2.4 Eco-effect analysis
The study operated partial-correlation analyses to characterize the eco-effects about EP of LEP, AEP, and HEP ecosystems responding to snow-onset and -end SP. The used means can exclude the confounding effects caused by other climatic variables (e.g., temperature and solar radiation) and the covariate effect (Piao et al., 2016) possibly existing between those interannual snow-onset and -end SP series of the same sites. The specific operation was deployed onto EP case by case (LEP, AEP, and HEP), with the performance of SP in the two different cases (snow-onset and -end SP) compared first and then the optimal SP case identified. Note that in subject to the relatively short terms of the examined EP and SP datasets, all of the partial correlation analyses were operated at the significance level of p<0.1 (indicated by *) for extracting more potential sites (i.e., 0.5°×0.5° geo-grids) with distinguishable responses.
With the basic method of analyzing the EP-SP ecological links determined, the answers to the two goal questions were sought then. First, statistics of the three kinds of EP cases that performed with the highest correlations with the corresponding- or surrounding-geo-grid-related snow-onset or -end SP were individually operated all over the study area. Then, we made their comparisons, with particular interest on the scenario showing more cases that were sensitive to the SP around – we termed this scenario as “neighborhood eco-effect”, by referring to the mechanism-approximate concept in social and economic sciences usually positing that neighborhoods have either direct or indirect effects on individual human behaviors (MacAllister et al. , 2001). Note that theoretically the correlation-based eco-effect exploring and comparison as above were based on two interannual EP and SP series at the individual geo-grid level, different from those traditional spatial correlation analyses based on two spatial maps of different variables. Thus, the issue of autocorrelation often met in the latter scenario (Macias-Fauria et al. , 2012) and its related solutions such as the Monte Carlo-based techniques (Livezey & Chen, 1983) could be omitted in this study. Further, after the existence of the neighborhood eco-effect was detected, a statistical comparison of its EP-SP cases in terms of the common eight directions – as “directional eco-effect” – was operated to reveal the primary orientations of SP driving EP, for different regions such as North America (>45°N) (NA) and North Eurasia (>45°N) (NE). Simultaneously identifying these two eco-effects was considered as the kernel theoretical basis of validating the existence of periconnection.
To quantify the traits of the neighborhood eco-effect in details, a 3×3 window for each geo-grid (0.5°×0.5°) was set to restrict the conditions in operating its partial-correlation analyses (termed as correlation window, whose eight surrounding geo-grids from the right upper one relate to north, northeast, east, southeast, south, southwest, west, and northwest in a clockwise way). That is, for each targeted geo-grid the correlations between its serial LEP, AEP, and HEP and the synchronous SP series of both the same geo-grid and all the eight geo-grids around were analyzed. Then, in order to characterize the features of the directional eco-effect, the sum of the occurrences for any a targeted surrounding geo-grid and its two neighboring geo-grids all at the edges of the correlation window in the derived neighborhood-accounted responses was used as an index. If the index is large, the directional eco-effect from the targeted geo-grid to the centered geo-grid is strong. After the two steps accomplished as above, the primary traits of periconnection could be characterized.