• Climatic warming produces significant gradual alterations in the timing of life-cycle events, and here we study the phenological effects of rainfall-pattern changes.
• We conducted ecosystem field experiments that partially excluded rain and runoff during the growing season in a Mediterranean forest and in a mediterranean shrubland. Studies of time-series of leaf-unfolding, flowering and fruiting over the last 50 yr in central Catalonia were carried out, and greenup onset in the Iberian Peninsula was monitored by satellite images.
• Experimental, historical and geographical changes in rainfall produced significant, complex and strongly species-specific, as well as spatially and temporally variable, phenological effects. Among these changes, it was found that in the Iberian Peninsula, greenup onset changes from spring (triggered by rising temperatures) in the northern cool-wet regions to autumn (triggered by the arrival of autumn rainfalls) in the southern warm-dry regions. Even in the mesic Mediterranean central Catalonia (NE of the peninsula) rainfall had a stronger relative influence than temperature on fruiting phenology.
• The results show that changes in rainfall and water availability, an important driver of climate change, can cause complex phenological changes with likely far-reaching consequences for ecosystem and biosphere functioning and structure. The seasonal shift in the Iberian Peninsula further highlights this importance and indicates that vegetation may respond to climate change not only with gradual, but also with abrupt temporal and spatial, changes in the timing of greenup onset.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Phenology monitoring, which is very simple to record ‘on-the-ground’ and constitutes a hobby for many people, has thus become a tool of great scientific value for assessing the effects of climate change (Sparks et al., 2000). However, the vast majority of studies into the effects these changes have on phenology have focused on warming as the main component of climatic change producing gradual phenological alterations. Less attention has been devoted to the fact that phenology is also responsive to other climatic changes such as rainfall and evaporation in at least most regions of the planet (Rathcke & Lacey, 1985; Defila & Clot, 2001). Likewise, the fact that phenological alterations could occur in an abrupt rather than gradual fashion when particular temperatures and aridity thresholds are surpassed has not been considered.
Water is needed for plant growth. It is required for both leaf and floral bud expansion, as well as for leaf and flower turgor maintenance under evapotranspirational demand (Galen et al., 1999). Rainfall is known to affect leaf and flower phenology even in regions with good water supply such as in the oceanic-continental gradients in Norway (Wielgolaski, 2001) or in seasonal tropical forests where flowering is often induced by rainfall (Rathcke & Lacey, 1985). Rainfall is, of course, more determinant in dry regions. It stimulates greenup onset and determines the duration of growth and flowering of some desert plants (Fox, 1990; Abd El-Ghani, 1997; Ghazanfar, 1997). It has also been shown to affect leafing and flowering phenology in Mediterranean plants (Peñuelas et al., 2002) and in trees in tropical dry forests (Borchert, 1994). There are, however, other studies that report that the effects of soil moisture on both the timing and duration of flowering in some ecosystems such as subalpine meadows (Dunne et al., 2003) or on leafing and flowering in some species from otherwise responsive Mediterranean ecosystems (Peñuelas et al., 2002) are nonsignificant.
Our study was based on the following six premises: climate projections predict for the next decades decreased rainfall and overall drier conditions in some regions of the globe such as the Mediterranean (IPCC, 2001); these regions have already experienced a progressive aridification over recent decades (Piñol et al., 1998; Peñuelas et al., 2002; Peñuelas & Boada, 2003); these regions present great geographical and temporal variability in rainfall and water availability (Peñuelas, 2001); diverse phenological responses have been described for different species (Peñuelas et al., 2002) and for different regions (Ahas et al., 2002); very few experimental studies at ecosystem level simulating the climate scenarios predicted for the next decades have ever been conducted; and the possibility of nongradual responses to climate changes exist. These six premises led us to: conduct ecosystem field experiments that excluded rain and runoff during the growing season in a Mediterranean forest and in a Mediterranean shrubland, thereby producing a 15–20% decrease in soil water availability (as forecasted over the next decades by most GCM (IPCC, 2001) for this region); carry out historical studies of time series of leaf-unfolding, flowering, fruiting, temperature, and precipitation over the last 50 years in central Catalonia, and; to monitor by means of satellite images greenup onset in the Iberian peninsula, a geographically and climatologically very diverse region with abrupt changes of vegetation types. In these studies, we studied the spatio-temporal variability in the phenological responses of different dominant species and regions to changes in rainfall patterns.
