Figure 1.
Map of the study area along the north Norwegian coast 65°–68°N, within the northeast Atlantic, coded for sampling year. Large points (n = 1220) are model data for the statistical analyses, whereas small yellow points (n = 403) are test data for evaluation of the models.
Figure 2.
The partial response plots of the best GAM (lowest AICc) for presence of the kelp Laminaria hyperborea.
Figure 3.
Interactions in the BRT model for kelp recovery.
Predictions based on the main interactions from the BRT model of kelp Laminaria hyperborea recovery: wave exposure x depth (left, interaction size = 48.9) and curvature x depth (right, interaction size = 40.9).
Figure 4.
GAM for sea urchin persistence.
The partial response plots of the best GAM without interactions for presence of sea urchin Strongylocentrotus droebachiensis.
Figure 5.
Interactions in the BRT model for sea urchin persistence.
Predictions based on the main interactions from the BRT model of occurrence/persistence of sea urchin Strongylocentrotus droebachiensis: depth x latitude (left, interaction size = 205.5) and minimum current speed x latitude (right, interaction size = 91.2).
Figure 6.
The extent of kelp recovery within the study area.
Size of the predicted area (in km2 on the left axis and in percent of the study area on the right axis) modeled to be areas with recovery of kelp (Laminaria hyperborea, lh) and with persistence of sea urchins (Strongylocentrotus droebachiensis, sd), given different cut-off probabilities for the best GAMs and for the BRT models. The low and high optimal threshold lines show the lowest and highest cut-off value according to the Youden index for the best models.
Figure 7.
Maps showing predicted kelp recovery and sea urchin persistence.
Predicted recovery of kelp Laminaria hyperborea (orange) and persistence of sea urchin Strongylocentrotus droebachiensis (green) based on the BRT models in region 1 (65°N, lower panel), region 3 (67°N, middle panel) and region 5 (69°N, upper panel). The map further shows land (beige), wave exposed areas where kelp never has been grazed (leather brown), field observed presences of L. hyperborera (brown circles), S. droebachiensis (violet circles) and absences of both species (open circles). White areas include shallow areas with predicted absences of the two species, areas deeper than 30 meters, as well as sheltered areas that are excluded from the predicted area. Scale 1∶100 000.
Table 1.
Estimates of recovered kelp area within 5 regions.
Figure 8.
Summer sea water temperature along latitude in the period 1990 to 2007.
Modeled sea water temperature at 10(left) and July (right) at latitudes between 63–70°N in the period 1990 to 2007, at an intermediate position along the coast-ocean gradient. The analysis is based on data from the Temperature Atlas developed by [28].