Simulating the Impacts of Drought and Warming in Summer and Autumn on the Productivity of Subtropical Coniferous Forests

Abstract

The impacts of drought and/or warming on forests have received great attention in recent decades. Although the extreme drought and/or warming events significantly changed the forest demography and regional carbon cycle, the seasonality quantifying the impacts of these climate extremes with different severities on the productivity of subtropical coniferous forests remains poorly understood. This study evaluated the effects of seasonal drought and/or warming on the net primary productivity (NPP) of subtropical coniferous forests (i.e., Cunninghamia lanceolata and Pinus massoniana forests) from Hengyang–Shaoyang Basin in southern China using the Ecosystem Demography model, Version 2.2 (ED-2.2) and based on the datasets from forest inventory, meteorological reanalysis, and remotely sensed products. The results showed that the goodness of fit of the DBH-height allometric equations was better than that of the default in ED-2.2 after model calibration; the ED-2.2 model qualitatively captured the seasonality of NPP in the subtropical coniferous forests; and the mismatch between simulated annual NPP and MODIS-NPP (MOD17A3HGF) became smaller over time. The effect of seasonal drought on NPP was greater than that of warming; the decline rate of NPP gradually increased and decreased with time (from July to October) under the seasonal drought and warming scenarios, respectively; NPP decreased more seriously under the combined drought-warming scenario in October, with an average decrease of 31.72%, than the drought-only and warming-only scenarios; seasonal drought had an obvious legacy impact on productivity recovery of subtropical coniferous forests, but it was not the case for warming. With the increase in drought severity, the average values of soil available water and NPP together showed a downward trend. With the increase in warming severity, the average values of canopy air space temperature increased, but NPP decreased. Seasonal drought and/or warming limit forest production through decreasing soil moisture and/or increasing canopy air space temperature, which impact on plant photosynthesis and productivity, respectively. Our results highlight the significance of taking into account the impacts of seasonal warming and drought when evaluating the productivity of subtropical coniferous forests, as well as the significance of enhancing the resistance and resilience of forests to future, more severe global climate change.

1. Introduction

The impacts of drought and/or warming on forests have received great attention in recent decades [1,2,3,4]. It is important to investigate the effects of future drought and/or warming events on forest productivity. Since the 20th century, with the intensification of global climate change, drought and/or warming events have occurred more frequently across the globe [3,5,6,7,8,9,10,11], and the trend is expected to be stronger in the future under climate change [10,12]. The forest is a vital component for carbon sink among the terrestrial ecosystems [13], and their demography (e.g., growth, mortality, and reproduction) may be influenced by climate extremes (e.g., drought and warming), and further causing changes in the regional carbon cycle [3,4,14]. Drought and warming are inextricably linked [15], and drought often occurs with warming [15], triggering a series of synergies and jointly affecting tree and seedling mortality [15,16,17,18,19] and forest productivity. Drought stress also increases the risk of fires, causes outbreaks of pathogens and pests, and affects the forest demography, which seriously affects forest productivity and carbon absorption [15]. Consequently, investigating the effects of drought and/or warming on forests is critical in a changing climate.

The current research on forest drought and/or warming varies by climatic zones. So far, research on drought and/or warming has focused primarily on tropical forests, such as the Amazon rainforest [7,15,20,21,22,23] due to the distinctive dry and rainy seasons by the tropical monsoon climate. The uneven precipitation leads to frequent extreme climate events such as drought in the tropical forests [24]. There are few studies on subtropical region where forests are also vulnerable to drought and/or warming [11,15]. Subtropical regions in China have experienced extreme temperature, and are accompanied by changing precipitation seasonally over the past few decades [11,25,26]. At the same time, the frequency and severity of drought and/or warming events in the subtropical regions will increase in the future [24,27,28]. Forest responses to drought and/or warming can be assessed by their resistance and resilience [29]. Resistance measures how an ecosystem changes immediately after a disturbance, and resilience evaluates the degree of ecological functioning that is restored to pre-disturbance equilibrium [30,31,32]. Therefore, it is important to quantify the sensitivity (resistance) and post-event recovery (resilience) of the subtropical coniferous forests to drought and/or warming under the climate change.

The timing of extreme drought and/or warming affects the response of ecosystems to disturbances [33]. Seasonal drought is frequent in the subtropical region throughout the summer, where a West Pacific high pressure system dominates the bottom layer of the atmosphere [34,35]. Seasonal drought often happens when vegetation is actively growing [34,36], drastically lowering vegetative productivity. Previous research has primarily focused on drought in summer [37]; however, there have been some studies on other seasonal droughts (e.g., spring and winter) [33,38]. Hengyang–Shaoyang Basin is a high incidence area of drought events in the subtropical region of China. High rates of seasonal drought in subtropical forests, with more severe drought in the summer and autumn are observed in Hengyang–Shaoyang Basin, China [39]. Quantifying the impacts of these seasonal climate extremes on subtropical forests in the Hengyang–Shaoyang Basin is still poorly understood.