Materials and Methods
Mediterranean forests In order to study the phenological responses of Mediterranean forest vegetation to a decrease in rainfall we conducted an experimental manipulation of a Mediterranean forest in Les Muntanyes de Prades (41°13′-N, 0°55′-E, southern Catalonia) on a south-facing slope (25% slope) at 930 m above sea level. The average annual temperature there is 12°C and the annual rainfall is 658 mm. Summer drought is pronounced from approximately mid-June to mid-September. We submitted four 10 × 15 m forest plots to a soil-water availability reduction of approx. 15% by using plastic strips and funnels that partially excluded rain throughfall (30% of the plot surface) and by ditch exclusion of water runoff (Ogaya & Peñuelas, 2003). Four other 10 × 15 m forest plots that were not submitted to any kind of treatment were considered as control plots.
An automatic meteorological station installed in between the plots monitored temperature, photosynthetic active radiation, air humidity and precipitation every half-hour. Soil moisture was measured every 2 wk throughout the experiment by Time Domain Reflectometry (Tektronix 1502C, Beaverton, Oregon, USA) (Gray & Spies, 1995). Three stainless steel cylindrical rods, 25 cm long, were left permanently and fully driven into the soil at four randomly selected places in each plot. The time domain reflectometer was connected to the ends of the rods in each measurement.
We monitored the presence or absence of flowering (flowers in anthesis) in two dominant species, Arbutus unedo and Phillyrea latifolia, once a week in 1999 and 2000. The intensity of flowering was visually estimated as the percentage of plants in anthesis.
In the two subsequent years, 2000 and 2001, we also measured the ratio of fruit mass to the total above-ground tree biomass in each species. The fruits were collected in 20 circular baskets (27 cm diameter with a 1.5 mm mesh) randomly distributed on the ground of each of the eight plots. The fruits were collected every 15 d from winter 1999 to winter 2001. The dry weight of the fruit was measured after being dried in an oven at 70°C until constant weight. The above-ground tree biomass was estimated through allometric relationships with the stem-diameter height at 50 cm measured in P. latifolia growing within the area of study but outside the plots. To estimate the biomass in Arbutus unedo we used the allometric relationship calculated a few years ago in the same area by Lledó (1990). The stem diameter was measured each winter in all P. latifolia and A. unedo trees growing in the plots.
Mediterranean shrubland In order to study the phenological responses of Mediterranean shrub vegetation to a decrease in rainfall we conducted an in situ manipulation of the climate in the shrublands of the hills of Garraf (41°18′ N, 1°49′ E) in central coastal Catalonia at 210 m above sea level on a SSE-facing slope (13°). We used a nonintrusive procedure consisting of automatic and transparent waterproof roofs at c. 0.2–0.5 m above the vegetation which covered the vegetation only when it was raining; in this fashion, we excluded rainfall in 5 × 6 m plots (three per treatment) for 2 months during the spring and autumn growing seasons (1999–2002) (Beier et al., 2003; Peñuelas et al., 2003). The latitude and location of these hills, near the coast, give a typical Mediterranean climate (during the 3 years of the study, the average annual temperature was 13.8°C and the annual rainfall 450 mm).
Meteorological measurements were made at the study site. Precipitation was measured with a standard rain gauge and soil moisture was measured at three fixed sampling points per plot every 1–2 wk, using Time Domain Reflectometry (TDR). Air 20 cm above ground and soil temperatures (0, 2 and 10 cm depth) were obtained by means of temperature sensors RTD Pt100 1/3 DIN (Desin Instruments, Barcelona, Spain) located in the experimental plots. Temperatures were measured every 10 min and the average of the measurements for the three sensors was recorded.
We studied the phenological responses in flowering in the two dominant species of these Mediterranean coastal shrublands: Erica multiflora L. (Ericaceae) and Globularia alypum L. (Globulariaceae). During the flowering seasons of 1999 and 2000, we measured the percentage of plants with functional (well-developed) flowers or functional flower-heads in E. multiflora and G. alypum, respectively.
