Drivers of chaparral photosynthetic rate reduction under modern drought conditions

Terrestrial vegetation communities are experiencing rapid and novel changes to photosynthetic rates under the changing climate. Chaparral, a semi-arid shrubland ecosystem of the Southwestern United States and Northern Mexico, is projected to experience substantial increases in aridity and stochastic precipitation. This study identifies the primary meteorological drivers of photosynthesis for three widespread chaparral shrub species—Adenostoma sparsifolium, Adenostoma fasciculatum, and Ceanothus perplexans—from 2019 to 2021. Monthly leaf-level carbon exchange rates, water potentials (WPs), and meteorological conditions were collected for each species. Average monthly primary productivity (n = 25) demonstrated vapor pressure deficit (VPD) as a significant limit to photosynthetic rates for A. sparsifolium and A. fasciculatum. VPD was also the most influential predictor of WP for all three species. These results suggest increasing atmospheric dryness as a key predictor for reduction in chaparral primary productivity, particularly for deeply-rooted, resprouting species. There are additional indications that VPD could exacerbate drought-related mortality for C. perplexans and A. sparsifolium by pushing WP to novel extremes. This study concludes that atmospheric dryness, across 3 years of differing soil water stress levels, was consistently a substantial physiological limitation for three common, chaparral species. Although this experiment occurred over a limited window and cannot assess climatic response trends, acute increases in air temperature and VPD within the region would exacerbate photosynthetic limitation for these species and may contribute to declining primary productivity in broader chaparral ecosystems.


Introduction
Variability in terrestrial-vegetation carbon uptake is one of the largest uncertainties of global carbon cycling models (Friedlingstein et al 2014, Tharammal et al 2019, Williams et al 2020).This uncertainty limits our ability to quantify global carbon dynamics and attribute the relative uptake potential of distinct vegetation communities.This is compounded for Mediterranean climates which have larger anticipated drought impacts under a changing climate (Loarie et al 2008, Kogan and Guo 2015, Polade et al 2017).In Mediterranean regions of the Southwestern United States and northern Mexico, chaparral is an evergreen shrubland community with a previously demonstrated capacity for significant carbon sink dynamics (Luo et al 2007).In California, chaparral is one of the dominant vegetation communities, covering 8%-10% of wildlands (Parker et al 2016), with related shrublands covering up to 41% (Malone et al 2016).These systems are at the highest state-wide risk of stochastic drought (Quero et al 2011, Underwood et al 2018).uptake and the lack of research on these communities is explicitly stated (CARB 2022 Scoping Plan for Achieving Carbon Neutrality 2022).In global terrestrial carbon uptake budgets, semi-arid ecosystems are the largest contributors to carbon uptake variability (Ahlström et al 2015).This variability is a product of the strong interannual carbon source/sink dynamics in semi-arid ecosystems, resulting from stochastic hydrological conditions and changing climate (Luo et al 2007, Ahlström et al 2015, Polade et al 2017).This variability is not well parameterized for semi-arid ecosystems, including chaparral, which limits models of future atmospheric carbon levels.

Study site
This experiment was conducted at the Sky Oaks Ecological Reserve in Warner Springs, CA (33 • 22 ′ 25 ′′ N, 116 • 37 ′ 19 ′′ W).Vegetation around the sampled site is dominated by A. fasciculatum, A. sparsifolium, and C. perplexans, with isolated populations of Arctostaphylos spp., Quercus berberidifolia, and other shrub species.The site is 1414 m above sea level, with an average precipitation of 394 mm yr −1 and an average seasonal temperature range of 8.26 • C to 23.2 • C (1997-2021).Interannual conditions during the sampling period differed substantially from historical conditions  collected by the nearby Ameriflux station US_SO2.This reserve was burned during the 2003 Coyote Wildfire, the first recorded wildfire within the past 150 years.Due to the site's remote location, it experienced little to no anthropogenic impacts during recovery.

