Temporal trends in the enhanced vegetation index and spring weather predict seed production in Mediterranean oaks

The extremely year-to-year variable production of seeds (masting) is an extended plant reproductive behaviour important for forest dynamics and food webs. The dependence of these episodes of massive seed production on recently or long-term photosynthesised carbohydrates, however, remains controversial. In this paper, we explore whether vegetation (tree canopy) changes, detected using EVI as a proxy of leaf area and photosynthetic capacity, can provide a reliable estimation of seed production. To complete this analysis, we also explored the effect of weather both in the trends of EVI and in acorn crop size. To this end, we compared the trends of the EVI and acorn production over 10 years (2000–2009) in five stands of Quercus ilex L. in Barcelona (Catalonia, NE Spain). We found that acorn production was mainly driven by a combination of: (i) a minimum initial threshold in the EVI values, (ii) an increase in EVI in the 9 ± 4 months prior to reproduction, and (iii) appropriate weather conditions (low water stress) during spring. These results indicated, apparently for the first time, that reproduction in masting species could be detected and partly predicted by remotely sensed vegetative indices. Our results suggested that this particular reproductive behaviour in Mediterranean oaks was driven by a combination of two factors, i.e. good and improving vegetation conditions, as shown by a minimum initial threshold and the increase in EVI needed for large seed crops, and the need of wet weather conditions during spring. Moreover, our results fully supported recent studies that have associated short-term photosynthate production with seed production.


51
The synchronous but erratic year-to-year production of seeds, i.e. masting, is a widely geographically and 52 taxonomically extended reproductive behaviour (Kelly and Sork 2002). Masting events have cascading effects 53 on several forest functions and processes (Ostfeld and Keesing 2000) such as seedling establishment (Negi et al.

71
Recent studies based on carbon isotopes suggest that seed production in several tree species in temperate 72 forests depends only on the carbohydrates photosynthesised during the months prior to seeding (Ichie et al. remotely sensed vegetation indices to predict seed crop production in forest ecosystems, particularly in masting 82 species (but see Camarero et al., 2010). In this sense, a successful methodology to predict seed crops would 83 allow managers to better plan management actions in the near future (e.g., wildlife conservation strategies).

84
The aim of this study was to explore whether the pattern of seed production in Mediterranean oak (Quercus 85 spp.) forests could be governed by the temporal dynamics of tree canopies, assessed by means of remote 86 sensing indices. We hypothesise that vegetation changes, detected using EVI, represent a reliable proxy of the 87 tree resources available for seed production. Therefore, if an accumulation of resources (e.g. carbohydrates and 88 nutrients) for several years is a prerequisite for a masting event, EVI should progressively increase during that 89 period of accumulation. Similarly, if a severe depletion of resources occurs after a masting episode, EVI should 90 decrease due to the self-thinning of the canopy. In addition, given the nature of the Mediterranean climate, we 91 hypothesise that weather conditions (e.g. drought) must be an important driver of the temporal dynamics of both 92 EVI and acorn crop size. To address these questions, we used a data set comprising 12 years (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)   We performed a two-step analysis to determine the relationships among weather, the EVI, and acorn 177 production. First, we explored the relationships of seasonally averaged SPEI and EVI time series with acorn 178 production to identify the season when SPEI has the largest effect on EVI. Then, we tested the correlation 179 between acorn production and the seasonally averaged time series of SPEI and EVI. We used Spearman 180 correlations for this procedure because the acorn-production data were not normally distributed. Second, we 181 used the smoothed EVI time series to test whether tree resources increased before large seed crops (i.e. masting 182 events) by looking for evidence of an increasing EVI prior to masting. We identified all changes in the trends 183 (e.g. from increasing to decreasing EVI) of the smoothed time series and considered an EVI trend to be the 184 period between two of these points of change (peaks or valleys). We chose not to calculate trends for fixed

191
To test the influence of weather, the EVI, and previous acorn harvests on acorn production, we constructed 192 a generalised linear mixed model (GLMM) fitted using the negative binomial distribution, using the natural the acorn production of the previous year as covariates. We also included the interaction of ΔEVI with EVI i and 196 EVI f to test for possible synergic effects of an increasing EVI and the initial or final level reached in the EVI.   (Table 1). No significant differences 218 were found for r p , r i , CV p , CV i , and temporal autocorrelation among plots (using bootstrapped standard errors of 219 the means). The average synchrony (r p ) in the mean annual EVI was 0.68 ± 0.03 (Table 1). The CVs of the EVI 220 values were very low (~0.05) and did not differ from zero (were not significant) due to error propagation, 221 although the r values were quite high. correlated with low and above-average EVI values, respectively (Figure 1). The correlation analysis confirmed 226 that the EVI could be driven by the cumulative effect of weather conditions during the previous nine months (r 227 = 0.50, P < 0.001; Table 3). Additionally, acorn crops were larger after wetter seasons and when they were 228 preceded by larger increases in the EVI during the previous months (Figures 1 and 2). Acorn crop sizes, 229 however, were still very low when these periods of increasing EVI began from very low EVI values and did not 230 reach above-average values (Figures 1 and 2).

231
The selected periods of increasing or decreasing EVI values (Figure 2, shaded periods)  followed an average period of increase in the EVI (a surrogate of potential photosynthetic capacity) of 9 ± 4 months but also required favourable weather conditions in spring (low levels of water stress, SPEI6 Jn ). In 255 addition to these two factors, the initial EVI value at the start of an EVI increase period was also important, i.e. 256 a large increase in the EVI did not lead to a masting event unless it departed from a minimum EVI i threshold 257 (Table 2; Figures 1, 2, and 3). This initial EVI threshold may suggest the need of a minimum amount of 258 previously stored resources to produce buds before a mast seeding event, while the importance of the seasonal  (Table 1, lags 1, 3, and 4) also support the premise that large crop sizes are driven by endogenous cycles of 274 resource accumulation-depletion. When taking into account the EVI temporal dynamics and SPEI, however, the 275 negative autocorrelation at lag 1 was not significant (Table 2) role of weather found in this study, presenting a logarithmic relationship with acorn production, suggests that 315 non-linear relationships between weather and seed production might be the cause of these differences in the 316 distribution of seed production and weather data. For example, if seed production can be modelled as an 317 exponential function of rainfall, the variability in seed production will easily be much higher than the variability 318 in rainfall. Similarly, a sigmoid relationship between weather and seed production would lead that bimodality to  . 2007), but to our knowledge, this study is the first to demonstrate that seed 334 production in forest ecosystems can also be detected using satellite imagery and weather data. Nevertheless, the 335 study of species with different cycles of flowering, pollination, and maturation of seeds (e.g. red oaks that 336 require more than one year from the onset of flowering to seed maturation), may require a different approach.

337
Likewise, the study of deciduous forests may also require different methodologies, because these species may 338 respond differently. For example, a drastic reduction in crown cover was detected in two birch species (Betula     Table 1. Synchrony (r p among plots and r i among trees), variability at the population (CV p ) and individual (CV i ) levels, and temporal autocorrelation (simple and 519 partial (P)) of acorn production and EVI time series (mean ± SE). No significant differences were found among plots. Values different from zero are in bold type.