Red Kite

7.2.6.1 Spatial distribution and migration direction

According to Karlsson et al. (2004) 500 Red Kites (Figure 57) leave Falsterbo on an average autumn season.

Based on the rangefinder data collected during the baseline at Falsterbo 12% of the birds have directions indi-cating that they will cross the Arkona Basin, whereas the vast majority of directions are concentrated around SW in the direction of Stevns Klint in Denmark (Figure 56, Figure 58). Although some birds may leave Sweden before they reach Falsterbo the above proportion is most likely a reasonable approximation of the number of Red Kites crossing, which using the mean figure from Falsterbo equals 60 birds. No figures on spring migration of raptors through the region are available, neither from this or other studies.

Rough-legged Buzzard

97 Figure 56. Migration tracks of Red Kite collected in the study area, spring and autumn 2013. Radar-based tracks are marked by blue lines, and rangefinder-based tracks by red lines.

Figure 57. Red Kite.

98 Figure 58. Sampled migration directions of Red Kite at Falsterbo, autumn 2013. Numbers on the Y-axes refer to sample size (number of recordings by laser rangefinder). Each wedge represents a sector of 15°. The mean direc-tion is indicated by the black line running from the centre of the graph to the outer edge. The arcs extending to either side represent the 95% confidence limits of the mean direction.

7.2.6.2 Migration altitude

Migrating Red Kites show typical patterns of flight altitude with a wide range of altitudes as the birds leave Fal-sterbo as compared to the arrival coasts in Denmark (Figure 59, Figure 60). Too few records of kites were made on FINO 2 to allow for computation of frequency distribution. The angle of descend in Red Kites seems to be quite shallow, yet beyond 20 km from the Swedish most if not all birds fly below 200 m altitude (Figure 60). The angles of descend differ significantly during different wind directions (Table 18).

Table 18. Results of homogeneity of slope test testing whether tracks of Red Kite during different wind directions have different responses of altitude to distance to land. Only the results for the interaction between wind direc-tion and distance to land are shown for the DHI rangefinder data collected from the Swedish coast autumn 2013.

99 Figure 59. Frequency distribution of altitude measurements of Red Kite by laser rangefinder at the Swedish south coast and at the Danish coast during autumn 2013.

Red Kite Swedish coast

100 200 300 400 500 600 700 800 900 1000

0

100 200 300 400 500 600 700 800 900 1000

0

100 Figure 60. Changes in sampled altitude of Red Kite by laser rangefinder during 2013 in relation to distance from the departure coast (north Germany in spring and south Sweden in autumn). Mean and confidence interval are given for different distances from departing coast and wind direction.

The GAMM flight model for Red Kite in autumn indicates that the birds flew higher in tail winds (northerly winds) and first ascend in altitude after leaving the coast, but further offshore start descending again (Figure 61). The predictive accuracy of the GAMM was high, with a good agreement between observed and predicted altitudes, a Spearman’s rank correlation of 0.72, when the model was evaluated on semi-independent data (Table 19, Figure 62). The adjusted R² indicated a reasonable fit (Table 19). The model successfully accounted for the strong temporal and spatial autocorrelation in the track data by using the correlation structure and ran-dom term (serial and spatial autocorrelograms and model diagnostics are shown in Appendix A).

According to the predictions the birds flew on average below rotor height of the 10 MW turbines during all wind conditions (Figure 63). However, as observation are recorded only within about 25 km from the departure coast, the predictions are most likely under-estimations and the flight altitudes at the wind farm could be ex-pected to be close to the predicted altitude at about 25 km from the departure coast (rotor height, Figure 63).

Graphs of the predictions including model standard errors are shown in the Appendix A.

Table 19. Significance and F-values for the fixed parametric (wind directions) and smooth terms included in the GAMMs for the Red Kite. Adjusted R-square indicates the variance explained by the model and the Spearman’s correlation coefficient the agreement between predicted and evaluated altitudes (by a split sample evaluation approach). Number of samples used in the analysis is shown on the bottom row.

Red Kite

101

F-value p-value

Smooth Distance to coast:

Wind direction

24.565 <0.01

Clearness 2.993 0.07

R-sq. (adj) 0.28

Spearman’s corr. 0.72

N of tracks (samples) 151 (574)

Figure 61. GAMM response curves for the Red Kite. Both a perspective plot (3d) and a contour plot (2d) are shown for the interaction term (the tensor product smoother). The response is on the scale of the linear predic-tor. The degree of smoothing is indicated in the title of the interaction term (of the perspective plot) and in the title of the Y-axis for the 1d smooth functions. The shaded areas show the 95% Bayesian confidence intervals.

Confidence intervals are not shown for the interaction term to improve interpretability.

102 Figure 62. Split sample evaluation results: predicted average flight altitudes of Red Kites against observed alti-tudes. The model was fitted on 70% of the tracks and was tested on 30%. The black line is a regression line based on a linear regression between observed and predicted altitudes. If the model would be perfectly calibrat-ed all points would lie on the dashcalibrat-ed line.

103 Figure 63. Average predicted altitude for Red Kites in relation to distance from the coast of Sweden during dif-ferent wind directions and wind speeds. All other predictor variables were set to mean values within the species specific data set. The lines are the predicted flight altitudes and the black rectangle indicates the rotor swept ar-ea by 10 MW turbines.