Over the two subsequent years, 2000 and 2001, we recorded all emerging seedlings belonging to these two species in eight 400 cm2 (20 × 20 cm) quadrats randomly sited in each plot.
Statistical analyses We employed the Kaplan-Meyer nonparametric method for the computation of survival curves and used the onset of flowering as a survival time. We thereafter used the Log-Rank test to assess treatment differences. Analyses of variance (anova) were conducted with the amount and duration of flowering, the fruit-litter biomass proportion or the number of seedlings in each plot as dependent variables, and the treatment application (and year) as the independent factor(s). The data of the percentage of plants flowering was transformed to arcsin (p)1/2 to reach the normality assumptions of the parametric statistical analyses. All these analyses were performed with the Statview (Abacus Concepts Inc., 1998), the SPSS (SPSS Inc., 2002) and the Statistica (StatSoft Inc., 2001) software packages.
We conducted a historical study of a complete set of phenological data from a site in central Catalonia (Cardedeu field station, 41°34′-N, 2°21′-E) where temporal and temperature-related changes in phenophases have been described in a previous study (Peñuelas et al., 2002). From 1952 to 2000, the dates of leaf-unfolding, flowering and fruiting were recorded for 103 plant species by one of us (P. C.) in the area surrounding the Cardedeu field station as part of a long-term monitoring program of the meteorology and phenology of the area. Monitoring was conducted by daily sightings from key points, plus surveys from a 5–10 km circuit within the 100 km2 study site. This study site is located in a plain with little topographic variability, and includes forests, fields and gardens. Phenological events were recorded with an estimated accuracy of ±1 d: six or more individual plants per species were monitored and the phenophase was recorded once the whole plant had reached the functional phenological stage.
For all available phenological records covering more than 20 yr (25 deciduous woody species for leaf-unfolding, 57 plant species for flowering and 27 plant species for fruiting), we calculated linear annual trends and their significance according to an F-test. Regression analyses were performed to determine the effect of precipitation in the months before the phenophase date. Multiple regression analyses were also performed to assess the relative role of these precipitations and temperatures. The above mentioned Statview, SPSS and Statistica software packages were used.
To assess greenup onset in the whole of the Iberian peninsula, we used images from MODIS (Moderate Resolution Imaging Spectroradiometer) on board NASA's Terra satellite (Zhang et al., 2003). The MODIS instrument possesses seven spectral bands specifically designed for land applications that have spatial resolutions ranging from 250 m to 1 km (Justice et al., 1997). Using daily multiangle, cloud-free and atmospherically corrected surface reflectances collected over 16-day periods, the MODIS bi-directional reflectance distribution function (BRDF)/Albedo algorithm generates one nadir BRDF-adjusted reflectance (NBAR) for each MODIS land band at 1-km spatial resolution (Schaaf et al., 2002). The day of the year (DOY) for greenup onset was identified by fitting the 2001 annual trajectory of the MODIS-enhanced vegetation-index data (EVI; Huete et al., 2002) collected over 16-day periods to a sigmoidal vegetation growth model (Zhang et al., 2003). EVI values were calculated from MODIS nadir BRDF-adjusted reflectance with 1-km spatial resolution as described in Zhang et al. (2003).
The map of the annual potential evapotranspiration for Spain in 2001 was produced from the FAO modification of the Penman-Monteith equation at a 0.2° spatial resolution by El Instituto Nacional de Meteorología de España.
Results and Discussion
In the Mediterranean forest experiment, where the drought treatment produced an average of 15% decrease in soil moisture and the second year of study had a longer summer dry period (rains started in October instead of September) (Ogaya & Peñuelas, 2003), we found annual and species-specific responses to our artificially provoked climate changes. Annual variation was appreciated in both species in flowering-time (delayed in the second year of the study in Arbutus unedo) and in the flowering-amount (lower in the second year of the study in Phillyrea latifolia), as well as in the effects of the drought treatment (more notable in the second year in Arbutus unedo) (Fig. 1a). Nevertheless, although rainfall exclusion affected Arbutus unedo (delayed flowering and fewer flowers), Phillyrea latifolia remained unaffected (Fig. 1a).