Forest inventory, modeling, and remote sensing techniques are commonly used to explore the extreme climates on forests [40,41,42]. Forest inventory data have been used to simulate/track drought-induced mortality [40,43]. In addition to traditional enhanced vegetation index, new developed remote-sensing products/indices (e.g., solar-induced chlorophyll fluorescence, and vegetation optical depth) are gradually applied for investigating vegetation productivity [41,44,45]. Cohort-based ecosystem demography (ED) models [46,47,48], which characterize the horizontal and vertical heterogeneity of forest ecosystems, are also frequently simulating drought impacts on the productivity of tropical forests [42]. Remote sensing specializes in capturing the past and current changes of forest attributes; however, process-based modeling projects the response of forest attributes to different future climate scenarios. Remote sensing and process-based modeling techniques improve quantifying and understanding of how forest productivity responds to extreme climate impacts under the historical, current, and future situations. Forest inventory, satellite data, and modeling could be combined to investigate the impacts of disturbances (e.g., drought and/or warming) on forest attributes (e.g., productivity); for instance, the forest inventory and freely available satellite product could provide data supporting model initialization and verification/validation, respectively [49].

In this study, we used the process-based ED-2.2 model, remotely sensed NPP product, and forest survey data to investigate the influence of different degrees of seasonal drought and/or warming (from July to September) on the productivity of subtropical coniferous forests (i.e., C. lanceolata and P. massoniana forests) from the Hengyang–Shaoyang Basin in southern China. The specific objectives are to (1) quantify the impacts of drought and/or warming in summer and autumn with different severities on the forest productivity; and (2) investigate the factors leading to seasonal drought- and/or warming-induced NPP decline.

2. Materials and Methods

2.1. Data Collection

In the Hengyang–Shaoyang Basin, one of the most frequently affected regions by drought and warming in Hunan Province, southern China, subtropical coniferous forests plots were chosen for this study. Mean annual temperature is 17.3 °C. Total annual precipitation is 1339.6 mm [50]. Additionally, this location is among the best places for coniferous species, such as Pinus massoniana and Cunninghamia lanceolata. We selected 10 forest plots with a plot size of 0.0667 ha from the fifth (2004) National Forest Inventory of China (

Table 1. Descriptions of subtropical coniferous forest plots in the study.

Plot	Stand Origin	Forest Type	Elevation [m]	Mean Age [a]	Mean DBH [cm]	Canopy Height [m]
1	Plantation forest	C. lanceolata	540	17	11.8	12.2
2	Natural forest	P. massoniana	220	23	11.1	10.0
3	Plantation forest	C. lanceolata	600	14	12.2	8.3
4	Natural forest	P. massoniana	230	17	12.8	8.0
5	Natural forest	P. massoniana	550	24	10.2	7.7
6	Plantation forest	P. massoniana	160	19	10.2	8.0
7	Natural forest	P. massoniana	400	20	11.8	12.5
8	Natural forest	P. massoniana	390	22	10.8	10.6
9	Natural forest	P. massoniana	290	31	14.0	11.7
10	Plantation forest	P. massoniana	430	31	12.2	9.8

). The selected plots meet the following criteria: (1) from Hengyang–Shaoyang Basin; (2) belong to C. lanceolata forest or P. massoniana forest; (3) not affected by other forest disasters (e.g., diseases and insect pests, wildfire, and ice-storm, etc.) from the inventory record; and (4) average diameter at breast height (DBH) ≥ 10 cm. We initialized the ED-2.2 model by using the forest inventory data in the 2004.

We used the meteorological reanalysis dataset MERRA-2 [51,52,53,54,55] (https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/) (accessed on 16 March 2022) with an hourly resolution and a spatial resolution of 0.5° × 0.625° and cycled the meteorological data from nearly a decade (2012–2021) as a baseline climate to drive the ED-2.2 model. The drought and/or warming scenarios were based on modifying MERRA-2 reanalysis datasets, following the meteorological drought levels of the Grades of Meteorological Drought (GB/T 20481-2017) [56] and the prediction of the level of future temperature increases in the 6th Assessment Report of IPCC [27]. Soil data are from the SoilGrids 2.0 database [57,58] and a global gridded data set [59].