Photosynthetic measurements
Photosynthetic rates were measured monthly for each species from August 2019 to August 2021.Five individuals from each species were randomly selected from the stand for sampling each month.Photosynthetic rates from leaves in full sunlight were measured across a range of descending artificial light levels using a Li-Cor 6400 Portable Photosynthesis System (LICOR Environmental, Lincoln, Nebraska).The chamber head was equipped with a red/blue LED array.Instrument parameters were replicated from de Lobo et al (2013): photosynthetic active radiation (PAR) in the chamber progressed from 1800 µmol photon m −2 s −1 -0 µmol photon m −2 s −1 at intervals of 200 µmol photon m −2 s −1 .Additional measurements were taken at 300 µmol photon m −2 s −1 and 100 µmol photon m −2 s −1 to improve resolution around the light compensation point.Each light level was maintained for a minimum of 2 min and a maximum of 4 min, ending at a stabilization point of 0.2 µmol CO 2 m −2 s −2 (de Lobo et al 2013).
The system was calibrated every four months to ensure accurate measurement of gas-based light response curves.Calibration included two-point references for both CO 2 and H 2 O. Zero points were determined using Airgas ultra-high purity N 2 (99.999%N 2 -Airgas: Randor, PA).A known H 2 O concentration was measured at a dewpoint of 15 • C using a LiCor LI-610 dewpoint generator (LiCor Biosciences: Lincoln, NE).A known CO 2 concentration was measured at 698.9 ppm ± 2% (Airgas: Randor, PA).Before each sampling period, zero points of H 2 O and CO 2 were checked using soda lime (NaOH and CaO) and drierite (CaSO 4 ) scrubbers.In addition, reference and sample gas concentrations were compared across each infrared gas analyzer between each light response curve sequence.
Raw photosynthetic measurements were further processed into light response curve parameters using a modified Michaelis-Menten equation (de Lobo et al 2013).This yielded model estimates of maximal photosynthetic rate, dark respiration, and light compensation point.Although there are several regression equations, a modified Michaelis-Menten relationship is often a simple and accurate regression strategy (de Lobo et al 2013).Parameters were solved using an iterative non-linear model in R (R Team 2013).Samples with a regression coefficient of two standard deviations within the mean were retained, and others were removed from the sample set.
The leaf area was measured optically and used to correct flux densities.Following each photosynthetic measurement, sampled leaf clusters were stored and refrigerated until they could be processed.The two-dimensional leaf area was measured using ImageJ.Undisturbed branch structure-similar to the profile present in the leaf chamber-and deconstructed leaflets-with the total exposed leaf area-were both measured.Both methods have been used in similar scenarios (Cosmulescu et al 2020).A. sparsifolium samples often had photosynthetic stem tissues during the growing season.These tissues were included in leaf area measurements when appropriate.

Water stress measurements
Water stress was measured via stem WP for each individual during monthly collections.From each of the five randomly selected individuals, two stems were sampled at both pre-dawn and solar noon.Samples were collected from healthy branches, approximately 10 cm from new growth.WP was measured using a PMS 1000 pressure bomb (PMS, Corvallis, Oregon) and nitrogen gas.This system was factory-calibrated at the onset of measurements in 2019.Samples were processed immediately in the field.Between collection and measurement, samples were stored in sealed bags in a cooler.
Stomatal conductance was corrected for irregular leaf shapes by determining a relationship between leaf area and boundary layer conductance for each species (Paw and Daughtry 1984).Irregular leaf shapes pose a problem for small chamber measurements of stomal conductance, as factory-calibrated boundary layer conductance is typically for a large, flat leaf occupying the full chamber width of the LI-6400 (LICOR Environmental, Lincoln, Nebraska).Correction to stomatal conductance relied on coating additional leaves with water and measuring the total resultant transpiration rate (assuming a stomatal conductance of zero and a relative surface humidity of 100%) (Paw and Daughtry 1984).Wetted samples were measured every second for 30 s, retaining the maximum transpiration rate.This was then translated into a boundary layer conductance and regressed against the leaf area.

Meteorological measurements
All sampling occurred within the previous Coyote Wildfire perimeter within a 200 m radius of the neighboring eddy covariance station (Ameriflux station US-SO2).General meteorological conditions were measured from the adjacent US_SO2 tower.Half-hourly measurements of wind speed, air temperature, relative humidity, incoming radiation, soil moisture (30 cm integrated sampling depth), and precipitation were connected to photosynthetic measurements at the nearest time increment.In rare cases when the main meteorological station was inoperable, conditions were sourced from a secondary tower 0.37 km from the photosynthetic sampling site (Ameriflux station US-SO3).Each tower is owned and maintained by San Diego State University.

