**3. Results**

#### **3.1 Functional characterization of Sierra Nevada oak woods**

The Pyrenean oak forests of Sierra Nevada showed a heterogeneous spatial behavior in terms of their EVI seasonal dynamics. In general, woods of the southern slope of Sierra Nevada displayed greater annual vegetation greenness and longer growing seasons than those from the northern slope (Figures 3 and 4). The seasonal EVI curve of the oak woods in the northern slope (Figure 3) showed a gradual increase in productivity that begins around March and that reaches its maximum peak in late May - early June (Figure 5e). Then, senescence takes place at a similar but slightly lower rate than growth. In contrast, the EVI curves of southern-slope woods (Figure 4) show a later but much steeper start of the growing season in late April - early May, reaching the EVI maximum value in June, as in the northern slope woods (Figure 5e). Then, EVI maintains a slowly decreasing plateau until around November, when a less pronounced end of the growing season than in the northern woods occurs.

Statistical comparisons of the EVI attributes (Figure 5) among oak woods also revealed the former differences. In general, Northern oak woods had significantly lower EVI\_mean values than southern ones (ANOVA: F=33.56; p=0.0000; n=177; Figures 5a and 6a). Dílar woods (Figure 3d) showed the lowest values and Poqueira (Figure 4c) the highest. The EVI\_sCV displayed greater values in the north than in the south (ANOVA: F=29.35; p=0.0000; n=177; Figures 5b and 6b) and a much greater dispersion of data in the north. Although MAX values (Figures 5c and 6c) showed significant differences between some oak woods (Kruskal Wallis: H=36.94; p=0.0000; n=177), there were no clear differences between the northern and southern woods. In general, Max values showed little inter-woods, but large intra-wood variation. We hypothesize that this larger intra-wood variation could be related to greater altitudinal range, such as in Alhama, Genil, Chico and Trevélez (Table 1). DMAX did not either significantly differ between the northern and southern woods, happening in May-June in all oak woods but coming about later with altitude. The increase of intra-wood variability with greater altitudinal variation was also observed in DMAX **(**Kruskal Wallis: H=64.61; p=0.0000; n=177; Figures 5e and 7a). Regarding MIN values, southern woods showed significantly higher values than northern woods (Kruskal-Wallis: H=126.05; p=0.000; n=177; Figures 5d and 6d), which is directly related to EVI\_mean (Figures 5a and 6a) and EVI\_sCV (Figures 5b and 6b). Contrary to DMAX, DMIN showed great variability both within and among woods (from November to April) (Figures 5f and 7b) (May-July), with earlier DMIN values in the northern woods than in the southern ones (Kruskal-Wallis: H=86.93; p=0.0000; n=177; Figures 5f and 7b).

p=0.873, n=177) and for EVI\_sCV a Box-Cox transformation (Shapiro-Wilk, W=0.983, p=0.031, n=177; Levene's Test F=1.951, p=0.055, n=177). The slight but not significant deviation from normality for the EVI\_sCV data did not affect results. For those attributes that even transformed did not fulfill normality (MAX, MIN, DMAX, and DMIN), the analysis was conducted using the non-parametric Kruskal-Wallis test. To determine which groups significantly differed from each other, we used multiple *post hoc* comparisons, using the Tukey test for EVI\_mean and EVI\_sCV, and the Bonferroni test for MAX, MIN, DMAX,

The Pyrenean oak forests of Sierra Nevada showed a heterogeneous spatial behavior in terms of their EVI seasonal dynamics. In general, woods of the southern slope of Sierra Nevada displayed greater annual vegetation greenness and longer growing seasons than those from the northern slope (Figures 3 and 4). The seasonal EVI curve of the oak woods in the northern slope (Figure 3) showed a gradual increase in productivity that begins around March and that reaches its maximum peak in late May - early June (Figure 5e). Then, senescence takes place at a similar but slightly lower rate than growth. In contrast, the EVI curves of southern-slope woods (Figure 4) show a later but much steeper start of the growing season in late April - early May, reaching the EVI maximum value in June, as in the northern slope woods (Figure 5e). Then, EVI maintains a slowly decreasing plateau until around November, when a less pronounced end of the growing season than in the northern

