Ecophysiological adjustments of a pine forest to enhance early spring activity in hot and dry climate

Climate change can impose large offsets between the seasonal cycle of photosynthesis and that in solar radiation and temperature which drive it. Ecophysiological adjustments to such offsets in forests growing under hot and dry conditions are critical for maintaining carbon uptake and survival. Here, we investigate the adjustments that underlie the unusually short and intense early spring productive season, under suboptimal radiation and temperature conditions in a semi-arid pine forest. We used eddy covariance flux, meteorological, and close-range sensing measurements, together with leaf chlorophyll content over four years in a semi-arid pine forest to identify the canopy-scale ecophysiological adjustments to the short active season, and long seasonal drought. The results reveal a range of processes that intricately converge to support the early spring peak (March) in photosynthetic activity, including peaks in light use efficiency, leaf chlorophyll content, increase in the absorption of solar radiation, and high leaf scattering properties (indicating optimizing leaf orientation). These canopy-scale adjustments exploit the tradeoffs between the yet increasing temperature and solar radiation, but the concurrently rapidly diminishing soil moisture. In contrast, during the long dry stressful period with rapidly declining photosynthesis under high and potentially damaging solar radiation, physiological photoprotection was conferred by strongly relaxing the early spring adjustments. The results provide evidence for canopy-scale ecophysiological adjustments, detectable by spectral measurements, that support the survival and productivity of a pine forest under the hot and dry conditions, which may apply to large areas in the Mediterranean and other regions in the next few decades due to the current warming and drying trends.


Introduction
The seasonality of ecosystem photosynthetic activity is characterized by the time of peak activity and reflects the integration of plant response to variations in temperature, radiation, and water availability (Garonna et al 2018, Park et al 2016, Xia et al 2015, Yang et al 2019. Such analyses indicate, in turn, that the timing of peak photosynthesis can display offset of days to months relative to the seasonality in the physical climate parameters, such as solar radiation and temperature. For example, the peak in gross primary productivity (GPP) can be delayed by ∼45 d with respect to the seasonal peak in radiation, or precede it by up to ∼75 d in the transitions from the arctic to the temperate climate zones (Park et al 2019). Another study found that peak GPP in pine forests advanced by 7 d for each increase of 10 W m −2 in annual mean radiation across Europe (Rotenberg and Yakir 2010). Consequently, while peak solar radiation in the Northern Hemisphere is around the day of year 172 (DOY; June 21), peak GPP in northern Finland was around DOY 205 (delayed by low temperatures) and around DOY 85 in Israel (advanced due to rapid drying). Similar shifts in seasonality were also associated with biome type, with peak activity in early May in savannas and mid-July in crops (Xia et al 2015). It was recently estimated that global warming results in gradual advance in the photosynthesis seasonality estimated at 1.66 d decade −1 across the Northern Hemisphere (Park et al 2019). Such trends are likely associated with a considerable increase in the relative influence of moisture constraint in determining the timing of peak photosynthesis (Garonna et al 2018).
The changes in the seasonality of photosynthesis can enhance forest productivity in some cases (Menzel et al 2006, Keenan et al 2014, Park et al 2019, but exceeds the ecosystem adjustment capabilities in other cases resulting in enhanced mortality (Allen et al 2015, Klein 2015, Anderegg et al 2019. These studies provide information on the spatial patterns of plant GPP and tree mortality. Much less information is available, however, on the ecophysiological processes that underlie the plant adjustments to changes in these forcing (McDowell et al 2008, Anderegg et al 2019. This is particularly so in the dry climates, including the Mediterranean and the semiarid regions where ecophysiological adjustments are key to survival (Peñuelas et  Hot and dry regions, such as the sub-humid Mediterranean, are characterized by short mild winters and rainfall periods followed by rapidly drying soils and long dry summers with high radiation intensity (Joffre andRambal 2001, Deitch et al 2017). These strong environmental forcing result in a large shift in the time of peak activity from July-August in temperate forests to March in semi-arid Mediterranean forests (Rotenberg and Yakir 2010). Furthermore, the peak productivity season (period with GPP within 15% of the peak level) can also be as short as 1.5 months (Helman et al 2017) and subject to suboptimal conditions because of the limitation by low temperature and radiation at one end of the productive season, and the rapidly drying soil at the other end. However, detailed information is still scarce on ecophysiological adjustments that enable the exploitation of such short and time-shifted productive period and, in turn, sustain the long stress period, especially at the canopy and ecosystem scales.