Statistical analysis
Statistical analysis used multiple regression and a generalized additive model (GAM) to determine meteorological drivers of photosynthesis, conductance, and WP.Assumptions of multiple regression were checked optically using base R plotting.GAM is a non-parametric statistical test, where relationships are allowed to vary using spline functions.Regressors between both models were kept consistent.For the visual display of GAM outputs, statistically significant predictors from the linear model were allowed to vary.Other regressors were controlled for at a constant mean value.Comparisons between species' photosynthetic capacity and seasonal changes were assessed using an ANOVA and post hoc Tukey test.Collinearity was assessed using a variable inflation index.Comparisons between models used the Akaike information criterion to determine an optimal model.Seasons were grouped nontraditionally, following previous research at the US-SO2 eddy covariance site (Luo et al 2007).Seasons were characterized as a wet season (1 November-28 February), dry season (1 July-31 October), and growing season (1 March-30 June).Statistical analysis was conducted using this grouping where appropriate.

Photosynthetic capacity
Sampled species demonstrated significantly different maximum photosynthetic rates.Across the full sample period, A. sparsifolium had an average maximal photosynthetic rate of 15.8 ± 6.9 µmol C m −2 s −1 , A. fasciculatum averaged 13.1 ± 6.9 µmol C m −2 s −1 , and C. perplexans averaged 9.5 ± 5.5 µmol C m −2 s −1 between 2019-2021 (figure 1).All species' photosynthetic differences were significant across seasons (F = 17.2, df = 212, p < 0.001).Within seasons, A. sparsifolium demonstrated significantly higher photosynthetic rates than C. perplexans in the wet and growing seasons (October-June) (figure 1) (wet: p = 0.01; growing: p < 0.001).A. sparsifolium and A. fasciculatum did not significantly differ from each other in any of the seasons.When comparing light curve model performance, the modeled Michaelis-Menten predictions exhibited more variation in the dry season when compared to the wet and growing seasons (figure 1).Most of the variation was seen at higher light levels, suggesting additional influences of strong light during stressful meteorological conditions.

Meteorological conditions
Meteorological conditions during the sampling period varied between different drought intensities.Rainfall during 2019-2021 was 464 mm, 412 mm, and 392 mm respectively, compared to an average rainfall of 394 mm from 1997-2021.The average daily maximum air temperature in summer for 2019-2021 was 28.6 • C, 29.3 • C, and 30.6 • C respectively.The average maximum daily summer temperature from 1997-2021 was 28.8 • C. Summer daily maximum VPD for 2019-2021 was 2.45 kPa, 2.68 kPa, and 2.56 kPa respectively, compared to an average of 1.68 ± 0.33 kPa from 1997-2021.

Meteorological controls of photosynthesis
A multiple regression analysis of meteorological conditions found differing drivers for each species.Maximal photosynthetic rates of A. fasiculatum and A. sparsifolium were significantly correlated with monthly average PAR (A. fasciculatum: Beta = 1.13, p < 0.01-A.sparsifolium: Beta = 1.23, p < 0.001) and VPD (A. fasciculatum: Beta = −0.78,p < 0.01-A.sparsifolium: Beta = −0.58,p = 0.04), but did not correlate with precipitation, soil moisture, or NDVI (table 1).The maximal photosynthetic rate of C. perplexans was correlated with monthly averaged PAR (Beta = 1.07, p < 0.01) and summed monthly precipitation (Beta = 0.78, p 0.04) but did not correlate with VPD, soil moisture, or NDVI (table 1).All models met assumptions of normality and homoscedasticity via visual inspection of diagnostic plots.Co-linearity was not found between any predictors using a variable inflation index threshold of five (table 1) (Fox and Monette 1992).

Meteorological controls of WP
VPD was strongly correlated with WP for all species at both pre-dawn and mid-day.All three species are significantly correlated with VPD for midday WP measurements.A. fasciculatum and C. perplexans were significantly correlated with VPD for predawn WP (table 2).Predawn WP in the summer of 2020 was the lowest recorded for all species.A. fasciculatum and C. perplexans experienced their lowest WP in October 2020.These average monthly values were negative −63.5 bar and −79.3 bar, respectively (figure 2).The lowest WP for A. sparsifolium was recorded in September 2020, at an average value of −38.6 bar.The highest WP for all species was reported in April 2020.C. perplexans had the highest average midday WP at −8.6 bar, A. fasciculatum reported −19.2 bar, and as reported −20.3 bar.Periods of extremely low WP correlated with periods of reduced or near-zero stomatal conductance (figure 2).