Statistical comparisons of the EVI attributes (Figure 5) among oak woods also revealed the former differences. In general, Northern oak woods had significantly lower EVI\_mean values than southern ones (ANOVA: F=33.56; p=0.0000; n=177; Figures 5a and 6a). Dílar woods (Figure 3d) showed the lowest values and Poqueira (Figure 4c) the highest. The EVI\_sCV displayed greater values in the north than in the south (ANOVA: F=29.35; p=0.0000; n=177; Figures 5b and 6b) and a much greater dispersion of data in the north. Although MAX values (Figures 5c and 6c) showed significant differences between some oak woods (Kruskal Wallis: H=36.94; p=0.0000; n=177), there were no clear differences between the northern and southern woods. In general, Max values showed little inter-woods, but large intra-wood variation. We hypothesize that this larger intra-wood variation could be related to greater altitudinal range, such as in Alhama, Genil, Chico and Trevélez (Table 1). DMAX did not either significantly differ between the northern and southern woods, happening in May-June in all oak woods but coming about later with altitude. The increase of intra-wood variability with greater altitudinal variation was also observed in DMAX **(**Kruskal Wallis: H=64.61; p=0.0000; n=177; Figures 5e and 7a). Regarding MIN values, southern woods showed significantly higher values than northern woods (Kruskal-Wallis: H=126.05; p=0.000; n=177; Figures 5d and 6d), which is directly related to EVI\_mean (Figures 5a and 6a) and EVI\_sCV (Figures 5b and 6b). Contrary to DMAX, DMIN showed great variability both within and among woods (from November to April) (Figures 5f and 7b) (May-July), with earlier DMIN values in the northern woods than in the southern ones

**3.1 Functional characterization of Sierra Nevada oak woods** 

(Kruskal-Wallis: H=86.93; p=0.0000; n=177; Figures 5f and 7b).

and DMIN. See Figure 5.

**3. Results** 

woods occurs.

Fig. 3. EVI seasonal dynamics (left Y axis, in gray) and 2001-2009 EVI trends (right Y axis, in black) for oak forests in the northern slope of Sierra Nevada. The horizontal "zero-trend" line shows the absence of significant trends. The two vertical dotted gray lines show the beginning and the end of the growing season.

Satellite-Based Monitoring of Ecosystem Functioning in Protected Areas:

**3.2 Functional changes in Sierra Nevada oak woods** 

the early-summer (Genil, Dílar, and Durcal).

**Area (ha)/ Pixels sampled (n)**

**Altitudinal** 

**range Aspect Slope**

Alhama 266/36 1443-1838 NE 25º 0 20 0/1 Genil 356/51 1272-1792 N 30º 0 29 0/14 Monachil 104/15 1630-1842 N 27º 0 8 0/1 Dílar 111/14 1594-1884 NW 31º 1 7 0/1 Dúrcal 58/4 1598-1833 W 28º 0 2 0/0 Chico 445/39 1459-1870 S 17º 1 21 0/9 Soportújar 46/4 1652-1755 SW 17º 1 1 0/0 Poqueira 105/5 1635-1888 SE 25º 1 2 0/0 Trevélez 167/9 1397-1880 E 24º 1 5 0/1

Table 1. Environmental traits and EVI\_mean trends during the 2001-2009 period in nine *Quercus pyrenaica* oak woods of Sierra Nevada National Park. Forest patches are named according the river basin where they are located: Alhama, Genil, Monachil, Dílar, and Dúrcal, in the northern slope; and Chico, Soportújar, Poqueira, and Trevélez in the southern

4d).

**Oak woods** 

slope.

Recent Trends in the Oak Forests (*Quercus pyrenaica* Willd.) of Sierra Nevada (Spain) 363

We found significant functional changes happening in the Sierra Nevada oak woods during the 2001-2009 period. Though we did not observe significant long-term trends in the annual synthetic EVI attributes, particular periods of the year did show significant EVI trends. The greatest significant trends occurred at the beginning of the growing season, when strong EVI decreases were observed (March-April), particularly in the northwestern slope (Figure 3). A clearly marked downward trend in productivity was observed between 7th April - 23rd April), which took place in four out of the five northwestern oak woods (Genil, Monachil, Dílar, and Dúrcal, Figures 3b, 3c, 3d, and 3e). Alhama oak wood (Figure 3a) was the only exception, displaying no long-term trends. Some northern woods also showed small positive EVI trends in November (Genil, Monachil, and Dílar; Figures 3b, 3c, and 3d) and in