Physiological stress can be enhanced during drought periods due, for example, to stomatal closure, reduced ET, and CO 2 exchange, together with the excessive incoming radiation, and can result in photodamage, carbon starvation, and hydraulic failure (Ogaya and Peñuelas 2003, Asensio et al 2007, McDowell et al 2008, Sala et al 2012, Klein et al 2014, Rowland et al 2015. At the leaf scale, studies show a range of strategies that can help in dealing with the excessive absorbed light (Ruban 2009, Cazzaniga et al 2013, Xu et al 2015. This includes a reduction in chlorophyll and an increase in carotenes and xanthophyll concentrations that reduce the solar radiation channeled to photosynthesis (Demmig-Adams andAdams 2006, Scartazza et al 2016) or enhanced photorespiration (Wingler et al 2000, Voss et al 2013, Eisenhut et al 2017. A recent study on pine needles in a semi-arid forest showed that a combination of reduced radiation absorption, a relative increase in photorespiration, and enhanced xanthophyll cycle, conferred resilience and full recovery from stress on the diurnal time scale (Maseyk et al 2019). Some adjustments to drought are also identified at the tree scale, such as the response to the combined effects of drought and increased vapor pressure deficit (Grossiord et al 2017), and a shift in carbon allocation from stem to root growth, together with tighter stomata and hydraulic functioning (Klein et al 2011, Fernández-De-Uña et al 2017. At the canopy scale information often relies on eddy covariance (EC) flux measurements of CO 2 , ET, and energy (Baldocchi 2003). Some EC-based studies have demonstrated that forests adjustment to longterm drought by shifting to efficient canopy cooling through enhanced sensible heat flux, substituting for the evaporative cooling common in temperate forests where soil water is available . Evaluating a wide range of EC sites in the semiarid regions of North America demonstrated a predictable link between variations in net ecosystem productivity (NEP) and the actual water availability for ET (Biederman et al 2016). Another study in these regions identified a precipitation threshold where NEP switches from carbon sink to source (Liu et al 2018). Canopy-scale EC measurements also demonstrated high ecosystem resilience to short-term heatwaves superimposed on the seasonal drought stress in dry forest sites (Tatarinov et al 2016).
Remote sensing (RS) could extend and supplement ecosystem-scale studies (Kerr and Ostrovsky 2003, Mulla 2013, Lausch et al 2016. For example, the chlorophyll/carotenoid index (CCI) showed an increase in the fraction of carotenoids in leaf pigments in an evergreen forest as a biochemical response to the drought conditions. Another study about the ecosystem response to heatwaves used sun-induced fluorescence (SIF) measurements and showed stomatal closure in the early stages, intrinsic damage to the photosynthetic system at the height of the stress, and rapid recovery following the event (Wohlfahrt et al 2018). Remote sensing includes satellite observations, aerial or ground level (close-range sensing) measurements at different temporal and spatial scales (White et al 2016, Toth andJóźków 2016). While satellite remote sensing is widely used it is often associated with limited spatial and temporal resolution and with caveats such as partitioning vegetation and soil, which is particularly relevant in sparse dryland vegetation. Close-range sensing from flux towers, drones, or mobile masts provides the opportunity of obtaining direct and continuous canopy scale spectral information at high spatial resolution (Tittebrand et al 2009, Clasen et al 2015, Mikita et al 2016. The limitations on forest productivity and survival associated with hot and dry climates are particularly significant in light of the predictions of warming and drying trend and increasing tree mortality in the Mediterranean and other regions ( , in other cases trees and ecosystems can acquire ecophysiological adjustments to cope with the harsh conditions. Here, we use tower-based sensing and flux measurement methodologies to address the questions: (1) What are the ecophysiological adjustments that support forest productivity in the early spring despite the decoupling from the seasonal cycle in solar radiation and operation in suboptimal temperatures; (2) what are the canopy scale adjustments that provide photoprotection when productivity is suppressed by drought while radiation increases. We hypothesize that seasonal changes in the light absorption capability and light use efficiency, which can enhance productivity in the low radiation and temperature but high soil moisture season, and their relaxations that provide photoprotection in the hot and dry period.

