has been made in converting climatically static vegetation simulators to climatically dynamic models, and guidelines are in use for managing future generations of the broadly dispersed Larix occidentalis of western North America and the narrow endemics, Picea chihuahuensis, P. mexicana, and P. martinezii of Mexico. Yet, for much of North America, comprehensive management guidelines do not exist. Our goal was to develop a statistically valid, climate-driven vegetation model suitable for land-use planning during a changing climate.