Goetz et al. 10.1073/pnas.0506179102. |
Fig. 5. Spatial distribution of significant trends in seasonal photosynthetic activity across Canada and Alaska from 1982 through 2003 using the Vogelsang test (compare with Fig. 4 in the text). This figure corresponds to Fig. 4 in the text in showing trends in photosynthetic activity (Pg) over boreal North America. The Vogelsang test was used to determine the presence of a trend and is offered as a comparison against the augmented Dicky-Fuller (ADF) tests presented in the paper. The Vogelsang test determines whether the presence of a trend in a time series is significantly different from zero (b ¹ 0, P £ 0.05), and is robust to the form of serial correlation in the time series, i.e., it does not over-reject in the presence of nonstationary errors due to strong serial autocorrelation.
Table 4. Magnitude of slopes for significant models (β ≠ 0, P £ 0.05) of tundra and forest vegetation cover types using the Vogelsang test
Slope magnitude* | Tundra, m ha | Forest, m ha |
Strong negative | 0.2 (1.3%) | 2.3 (12.2%) |
Negative | 1.5 (9.0%) | 11.6 (60.7%) |
Near zero† | NA | NA |
Positive | 13.1 (79.8%) | 4.3 (22.3%) |
Strong positive | 1.6 (9.9%) | 0.9 (4.8%) |
This table corresponds to Table 2 in the text and summarizes the magnitude of the significant slopes by land cover using the Vogelsang test. See Figure 5 for more information.
* Magnitude categories correspond to Fig. 4.
†
The Vogelsang test does not discriminate between stationary and nonstationary models making a near zero slope comparison with the ADF tests impossible (where nonsignificant stationary models were included as "near zero" in Table 2).