Study site and period
This work was conducted in Yatir forest (31 • 20 ′ N; 35 • 30 ′ E; 600-900 m above sea level; 2800 ha), which is located at the northern edge of the Negev desert in southern Israel. The forest is dominated by Aleppo pine (Pinus halepensis Mill.), with smaller proportions of other pine species and cypress and little understory vegetation. Stand density is about 300 trees per hectare, mean tree height is 11 m and Leaf Area Index (LAI) is about 1.70 (Sprintsin et al 2011). The native background vegetation is sparse shrubs with a total vegetation height of 0.30-0.50 m (Grünzweig et al 2003). The annual mean solar radiation load is 238 Wm −2 . The annual mean air temperature is 17.7 ± 0.5 • C. Mean air temperature in January (the coldest month) and July (the hottest month) is 10 • C and 25 • C, respectively. The annual mean precipitation is 279 ± 90 mm.

EC and meteorological measurements
A 19 m high flux tower at the center of the Yatir forest is operated since 2000, continuously measuring fluxes of CO 2 (net ecosystem exchange, NEE), water vapor (evapotranspiration, ET), as well as other meteorological parameters such as incoming photosynthetic active radiation (PAR in ; 400-700 nm), air temperature, vapor pressure deficit (VPD) and soil water content (SWC) Yakir 2010, Tatarinov et al 2016), using the Euroflux methodology and data quality control (Aubinet et al 1999). The ecosystem gross primary production (GPP) was calculated from NEE measurements after correction for nighttime respiration (Tatarinov et al 2016).

Reflectance measurements
Since May 2012 two close-range radiometers (SKR 1850, Skye Instruments LTD, U.K.) were installed on the flux tower, facing the atmosphere and the ecosystem. With A diffuse cap, the sensors had a 90 • viewing angle, which captured ∼90% of the incoming radiation and allowed focusing on the canopy with the minimal signal from the soil. The sensors measured downwelling and upwelling irradiance in three visible (530 ± 11.5 nm, 570 ± 10.1 nm, 659.4 ± 11.3 nm) and one NIR (858 ± 10.7 nm) bands and the data were integrated over 30 min intervals to be compatible with flux measurements. The canopy reflectance ρ, was calculated as the ratio of the upwelling to downwelling irradiance at each spectral band. Based on the reflectance measurements, we used two approaches to estimate the fraction of absorbed PAR (fAPAR) and, in-turn, the actual absorbed photosynthetic radiation as APAR = fAPAR × PAR in .
Firstly, the widely used Normalized Difference Vegetation Index (NDVI) was calculated as (Rouse et al 1974): where ρ NIR and ρ red refer to the reflectance at spectral bands centered at 858 nm and 659 nm, respectively. NDVI is strongly correlated with leaf area, which is one of the main factors influencing canopy light absorption, and is linearly correlated with fAPAR (Ruimy et al 1994, Fensholt et al 2004. Therefore, APAR could be estimated as: Light use efficiency (LUE) was derived from APAR NDVI and GPP (Monteith 1972): Secondly, the canopy chlorophyll absorption coefficient in the green range was calculated as (Gitelson et al 2019): where ρ green is the reflectance in the spectral band centered at 570 nm. Since α chl indicates the ability of light absorption by chlorophyll, its product with PAR in could be used to estimate APAR: Note that NDVI and α chl are both empirical indices and we used the more common approach to obtain NDVI-based LUE (equation (3)) and compared to published values. But we took advantage of the higher sensitivity of the newer index α chl (figures 1(d), S2 (available online at https://stacks.iop.org/ERL/15/114054/mmedia)) and its direct relationships to chlorophyll absorption (Gitelson et al 2019) to follow the seasonal variation and better identify the timing of its seasonal peaks and minima.