In the Mediterranean shrubland experiment we found that the drier conditions had different effects on the flowering phenologies of Globularia alypum and Erica multiflora, the two dominant coastal shrubs we studied. Whereas in G. alypum drought delayed the flowering time and decreased the number of flowering plants, in E. multiflora, drought delayed the onset of flowering only in the second year of study (Fig. 1b), a wetter year (511 mm) than the first one (370 mm). While drought reduced the length of the flowering period in G. alypum, it extended it in E. multiflora in the first year of study (Fig. 1b). These noticeable effects on flowering span and flower production show that rainfall and the corresponding water availability are crucial for the flowering of these Mediterranean species and also that, in addition to the species-specificity of the response to changes in rainfall, there are seasonal and annual specificities that depend on temperature and the amount and distribution of rainfall throughout the season and year in question.
These species-specific responses in the timing and amount of flowering produced different reproductive responses in each of the studied species. In the rainfall exclusion experiment conducted in the Mediterranean forest, whereas the decreased flowering of Arbutus unedo (Fig. 1a) was followed by a clear trend towards fewer fruits (Fig. 2a), the absence of changes in the flowering of Phillyrea latifolia was followed likewise by no differences in the amount of fruits (Figs 1a and 2a). In the rainfall exclusion experiment in the Mediterranean shrubland, we hypothesized that the delays in flowering found for G. alypum might result in a reduction in seed numbers and seedling recruitment as a consequence of a reduction in pollination due to a scarcity of pollinators late in the season (Santandreu & Lloret, 1999). Indeed, a reduction in recruitment seedling was found for both studied species (Fig. 2b). This reduction might also occur due to injuries to flowers, fruits and seeds caused by frost in these autumn-winter flowering species. Rainfall changes may also affect final seedling recruitment through direct effects on germination and survival. Whichever the cause, we found a clear decrease in seedling recruitment in both drought plots; this was especially significant in G. alypum (Fig. 2b), the species which showed the most significant changes in the timing and amount of flowering (Fig. 1b). Since the responses to rainfall in species-specific seedling recruitment determine the pattern of species’ relative abundance in the community, our results suggest that new dominance patterns may appear as a consequence of these rainfall-mediated phenological changes.
Similar conclusions on the importance of rainfall and on the species- and annual-specificity of responses were obtained when we conducted a study of a complete phenological data set from a nearby site in central Catalonia (Cardedeu field station, 41°34′-N, 2°21′-E). During the period 1952–2000, precipitation significantly affected the timing of different phenophases and accounted for a significant amount of the phenological variation detected in several species. However, as previously seen for the effect of warming on this same phenological dataset (Peñuelas et al., 2002), these rainfall responses were strongly species-specific. Figure 3 shows the variable responses to rainfall found in this time-series data set. Most species advanced leaf-unfolding and flowering after increased precipitation throughout the previous months. For example, rainfall accounted for up to 48% of the leaf-unfolding variance in Ulmus minor, a species that advanced its leaf-unfolding by almost one week for every 100 mm of extra precipitation falling over the previous 5 months. There were, however, some species which showed no significant changes in response to higher rainfalls (Fig. 3), and one species (Vicia faba) even significantly delayed flowering after extra precipitation (Fig. 3). For a few species, rainfall accounted for an even greater part than temperature of the variance in flowering (3 species out of 57) or fruiting (10 species out of 27). The species which showed the most significant impact caused by increased precipitation were the least drought-tolerant species (e.g. Ulmus minor for leaf phenology) or agricultural plants without irrigation (e.g. Hordeum vulgare, for flowering; Zea mays, Prunus dulcis and Avena sativa for fruiting). The influence of precipitation on leaf-unfolding date for overall species was weaker than that of temperature (standardized regression coefficient of –0.15 vs –0.48), although it was still strongly significant (P < 0.0001). The overall influence on flowering dates, although also significant (P < 0.001), was weaker than that of temperature (standardized regression coefficient of –0.09 vs –0.29). On the contrary, the relative influence of precipitation on fruiting was stronger than that of temperature (standardized regression coefficient of –0.20 vs –0.14).