Meteorological controls of stomatal conductance
Stomatal conductance for all species was correlated with hydrological predictors.All species were significantly correlated with soil moisture conditions (table 2), however only A. sparsifolium and A. fasciculatum were significantly correlated with VPD.All three models demonstrated strong predictability (table 2) (figure 3), with the highest stomatal conductance coinciding with the highest WP event in April 2020.During this event, A. sparsifolium had the highest stomatal conductance at 0.69 mmol H 2 O m −2 s −1 , A. fasciculatum at 0.64 mmol H 2 O m −2 s −1 , and C. perplexans at 0.57 mmol H 2 O m −2 s −1 .This event proceeded with the highest maximal photosynthetic rates seen in May 2020.Periods of near-zero stomatal conductance were observed for all species in the late fall in early summer of each year (between August and November).
Table 2. Hydro-physiological multiple regression: multiple linear regression of pre-dawn water potential against monthly average soil moisture, VPD, and precipitation.Multiple linear regression of average stomatal conductance against average, monthly soil moisture and VPD.Meteorological variables were collected from a neighboring Ameriflux station.Stomatal conductance was corrected for irregular leaf shape and surface area with a boundary layer conductance regression (supplemental 1).

Discussion
Productivity in the observed species demonstrated sensitivity to both the atmospheric and soil components of a drought scenario.A. fasiculatum and A. sparsifolium both exhibited significantly reduced photosynthetic capacity at higher VPD, while C. perplexans exhibited diminished capacity with less monthly rainfall.This effect was observed across both sampling years.This combination of multiple climate-related drivers may significantly hinder photosynthesis for these chaparral species under expected climate regime changes.The region is projected to have increased maximum summer temperatures and stochastic precipitation (Kogan and Guo 2015, Richman and Leslie 2015, Polade et al 2017).These results differ from existing studies that cite soil moisture as the primary predictor of photosynthesis for the studied chaparral species.Soil moisture was not a significant predictor of photosynthetic rate for any of the sampled species.While this is contrary to similar species-level measurements in chaparral (Mooney et al 1975, Kolb andDavis 1994), soil moisture in similar semi-arid shrublands can significantly lag behind precipitation and have hold-over effects between seasons or years (Jia et al 2018).The productivity of C. perplexans did significantly correlate with monthly summed precipitation, which parallels soil moisture at the near surface.Other results agreed with photoperiod and air temperature as major controls of photosynthetic rate, however, VPD was a more significant predictor when compared to air temperature.This distinction between soil and atmospheric drivers under contemporary conditions highlights the role atmospheric-water-stress can induce under acute drought (Richman and Leslie 2015).Similar drought conditions are expected to become more prevalent in the region (Polade et al 2017).As this study had a limited monthly sample size (n = 25), future work should assess if this signal is representative of decadal-scale changes of meteorological photosynthetic controls.
Significant differences in productivity followed hypothesized patterns by species functional type and may confer differing resilience across disturbance regimes.Resprouting species had a better correlation with atmospheric predictors, suggesting they may draw upon semi-stable groundwater sources, moderated by atmospheric conditions and stomatal regulation.In addition, resprouting species are more common in old-growth stands (Pratt et al 2014).Whereas seeding species, C. perplexans, productivity correlated with monthly precipitation.This was hypothesized due to their shallow root system.These functional type differences may exacerbate or buffer climate effects in specific chaparral regions depending on the community composition (Pratt et al 2008, 2014, Jacobsen and Pratt 2018).Water stress levels, particularly during recovery from wildfire, has been cited as a major contributor to long-term compositional changes, resprouter mortality, and type conversion in chaparral (Pratt et al 2008, Syphard et al 2019).Overall, this study suggests that the established connection between wildfire, water stress, and loss of resprouting species would weaken compensatory ground-water access, and create larger dependance on soil water stress conditions for these three common species (Jacobsen andPratt 2018, Syphard et al 2019).Future work could focus on developing larger samples of various functional type species to assess the overall climate resilience conferred by community composition.
Although not a focus of this study, there were signals of photooxidation during the dry season which may introduce additional photosynthetic limitations.During dry-season measurements, high light levels lessened following photosynthetic responses (figure 1).This is demonstrated in the lower agreement between predicted and observed rates of photosynthesis at high light levels (figure 1).Stressed individuals may be more likely to damage photosystems through oxidation during extreme temperature events.The collocated species, Quercus pubescens, exhibited similar oxidated damages when exposed to prolonged drought conditions (Saunier et al 2018).However, the precise drivers of this signal cannot be determined from this study, as disparity in the Michaelis-Menten model fit during the dry season may be a methodological artifact (de Lobo et al 2013).At sampling onset in the early morning, exposure to light levels higher than ambient conditions may also induce stomatal opening and artificially increase the measured photosynthetic rate (Shimazaki et al 2007).This is supported by observed peaks in stomatal conductance at sampling onset (high light levels) (supplemental 1).Overall, this study did not find conclusive effects of photooxidation during acute drought conditions.
WP varied substantially between each of the sampled species.C. perplexans showed the highest overall annual variation in WP, as well as the lowest values.The minimal value of the average monthly midday WP for C. perplexans (−80 bar in August 2020) exceeded reported values for the same species in two summers of 1970 (Burk 1978), and six chaparral species sampled during the 2011 drought (Quero et al 2011).This increase in water stress signals changes in C. perplexans resilience to modern climate.A. sparsifolium exhibited relatively little variation in annual WP.This is interesting, given that A. sparsifolium is experiencing substantial adult mortality and lack of recolonization following wildfire throughout its range (Wiens et al 2012).Although A. sparsifolium reported a higher WP than A. fasiculatum, there is evidence that A. sparsifolium is less resilient to cavitation and embolism (Redtfeldt and Davis 1996).A. fasiculatum's reported WPs were relatively variable but were comparable to previous studies (Burk 1978, Redtfeldt and Davis 1996, Quero et al 2011).
A limited relationship to NDVI supports the need for meteorological-based estimates of photosynthesis over purely optical remote sensing methodologies.Chaparral systems often have a less inherent correlation with common remote sensing indices, such as NDVI, due to the predominance of evergreen species (Bounoua et al 2000, DuBois et al 2018).The results of this study support other efforts to focus on biogeochemical-meteorological models of chaparral productivity.Although a limited number of species were sampled, there is also support for community composition of resprouting and seeding species as a decoupling predictor for the expected effects of drought and temperature extremes.Future work on remotely sensed indices in chaparral may benefit from using hyperspectral datasets.Narrow-band measurements of shortwave infrared reflectance have been connected to water status in chaparral (Ustin et al 1998, Serrano 2000), which more closely parallels the measured hydrological condition.
The effects of increased aridity in this chaparral system clearly impacted photosynthetic capacity and water stress for all studied species between 2019 and 2021.These effects have the potential to significantly limit future photosynthetic rates under projected drought conditions (Richman andLeslie 2015, Polade et al 2017).However, a longer-term assessment would determine if this signals a significant change in climatic response.Differences were observed across species functional types that may have complicated interactions when accounting for concurrent changes to successional and disturbance regimes.This study demonstrates that the largest changes in photosynthetic rate for these three species are correlated with increases in VPD.Under recent drought conditions in Southern California, increasing atmospheric dryness has been a major driver of carbon sequestration for chamise-redshank shrubland communities.

