2.3.3. Drought Stress Determination

According to [56], drought refers to conditions characterized by low available soil moisture and high atmospheric VPD. As such, in quantifying the dry periods in our study, we used the Standardised Precipitation–Evapotranspiration Index (SPEI), which takes into account both precipitation and potential evapotranspiration for the determination of drought over the main growing season. The SPEI was calculated for various lags (1, 3, 6, 12, and 24 months) from monthly records. We used the period 1981–2010 as the baseline period for the SPEI computation. The computation of SPEI was performed according to [57,58], using the R package SPEI [59]. The SPEI represents an anomaly in the climatological water balance, which is given by the difference between precipitation and reference evapotranspiration. The input data for the SPEI derivation were obtained from climate reanalyses ERA5 monthly averages available at 0.25◦ spatial resolution, where the reference evapotranspiration was computed from air temperature at 2 m, air dew point temperature in 2 m, wind speed in 2 m converted from the original 10 m height using a logarithmic profile law, and shortwave incoming radiation, following the methodology by [60]. In determining the occurrence of dry and wet events in our study, SPEI classification based on [58] was used (Tables 3 and 4).

**Table 3.** The Standardised Precipitation-Evapotranspiration Index (SPEI) categories based on the classification of SPEI values by [58].


**Table 4.** Categorization of dryness/wetness using Standardised Precipitation-Evapotranspiration Index (SPEI) indices for CZ-BK1 and CZ-RAJ stations in years (2014–2016).


2.3.4. Light Response Curve Fitting

Light response curves (LRC) of daytime GPP were fitted at half-hourly time resolution using the logistic sigmoid approach by [61]:

$$\text{GPP} = 2 \times \text{GPP}\_{\text{max}}(0.5 - \frac{1}{1 + \varepsilon x \, p(\frac{-2\text{aPAR}}{\text{GPP}\_{\text{max}}})}) \tag{3}$$

where PAR is the photosynthetically active radiation, *α* is the apparent quantum yield in mol (CO2) mol−<sup>1</sup> (phot.) that describes the effectivity of photosynthesis at low light conditions, and the GPPmax is the asymptotic maximum assimilation rate at the light saturation point in μmol m<sup>2</sup> s−1. The fitting was done separately for half-hourly values from years with normal conditions (2014 and 2016) and the year with drought stress conditions (2015). To eliminate night-time measurements, we used a PAR threshold 10 μ, i.e., light compensation irradiance of 10 μmol m2 s−1, representing light compensation point of NEE.

Since VPD and soil moisture affect GPP [49,62], the effects of VPD and SVWC on the LRC residuals (GPP values after the removal of the strongest PAR dependence) were

analysed for both spruce forest sites within the period under study. A piecewise regression was performed on the LRC residuals versus VPD and SVWC to determine the site-specific environmental stress thresholds at which GPP declined during the growing season periods of normal and drought affected years.

#### 2.3.5. Piecewise Regression Analyses for the Assessment of Drought Effect

Since the LRC model does not account for changes in both VPD and SVWC, piecewise regression of residuals against these environmental variables were analysed. This allowed to determine the response of the studied spruce forest ecosystems with contrasting climates to severe drought conditions over the growing season for normal and drought-affected years. A piecewise regression analysis as part of the R package in 'segmented' [63] and the Davies test [64] were used to detect the breakpoints in the regression and also to test for the significant differences in slope parameters of a plot of residuals from the LRC model to drought conditions (high VPD and low SVWC).

#### **3. Results**