The southern oak woods (Figure 4) also showed a decrease of vegetation greenness at the beginning of the growing season (except Poqueira, Figure 4c), but less deep than in the northern woods. In addition, EVI increases were observed in middle to late summer in three out of four southern woods (Soportújar, Poqueira, and Trevélez (Figures 4b, 4c, and

**Environmental traits # of pixels with EVI\_mean trends** 

**Positive Sen's slope (+)**

**Negative Sen's slope (-)** 

**M-Kendall Significant (p≤0.15) (+/-)** 

Fig. 4. EVI seasonal dynamics (left Y axis, in gray) and 2001-2009 EVI trends (right Y axis, in black) for oak forests in the southern slope of Sierra Nevada. The horizontal "zero-trend" line shows the absence of significant trends. The two vertical dotted gray lines show the beginning and the end of the growing season.

#### **3.2 Functional changes in Sierra Nevada oak woods**

362 International Perspectives on Global Environmental Change

**a) Chico**

0 0.1 0.2 0.3 0.4 0.5

0 0.1 0.2 0.3 0.4 0.5

0 0.1 0.2 0.3 0.4 0.5

0 0.1 0.2 0.3 0.4 0.5

**EVI**

1 Jan.

beginning and the end of the growing season.

17 Jan.

2 Feb.

18 Feb.

6 Mar.

22 Mar.

7 Apr.

23 Apr.

9 May.

25 May.

10 Jun.

26 Jun.

Fig. 4. EVI seasonal dynamics (left Y axis, in gray) and 2001-2009 EVI trends (right Y axis, in black) for oak forests in the southern slope of Sierra Nevada. The horizontal "zero-trend" line shows the absence of significant trends. The two vertical dotted gray lines show the

12 Jul.

28 Jul.

13 Aug.

29 Aug.

14 Sep.

30 Sep.

16 Oct.

1 Nov.

17 Nov.

3 Dec.

19 Dec.

**EVI**

1 Jan.

17 Jan.

2 Feb.

18 Feb.

6 Mar.

22 Mar.

7 Apr.

23 Apr.

9 May.

25 May.

10 Jun.

26 Jun.

**d) Trevélez**

12 Jul.

28 Jul.

13 Aug.

29 Aug.

14 Sep.

30 Sep.

16 Oct.

1 Nov.

17 Nov.

3 Dec.

19 Dec.

**EVI**

1 Jan.

17 Jan.

2 Feb.

18 Feb.

6 Mar.

22 Mar.

7 Apr.

23 Apr.

9 May.

25 May.

10 Jun.

26 Jun.

**c) Poqueira**

12 Jul.

28 Jul.

13 Aug.

29 Aug.

14 Sep.

30 Sep.

16 Oct.

1 Nov.

17 Nov.

3 Dec.

19 Dec.

**EVI**

1 Jan.

17 Jan.

2 Feb.

18 Feb.

6 Mar.

22 Mar.

7 Apr.

23 Apr.

9 May.

25 May.

10 Jun.

26 Jun.

**b) Soportújar** 

12 Jul.

28 Jul.

13 Aug.

29 Aug.

14 Sep.

30 Sep.

16 Oct.

1 Nov.

17 Nov.

3 Dec.

19 Dec.





**EVI Trends**

**EVI Trends**

**EVI Trends**

**EVI Trends**

We found significant functional changes happening in the Sierra Nevada oak woods during the 2001-2009 period. Though we did not observe significant long-term trends in the annual synthetic EVI attributes, particular periods of the year did show significant EVI trends. The greatest significant trends occurred at the beginning of the growing season, when strong EVI decreases were observed (March-April), particularly in the northwestern slope (Figure 3). A clearly marked downward trend in productivity was observed between 7th April - 23rd April), which took place in four out of the five northwestern oak woods (Genil, Monachil, Dílar, and Dúrcal, Figures 3b, 3c, 3d, and 3e). Alhama oak wood (Figure 3a) was the only exception, displaying no long-term trends. Some northern woods also showed small positive EVI trends in November (Genil, Monachil, and Dílar; Figures 3b, 3c, and 3d) and in the early-summer (Genil, Dílar, and Durcal).

The southern oak woods (Figure 4) also showed a decrease of vegetation greenness at the beginning of the growing season (except Poqueira, Figure 4c), but less deep than in the northern woods. In addition, EVI increases were observed in middle to late summer in three out of four southern woods (Soportújar, Poqueira, and Trevélez (Figures 4b, 4c, and 4d).


