Past, Present, and Future Climate Change: Can Palaeoecology Help Manage a Warming World? - Biodiversity Conservation and Environmental Change - page 100

dispersal rates, and furthermore, some species may be able to adapt their climate tolerances,
thus reducing or eliminating the need to move to track suitable climate space (Skelly
et al. 2007).
Testing model predictions using hindcasting
Palaeoecological records have demonstrated distributional responses to past climate change,
providing essential information on sensitivity and migration rates that could potentially be
used in planning of reserve configurations, migration corridors, and protected area networks
(Bush 2002, Bush and Lovejoy 2007). Using palaeoecological, ecological and experimental
data alongside modelling and simulation, much progress has been made in estimating the
relationships between climate and distribution, in reconstructing climate refugia and migra-
tion pathways, and the genetic and evolutionary consequences of distribution change (Rob-
erts and Hamann 2011, Svenning et al. 2011, Hampe et al. 2013, Gavin et al. 2014).
Predicting future changes in distribution involves first simulating current distribution,
based on present climate data, then modelling distribution changes based on future climate
scenarios. Such species distribution models (SDMs) simulate current species distribution
statistically, based on correlations with current climatic conditions (niche-based or climate
envelope models) or mechanistically, by modelling the relationship between physiological
tolerance, climate parameters, and spatial distribution (process-based models) (Pereira et al.
2010, Roberts and Hamann 2011, Svenning et al. 2011). Some distribution models work at the
level of the biome rather than the species level. Dynamic vegetation models (DVMs) are com-
plex biome-level models that integrate processes such as photosynthesis, respiration, com-
petition, nutrient cycles, and fire, enabling macro-level changes like the expansion and
contraction of forests to be modelled (Pereira et al. 2010, Prentice et al. 2011). There are vari-
ous uncertainties associated with all of these modelling approaches; for example, climate
envelope models rest on the assumption that species distribution is in equilibrium with cli-
mate and unaffected by other factors like disturbance, predation, and competition, whereas
this may not be the case (Holt and Barfield 2009, Parmesan et al. 2011). Mechanistic models
require detailed physiological data that is often estimated because experimental and obser-
vational work is still in progress.
Confidence in predicting future distribution based on scenarios of climate change in the
coming decades is much increased if models can successfully simulate past distribution
changes in previous warm periods. Using palaeoclimatic records, SDMs and DVMs can be
used to 'hindcast' distribution changes that are already known from the palaeo-record, thus
testing model validity and increasing confidence in their ability to predict future distribution
changes (Nogués‐Bravo 2009). For example, Roberts and Hamman (2012) assessed the real-
ism of bioclimatic envelope models for 14 biomes in western North America, using palaeo-
ecological data from over 1,400 sites. The models could effectively simulate known ecosystem
distributions from about 6,000 years ago, and the authors were confident of the models pre-
dictions for distribution changes over the coming century. The models became increasingly
inaccurate further back in time, attributed to the no-analogue climates of the Last Glacial