Leaf reflectance at the NIR 780 nm (775-785 nm) and red-edge 742 nm (736-747 nm) bands were measured monthly by Polypen UVIS spectrometer (PSI spol. s r.o., Drásov, Czech Republic) during 2018-2019. Leaf chlorophyll content (LCC) was estimated non-destructively from the red-edge chlorophyll index (CI re = ρ NIR /ρ re −1) (Gitelson et al 2003, Ciganda et al 2009, after calibration with two sets of needles sampled on April 23 and May 30, 2018. During the calibration, the LCC of the two sets of the sampled needles was measured analytically. For each sample, 1 cm 2 needle tissue (about 50 mg) were immersed in 1 ml of 80% acetone, then ground using a ball mill (two 3 mm balls; Retsch, Hann, Germany) at a frequency of 25 s −1 for about 5 min, and followed by centrifuge (5424 R, Eppendorf AG, Hamburg, Germany) at 14 000 rpm for 10 min. The supernatant was transferred into a 96 well plate (300 µl), where its absorbance was measured at 663.6 nm and 646.6 nm by a microplate reader (Molecular Devices, Sunnyvale, CA, USA). After subtraction of the absorbance of a blank sample, the concentration of chlorophyll a + b was calculated (Porra et al 1989).
The detailed relationship between the CI re and the analytical LCC can be found in supplementary materials (figure S1), from which LCC can be estimated as: With determination coefficient R 2 = 0.92 and RMSE of 3.7 µg cm −2 .

Evaluation of seasonal trends and statistical analysis
Canopy scale fluxes of CO 2 were computed from the high-resolution EC measurements using EddyPro software 7.0.5 (LICOR Biosciences, Lincoln, Nebraska) and averaged over 30 min interval.
All canopy flux measurements were processed at a daily scale in two steps: Firstly, to eliminate the bias from the diurnal variations, the average of the daytime measurements were obtained. Secondly, a moving average was applied to the multiannual set of daily scale data to identify season peaks and minima. A 7 d running average provided the optimal representation of the seasonal changes.
To evaluate the seasonal cycle in the radiation and CO 2 uptake (GPP) data, daily sum (daytime only) of the PAR in , APAR, and GPP data were used.
The period of peak productivity was determined, for each year, as the period during which photosynthetic CO 2 uptake (GPP) was above 85% of its annual maximum values. The same threshold was also used to define the peak period for other variables.
Statistical analysis and visualization were carried out using R software (version R-3.4.1) and the 'ggplot2' package.

Results
In the results below we report on the seasonal offset between the peak in photosynthesis and its main driving forces, such as radiation and temperature. Afterward, we proceed to report on the factors that could help sustain high GPP despite such offset (including chlorophyll content and light absorption efficiency).
The seasonal variations in the meteorological and ecophysiological parameters for the study period 2013-2016 are reported in figure 1. In general, more than one distinct seasonal cycles are observed. One cycle includes the meteorological parameters dominated by solar radiation, such as PAR in , APAR, temperature, and VPD, peaking in July, with a minimum in January. In contrast, the ecophysiological parameters showed clear offsets to the climatic parameters. SWC, NDVI, α chl , ρ NIR reached a peak in late winter (January-February), and decreased with the onset of the dry period. GPP and LUE peaked together about one month later in March, while ρ red was at its minimum. The productive period where GPP was higher than 85% of its peak values, was only about 60 d. Precipitation occurred only from November to April, with annual precipitation of 323 mm, 358 mm, and 352 mm for the three hydrological years (1st October-31st September, 2013-2016). Notably, GPP and LUE were strongly coupled and were characterized by a sharp and narrow peak at the end of March, followed by a rapid decrease, from 0.5 mol m −2 d −1 at peak season to 0.05 mol m −2 d −1 around July (i.e. a 10-fold decrease), with the low values remaining through the long dry season.
Chlorophyll absorption coefficient α chl showed a seasonal cycle synchronously with GPP and with similarly narrow peak, but about 1-2 months in advance ( figure 2(a)). Maximal α chl occurred around February while maximal GPP around the end of March. Such offset was less clear during the time of the minima in these parameters at the end of the dry season in October.
The APARα chl , [α chl × PAR in ] was tightly coupled with GPP during the December-March buildup period leading to the seasonal maximum in GPP, but with a time delay and slower decrease from around July to December ( figure 2(b)). The slope of GPP as a Canopy reflectance in NIR, ρ NIR , decreased with increasing incident irradiance PAR in showing strong negative linear correlation (R 2 = 0.74) across the seasonal cycle (figure 4). This relationship was particularly strong as PAR levels increased in summer.