In a nearby locality, 50 km north of the Garraf shrubland experimental site, Santandreu & Lloret (1999) reported a much shorter flowering period for Erica multiflora than the one we describe here (Fig. 1b). In spite of its proximity, this locality is colder and wetter: 10.2°C mean annual temperature and 928 mm mean annual precipitation, against the 13.8°C and 450 mm of the Garraf site. These differences for the same species led us to consider another aspect of these complex phenological responses to water availability: the important geographical variation in phenology caused by the great local and regional variability of climatological and microclimatological conditions.
In order to tackle this local and regional variability and, even more importantly, in order to approach the possibility of nongradual geographical phenological shifts, we assessed greenup onset for the whole of the Iberian Peninsula. The phenological map of greenup onset obtained from MODIS 16-d period coverage shows a variable range of greenup onset dates from north to south and from west to east, with an abrupt ‘phenotonic’ change in central Spain and along the eastern coastline. In wet and cool northern areas (lower water dependence and more rainfall, Fig. 4b), greenup occurs in spring (Fig. 4a), triggered by rising temperatures. On the other hand, in the dry and warm south and eastern coastal areas (higher water dependence and less rainfall, Fig. 4b), greenup occurs in autumn (Fig. 4a), triggered by the arrival of autumn rainfall. The abrupt ‘phenotonic’ limit between the northern spring greenup and the southern and eastern autumn greenup corresponds to the existence of mountain ranges. Mountain ranges in the south still have greenup onset in spring (Fig. 4a). The map of potential evapotranspiration (Fig. 4b), which can be interpreted as a proxy indicator of water dependence, also separates the cool northern regions from the warm southern regions. This contrast in phenological responses to rainfall between cool-wet and warm-dry regions seem general throughout the globe and confirms previous models that consider leaf onset as mainly dependent on temperature at temperate and higher latitudes, but controlled by water availability at lower latitudes (Botta et al., 2000).
This MODIS-generated greenup onset image (Fig. 4a) also shows a complex spatio-temporal variation that comes about not only through climatic variability but also through variations in community composition, soils and land management. For example, greenup in some north-eastern and southern areas at the end of the year (Fig. 4a) is explained by agricultural practices (irrigation and cereal crops). The greening-up of rice fields of the Ebro Delta (NE in the map) at the same time as the high peaks of the Pyrenees (France border) is a paradigm of these irrigation effects. Moreover, some vegetation types exhibit bimodal and even multiple modes of growth and senescence within a single annual cycle, thereby making the regional phenological pattern even more complex. In all cases, and although this remote-sensing data is of much coarser resolution than the phenologically strongly species-specific ‘on-the-ground’ small-scale observation data, the greenup onset image shows geographically and ecologically coherent patterns that are consistent with the known phenological behavior of these Iberian regions.
These complex species-, seasonal-, annual-, local- and regional-specific phenological responses to changes in rainfall are very likely to lead to long-term asynchronies in intraspecific and interspecific interactions prompted by the changes in rainfall predicted by Global Climate Models (IPCC, 2001), just as has been described for responses to warming (Peñuelas & Filella, 2001). Potentially, population and evolutionary dynamics, community structure and ecosystem functioning will be altered. The complexity of these water-mediated phenological changes, which include possible nongradual shifts, complicates the predictions and the modeling of their ecological responses at different spatial and temporal scales. Nevertheless, together these results highlight the need to consider changes in rainfall and water availability as another important driver of climate change that may lead to significant phenological changes and subsequent changes in ecosystems and the functioning and structure of the biosphere.
This research was supported by EU Environment-program CLIMOOR and VULCAN (EVK2-CT-2000-00094) grants and by Spanish MCYT (REN 2000-0278, REN 2001–0003, and REN2003-04871) grants. We would like to thank Antonio Mestre for his technical assistance and the Instituto Nacional de Meteorología from Spain for the data and the maps of potential evapotranspiration and precipitation from 2001. We also want to thank Prof. F. Rodà for his helpful insights.