1. Research into global change ecology is motivated by the need to understand the role of humans in changing biotic systems. Mechanistic understanding of ecological responses requires the separation of different climatic parameters and processes that often operate on diverse spatiotemporal scales. Yet most environmental studies do not distinguish the effects of internal climate variability from those caused by external, natural (e.g. volcanic, solar, orbital) or anthropogenic (e.g. greenhouse gases, ozone, aerosols, land-use) forcing factors.
2. We suggest extending the climatological concept of ‘Detection and Attribution’ (DA) to unravel abiotic drivers of ecological dynamics in the Anthropocene. We therefore apply DA to quantify the relative roles of natural versus industrial temperature change on elevational shifts in the outbreak epicentres of the larch budmoth (LBM; Zeiraphera diniana or griseana Gn.); the classic example of a cyclic forest defoliating insect.
3. Our case study shows that anthropogenic warming shifts the epicentre of travelling LBM waves upward, which disrupts the intensity of population outbreaks that occurred regularly over the past millennium in the European Alps. Our findings demonstrate the ability of DA to detect ecological responses beyond internal system variability, to attribute them to specific external climate forcing factors, and to identify climate-induced ecological tipping points.
4. In order to implement the climatological concept of ‘Detection and Attribution’ successfully into modern global change ecology, future studies should combine high-resolution paleoenvironmental reconstructions and state-of-the-art climate model simulations to inform inference-based ecosystem models.

To extend the above study beyond ‘Detection’ to ‘Attribution’, we assess the elevational dynamics of LBM outbreaks on simulated Alpine winter temperature fluctuations from ‘The Community Earth System Model-Last Millennium Ensemble’ (CESM-LME; Otto-Bliesner et al., 2016). Using ~2° resolution in the atmosphere and land components, CESM-LME expands on the CMIP5 and earlier LM model simulations by providing the largest ensemble of last millennium simulations with a single model to date. The CESM-LME uses the CMIP5 climate forcing reconstructions (Schmidt et al., 2011) and contains both ‘full forcing’ simulations containing all last millennium forcing agents, as well as ensembles of simulations with each forcing individually (greenhouse gases, ozone and aerosols, land-use, volcanic eruptions, solar variability, orbital changes). In using the most recent version of the comprehensive CESM (Hurrell et al., 2013), the herein employed state-of-the-art multi-model climate ensemble approach adds understanding into the role of internal variability versus external forcing in generating climate variations over the last 12 centuries (Otto-Bliesner et al., 2016). Annual winter temperatures of the monthly means from December–February and averaged from 12 grid points over the Alpine arc (6–16°E and 44–48°N) were extracted from control and forced model runs. With the exception of combined ozone and aerosol simulations (1851/52 to 2004/5 CE), all other model output reaches back to the winter 851/52 CE. The climate simulations were coupled with a spatial model of LBM population dynamics (Johnson et al., 2010), in which population growth rates were a function of winter temperatures.

The effects of anthropogenic climatic forcing on LBM population dynamics were explored by a discrete time tri-trophic model where the insect’s pre-dispersal population dynamics are coupled with pre-dispersal population dynamics of parasitoids and the quality of the larch host tree foliage. LBM population dynamics are most consistent with regulation by both parasitoids and host needle quality (Baltensweiler & Rubli, 1999). Parameter values were empirically-derived and the model was selected based on the best fit to empirical population dynamics of LBM (Turchin et al., 2003). The model was spatially-extended by linking multiple populations with moth and parasitoid dispersal using exponential dispersal kernels. This model has been used to capture the recurrent west-east traveling waves of LBM outbreaks across the Alpine arc in the mid- to late-20^th century (Johnson et al., 2004). The method was used later to demonstrate that the travelling LBM outbreak waves vary along an elevational gradient, and that the elevation of the outbreak epicentre is affected by winter temperatures (Johnson et al., 2010). Previous studies have shown that elevation has a unimodal impact on the growth rates of LBM populations (Baltensweiler & Rubli, 1999). The spatially-extended population model suggests that outbreak epicentres occur at or near elevations where the LBM growth rate is greatest (Johnson et al., 2010).

In this study, we use the relationship between growth rate and elevation estimated by Johnson et al. (2010), and link the temperature component of the population model to the simulated temperature anomalies. From the simulations we calculate epicentre elevations using wavelet phase analysis (Grenfell et al., 2001). Phase angle values range from -180° to 180°. As a population cycles from a trough to a peak, the phase angle increases from -180° to 0°. As a population cycles from a peak to a trough, the phase angle continues to increase from 0° to 180° (Johnson et al., 2004). In this study, the elevation with the largest phase angle is identified as the location of the traveling wave epicentre. The magnitudes of LBM population cycles are estimated by identifying the maximum population density over 10-year intervals, which approximates the typical 8–9-year length of LBM cycles (Turchin et al., 2003). The population dynamics are simulated across a potential elevational range from 400–2200 m asl. The one-dimensional gradient is separated into 181 locations at 10 m intervals. Dispersal across elevation is modelled with isotropic movement. The effects of different climatic forcing agents on the elevation of epicentre and magnitude of cycles is explored by simulating LBM population dynamics under each scenario. In each simulation, LBM population dynamics are coupled with winter temperature anomalies from the respective forcing scenario. Epicentre elevations are estimated from the simulated LBM population dynamics using wavelet and phase angle analysis (Johnson et al., 2004). Model performance is assessed by comparing the likelihoods that empirical phase angle-elevation relationships (Johnson et al., 2010) would be produced by the population model under each of the various climate forcing scenarios. All analyses are performed in Matlab R2012b (see Johnson et al., 2004, 2010 for methodological details). Supplementary to the above, we refer to Otto-Bliesner et al. (2016) and Johnson et al. (2004, 2010) for gaining further insights into ‘The Community Earth System Model-Last Millennium Ensemble’ (CESM-LME) modelling project and the ecological LBM model, respectively.

See original publication for details.