Figure captions:

Figure 1 . Examples of stress induced behavioural shifts. A) Mean values of dive duration and distance travelled of individuals of ringed seal (Pusa hispida ) in Svalbard (Norway) before and after a significant decline in sea-ice extension, and thus in their habitat quality. Both behaviours showed significant increases in response to the stress. Bars indicate standard error. Data adapted from Hamilton et al. (2015). B) Probability of aggressive encounters between heterospecific and conspecific butterflyfishes (Chetodon spp.) calculated in 4 regions across the central Indo-Pacific, before and after a mass coral bleaching event in 2006 (reduction in food resources for the fishes). Both behaviours showed significant decreases. Data adapted from Keith et al. (2018). Bars are 95% confidence intervals. *Coral cover % change showed refers only to one of the four sites (Christmas Island site, Indian Ocean; 105.6° E, 10.4° S) for simplicity.

Figure 2 . Examples of stress induced shifts in morphological traits. A) Mean estimated body mass of female polar bear individuals (Ursus maritimus ) in western Hudson Bay from 1980 through 2007 (dashed line indicates fit of linear regression [r=-0.549, p<0.01], bars indicate standard error); the area experienced progressively earlier dates of sea-ice breakup, that is a decline in duration of the favorite sea-ice habitat of polar bears. Data adapted from Stirling & Derocher 2012. B) Comparison of the mean of individual asymmetry index for femoral pores among Italian wall lizard (Podarcis siculus ) sampled in hazelnut orchards with no history of pesticide use (control) vs orchards regularly treated with pesticides (treatment). The pollution stress induced an increase in the fluctuating asymmetry for femoral pores (bars indicate standard error). Data adapted from Simbula et al. 2021.

Figure 3. Theoretical example of a timeline to collapse. Here we posit a population of seabirds inhabiting an area where prey resources (e.g. fish stocks) begin a continuous decline (A). The curves in B and C represent average values of a behavioural (B) and morphological trait (C) calculated from a pool of individuals in the population through time. The red curve in D shows the abundance of the population. First a shift is observed in the behaviour (time point T_Bs), where the average foraging distance increases compared to the average measured during stable conditions B_s (B). The foraging distance will increase until it reaches a physiological limit (time point T_Be), defining the time interval where a continuous change is observable (I_B). After, or during such time, we will observe a decrease in average body size compared to that measured during stable conditions M_s (C), at time T_Ms. The body size will change until its physiological limit (T_Me), defining the time interval where such continuous change is observable (I_M). Later, the abundance trend of population will show alterations in the pre-decline indicators such as Early Warning Signals (EWSs), that will start to be observable at time point T_As, and will last until T_Ae, defining the time interval I_A. Subsequently, the continuous decreases to extinction (D) will begin at time point T_Es, and end will end with the extinction of the population T_Ee, lasting the time interval I_E. The first occurrence of the signals projected on the lower Time axes shows the sequence in the category of observable signals of stress starting at the individuals’ level (B, C) and propagating to the population level (D).

Figure 4. Theoretical example of a timeline to collapse. The grey curve represents a continuous growth of a given environmental stress on the population. The blue and green curves represent average values of a behavioural and morphological trait calculated from a pool of individuals in the population. The red curve represents the number of individuals. B_s and M_s are respectively the average measure of the considered behavioural and morphological traits in stable conditions. The small black dotted lines project the starting point of the shifts in morphological traits and abundance dynamics on the behaviour (point 1) and morphological trait (point 2) curves. Projected on the vertical axis, those points identify B_x and M_x: the values of behavioural and morphological metrics at the time of the onset of the next signal along the timeline. The interval of change (brackets) from the average values defines the intrinsic buffering capacities of behaviour (C_b,) and morphological traits (C_m).

Figures.