LCC showed a clear seasonal cycle with the minimal value in October (15 µg cm −2 ) and the peak in March-April (45 µg cm −2 ) (figure 5).

Discussion
The results reported above demonstrate two main components: First, the result of the photosynthetic CO 2 uptake fluxes (GPP) show the unusually early (March) and short (>60 d) peak activity period in the study ecosystem, as compared with the commonly observed peak time (June to August) and length (often >120 d) in temperate forests where it coincides with high solar radiation and temperature. This represents a shift from radiation-dominated to moisturedominated ecosystems along the latitudinal and climatic gradient Yakir 2010, Park et al 2019). Second, the additional measurements provide evidence that while the early productivity period under suboptimal conditions is imposed by the water limitations, it is facilitated by synchronizing multiple factors, such as increased chlorophyll content, enhanced light use efficiency, and adjusted leaf inclination at just the right time.
We focus on canopy-scale adjustments, which address two key requirements: Optimizing productivity in the unusually narrow productive window, when soil moisture is still sufficient and radiation and temperature are least limiting, and providing photoprotection in the long dry and hot summer. The adjustments to these contrasting requirements are clearly reflected in the sequence of the seasonal cycles of key ecophysiological parameters as summarized in figure 6.
Optimizing offseason productivity (the Goldilocks effect): The unusually narrow peak in photosynthetic activity (GPP) in spring (figure 1(e)) seems to reflect environmental and ecophysiological evolution: (1) As the limitation of soil moisture is reduced at the end of winter, photosynthesis (GPP) increases synchronously with the increase in temperature and incoming solar radiation (PAR in ) into the early spring.
(2) While PAR in and temperature continue to increase into the early summer, the rapid soil drying and high evaporation demands (VPD), quickly suppress the photosynthetic activities. This results in the observed narrow productivity peak in March (figure 2(a)). (3) To exploit this imposed narrow window of opportunities to optimize productivity, the results show that a range of ecophysiological resources are mobilized and are well synchronized to form what can be termed a 'Goldilocks effect' that underlies the high productivity of the dry forest.
One of the distinct aspects that supports the observed GPP seasonality is the temporal change in light use efficiency (LUE; figure 1(e)). While in temperate sites the highest LUE coincides with the lowest GPP in winter (e.g. Garbulsky et al 2008), here it coincides with the maximum GPP in spring. This is followed by a decrease in LUE to low values in the dry season that seems to reflect downregulation of biochemical activities, associated with the near-zero stomatal conductance (Maseyk et al 2008(Maseyk et al , 2019. Irrespective of the changes in the seasonal patterns, the magnitude of the observed LUE values (ranging between 0.002 and 0.02) was consistent with published estimates from different biomes (Garbulsky et al 2011).
Leaf development in Yatir has previously been shown to begin in spring and occur throughout the  dry and stressful summer, allowing to maximize the contribution of the new fully developed leaves, with peak concentrations of leaf nitrogen (Maseyk et al 2008), to maximize GPP in the following short active season. The seasonality of leaf development was composed of two components. The first was reflected in the seasonal patterns of LAI that reached a maximum already in the fall (Sprintsin et al 2011). However, as LAI reached its seasonal maximum, LCC was still at its minimum and only then started to increase, reaching a seasonal maximum in spring (∼3-fold increase; figure 5), such that the combined effects of high LAI times LCC coincided with the peak of photosynthetic activity.
LAI and LCC, together with structural effects influence the canopy absorption coefficient (Gitelson et al 2019), which was estimated here as α chl . We observed temporally synchronous behavior of GPP and α chl but with the increase in α chl leading GPP by 1-2 months ( figure 2(a)). Note that although NDVI

Figure 6.