Figure 1 .
Figure 1.Photosynthetic light response curve: photosynthetic rates for each species across a range of light levels and seasons, averaged between 2019 and 2021 (n = 25).(A) Wet season (1 November-28 February); (B) dry season (1 July-31 October); and (C) growing season (1 March-30 June).Solid regression lines were determined with an iterative Michaelis-Menten equation fitting program.Error bars depict standard error.

Figure 2 .
Figure 2. Annual dynamics-(A) Average maximal photosynthetic rate as determined by a Michaelis-Menten equation for each species during the sampling period.Error bars depict standard errors.(B) Average monthly vapor pressure deficit (VPD-orange dashed) and average air temperature at 2 m (red, solid line) during the sampling period.(C) Summed bi-weekly precipitation (blue columns) and soil moisture measured across 30 cm of depth (light blue line).(D) Average monthly pre-dawn water potential of each sampled species.Error bars depict standard errors.

Figure 3 .
Figure 3. Generalized additive model: non-parametric depictions of the most significant meteorological predictors for maximal photosynthetic rate and average stomatal conductance for each species.(Top) Maximal photosynthetic rate predicted by average monthly VPD and summed monthly precipitation for A. fasiculatum, A. sparsifolium, and C. perplexans.(Bottom) Average stomatal conductance predicted by monthly VPD and soil moisture for the same species.

Table 1 .
Photosynthetic multiple regression: multiple linear regression of maximal photosynthetic rate with monthly summed precipitation and average monthly soil moisture, VPD, PAR, and NDVI.Meteorological variables were collected from a neighboring Ameriflux station.NDVI was calculated as an 8 d average of Landsat imagery to minimize cloud coverage.Using a variable inflation index (VIF) threshold of five, no collinearity was observed.