Table 1. Environmental traits and EVI\_mean trends during the 2001-2009 period in nine *Quercus pyrenaica* oak woods of Sierra Nevada National Park. Forest patches are named according the river basin where they are located: Alhama, Genil, Monachil, Dílar, and Dúrcal, in the northern slope; and Chico, Soportújar, Poqueira, and Trevélez in the southern slope.

Satellite-Based Monitoring of Ecosystem Functioning in Protected Areas:

during the start of the growing season than in the northern woods (Figure 3).

Our study also showed that though the oak woodlands of Sierra Nevada have not experienced significant changes of the EVI\_mean during the 2001-2009 period, they have suffered seasonal functional changes that mainly affected the beginning of the growing season. In contrast to this relative stability of annual mean vegetation greenness (EVI\_mean) since 2001, previous evaluations showed a significant increase in vegetation greenness throughout the eighties and nineties in Sierra Nevada (see Alcaraz-Segura et al., 2008b for the 1981-2003 period, and Alcaraz-Segura et al., 2009b for the 1982-2006 period). Such evaluations used the GIMMS-AVHRR (Global Inventory Modelling and Mapping Studies - Advanced Very High Resolution Radiometer) NDVI dataset. Though there is some debate on the existence of a long-term bias in the GIMMS dataset towards NDVI increases in some

**4. Discussion** 

**woods of Sierra Nevada National Park** 

Recent Trends in the Oak Forests (*Quercus pyrenaica* Willd.) of Sierra Nevada (Spain) 365

**4.1 Baseline conditions and trends in the ecosystem functioning of the Pyrenean oak** 

Our approach, based on a time series of satellite-derived images of the EVI, provided a description of how different attributes of ecosystem functioning change across the remaining locations of Pyrenean oak woodlands in Sierra Nevada. This reference description provides the baseline conditions of ecosystem functioning that can be used to assess the effects of environmental changes on ecosystems processes. The Pyrenean oak woodlands of Sierra Nevada showed a unimodal EVI seasonal dynamics with a unique and well-defined growing season centered in summer and winter minima, as observed in previous works (Alcaraz-Segura et al., 2009a). Differences among locations mainly occurred during the winter non-growing season and at the beginning of the growing season (spring) and were mainly related to the location in the north or south slopes of Sierra Nevada. The lower EVI\_mean values in the northern oak woods (Figure 5a) are closely linked to the presence of lower winter MIN values than in the southern woods (Figure 5d) and with the more abrupt EVI decrease during the autumn. In contrast, southern woods maintained relatively high EVI values throughout their longer growing season (Figure 4). The greater annual vegetation greenness of southern woods is probably due to the greater incidence of solar radiation that favors longer growing seasons, milder temperatures during the winter, and an extra water supply from humid air masses coming from the Mediterranean sea that compensate the very high evapotranspiration rates during the summer, in comparison to the colder and more continental locations of the northern slope (Costa Tenorio et al., 2005). Contrary, summer maximum EVI values (MAX) would not cause significant differences in annual vegetation greenness between the northern and southern locations. In consequence, the northern slope shows much greater seasonality (EVI\_sCV) than the southern slope since MAX values are similar in both orientations, though the northern woods showed lower MIN values than the southern ones (Figure 5d). From the analysis of the shape of the EVI seasonal curves and according to previous studies (Alcaraz-Segura et al., 2009a), the main limiting factors for vegetation greenness in the oak woodlands of Sierra Nevada are low winter temperatures and lower solar irradiation in the northern slope, which favors a longer presence of snow (Figure 5d). An important point to consider is that the greater vegetation greenness of the southern woodlands during the non-growing season is not related to the activity of the oak trees (because they are winter semi-deciduous), but to the shrubs and herbaceous vegetation occupying the undergrowth vegetation and the patches without trees (Figure 8). In the same way, since the snow melt happens faster and earlier in the southern woods, undergrowth vegetation is also responsible for the earlier and more pronounced rise in vegetation greenness

Fig. 5. Functional characterization of the oak woods of Sierra Nevada (Spain) based on the EVI attributes for the 2001-2009 period. Letters show significant differences in *post hoc* comparisons. a) EVI annual mean, an estimator of annual primary production; b) EVI seasonal Coefficient of Variation, a descriptor of seasonality; c) Maximum and d) Minimum EVI annual values, indicators of the maximum and minimum photosynthetic activity; Dates when the e) Maximum and f) Minimum EVI values are reached, indicators of phenology.