A conceptual presentation of the sequence of key seasonal cycles in (a) soil moisture, the rate of photosynthetic activity, solar radiation, and temperature; and in ecophysiological adjustments in (b) light absorption coefficient, light use efficiency, leaf chlorophyll content, and canopy NIR reflectance. It highlights the large offset (several months) between photosynthetic activities and its driving forces (light and temperature), which is imposed by the water availability (soil moisture). The ecosystem's ability to adjust the timing of peak photosynthesis requires the convergence of a range of ecophysiological adjustments (termed here the Goldilocks effect). The lines are a fit to mean data at the study site. The table indicates the peak timing of the above and additional parameters, which are discussed in more detail in the text.
also provides an indicator of canopy absorbing (equation (2)), α chl was found to offer tighter relationships with the seasonal changes in GPP (figures S2(a), (b)). This temporal offset is, in fact, advantageous in ecophysiological terms as it enhances canopy absorption, and therefore photosynthetic potential, in late winter when SWC is at maximum but incident irradiance is limiting. The importance of this early adjustment is clearly evident when the product of α chl and PAR in is examined ( figure 2(b)).
Indeed, the product of α chl and PAR in , which provided an intrinsic measure of APAR in in the system showed a tight coupling with GPP during the early increasing phase in late winter to early spring ( figure 2(b)). The linear GPP vs. [α chl × PAR in ] relationship is consistent with the Monteith model which assumes that productivity must be positively dependent on absorbed radiation (Monteith 1972), as has been observed in other ecosystems (Turner et al 2003, Lagergren et al 2005, Garbulsky et al 2011. The seasonal-scale slope of the linear regressions indicated the seasonal mean LUE (figure 3). The lower LUE observed in winter 2013-2014, was associated also with lower maximum values of GPP and [α chl × PAR in ]. This is consistent with the lower annual precipitation (30 mm, and over 10% less than other study years), and lower minimum SWC at the end of winter in that year compared with the following two years (17% vs 20%), providing confidence in our APAR in and LUE estimates.
Photoprotection under stress: While the importance of increasing α chl in the low PAR period is evident, the overall relationships of α chl and GPP are more complex (see figure S2). The strong correlation between these parameters noted above also indicated a strong seasonal hysteresis with both low and high GPP values associated with similar α chl before or after the March GPP peak, respectively. It seems that variations in α chl are part of the canopy-scale adjustments to seasonal changes in ambient conditions, enhancing absorption when PAR in is low, but depressing absorption and providing some photoprotection when PAR in is high and GPP is in decline as the summer drought develops.
Structural changes can also enhance absorption or provide photoprotection, and reflectance in the NIR range, ρ NIR , where chlorophylls do not absorb, roughly represents canopy scattering coefficient that related to canopy structural properties such as LAI, and leaf inclination angle (Gitelson et al 2019). This notion is supported by the strong negative correlation between PAR in and canopy ρ NIR (figure 4), with ρ NIR peaking in winter during the lowest annual PAR in and decreasing to a minimum in summer during the highest PAR in . These changes do not coincide with those in LAI discussed above (Sprintsin et al 2011), and seasonal changes in solar angles are expected to have a small effect (Hinojo-Hinojo and Goulden 2020). We, therefore, suggest that the observed seasonal changes in ρ NIR are related to some changes in leaf orientation to a more horizontal angle in winter to maximize absorption, resulting in enhanced NIR scattering despite low PAR in at that time. In contrast, more vertical leaf orientation in summer reduces absorption and ρ NIR and, in turn, reduces over-excitation and photodamage during the high PAR in -low GPP period.
Finally, we note that besides the structural changes, the large reduction in LCC observed in the summer (figure 5) provides additional photoprotection by reducing α chl and minimize excessive light absorption. This is supported by canopy reflectance in the visible (red) range that was about 1.5-fold higher during the dry period ( figure 1(e)). Additional photoprotection mechanisms at the leaf-scale at the same site have also been recently reported (Maseyk et al 2019).
The results discussed above have important implications for the future of forests in large parts of the world. This is because evolutionary-based genetic modifications are too slow to adapt to the rate of the warming and drying underway in many regions. In contrast, the phenotypic plasticity we report here in common Mediterranean pine trees, clearly shows that such trees can make large adjustments to offseason productivity and provide photoprotection under stress, to offer an effective basis for forest survival and continued productivity under increasingly stressful environments.

Conclusion
As global climate change is expected to expose large land areas to increasing hot and dry conditions, our results provide evidence for canopy-scale ecophysiological adjustments that boost the resilience and productivity of pine forests under harsh conditions. These adjustments reflect a combination of avoidance by a shift in the timing of activity, balanced by enhancing productivity under the sub-optimal temperature and PAR conditions through optimizing chlorophyll content, LAI, and LUE.