2.2. Model Configuration and Simulation Setup

2.2.1. Model Overview

The Ecosystem Demography models (e.g., ED-1.0, ED-2.0, and ED-2.2) have been developed for more than 20 years [46,47,48]. In this study, we used the stable version for the Ecosystem Demography model (i.e., ED-2.2) [46,47,48] to simulate the effects of drought and/or warming with different severities on the NPP of subtropical coniferous forests from southern China. The ED-2.2 model is a cohort-based ecosystem model with integrated submodels of forest demography (i.e., growth, mortality, and recruitment) as well as enthalpy, water, and carbon cycles [46,47,48]. The cohort-based ED-2.2 represents the vertical and horizontal heterogeneity of forest ecosystems [48], and allows vegetation composition to change with time [60]. ED-2.2 is a collaborative open-source model, and its software and further development are publicly available (https://github.com/EDmodel/ED2) (accessed on 16 March 2022) [48]. Recently, the Ecosystem Demography model and its derivatives as well as coupled with other models have been calibrated, validated, and applied across the temperate forests [61,62,63,64], subtropical forests [65], and tropical forests [42,66,67,68,69].

2.2.2. Plant Functional Type

Plant functional types (PFTs), defined by biologically similar vegetation groups, used in the ED-2.2 model [46,47,70]. The ED-2.2 model includes three PFTs of tree species (e.g., PFT2/4/7) that are suitable for the subtropical coniferous forests in the study sites: PFT2 (early successional, non-shade-tolerant, broadleaved), PFT4 (late-successional, shade-tolerant, and broadleaved), and PFT7 (non-late successional, non-shade-tolerant, and coniferous). The species successional types are based on the Regulations on the Inventory of Forest Resources in Hunan Province, China [71].

2.2.3. Model Calibration

In order to localize the model, we corrected some ecophysiological and allometric parameters of the model (

Table 2. Ecophysiological and allometric parameters for the plant functional types.

Plant Functional Type	SLA [m2 kgC−1]	b1Ht [m]	b2Ht [cm−1]
PFT2	24.5	0.026	0.715
PFT4	36.6	0.026	0.717
PFT7	7.4	51.037	−0.014

SLA: specific leaf area; b1Ht: height allometric intercept; b2Ht: height allometric slope.

). In the ED-2.2 model, specific leaf area (SLA) is a more sensitive parameter for plant functionality [72,73], and we modified the SLA of PFT2, PFT4, and PFT7 to replace the default parameters of the model according to the existing studies from Zhao et al. (2009) [74] and Zhu et al. (2022) [65], in which the parameters used are from the TRY global plant functional trait databases [75,76]. According to the existing research [65], the allometric growth equation of PFT7 was modified. We also modified the DBH-height allometric equation (i.e., a modified Weibull function, Equation (1)) of PFT2 and PFT4 in the ED-2.2 model [48] by using forest inventory data (Table 1):

$$\mathrm{Height}={\mathrm{H}}_{\mathrm{ref}}·\left(1-\exp \left(-\mathrm{b}1\mathrm{Ht}·{\mathrm{DBH}}^{\mathrm{b}2\mathrm{Ht}}\right)\right)$$

(1)

where Height is tree height (m); DBH is diameter at breast height (cm); H_ref was reference height [61.7, m]; b1Ht [m] was allometry intercept; b2Ht [cm⁻¹] was allometry slope.

After calibration, the RMSE and MAE of the corrected DBH-height allometry for PFT2 decreased by 34.7% and 57.4% compared with the default allometry, respectively. The RMSE and MAE of the corrected DBH-height allometry for PFT4 decreased by 35.8% and 58.8% compared with the default allometry, respectively (

Table 3. Calibration of DBH-height allometric equation for PFT2 and PFT4.

Plant Functional Type	Allometric Equation	RMSE	MAE
PFT2	Height = 61.7·(1 − exp(−0.035·DBH0.695)) A	2.58	6.66
	Height = 61.7·(1 − exp(−0.026·DBH0.715)) B	1.68	2.84
		−34.7%	−57.4%
PFT4	Height = 61.7·(1 − exp(−0.042·DBH0.522)) A	2.58	6.67
	Height = 61.7·(1 − exp(−0.026·DBH0.717)) B	1.66	2.75
		−35.8%	−58.8%

A and B indicate the default and calibrated DBH-height allometric equations, respectively. The percentage represents the rates of change in RMSE and MAE between before and after calibration.

).

2.2.4. Model Verification and Validation

For the verification of NPP, we simulated and extracted the monthly NPP values from 2005 to 2021, plotted them as an NPP inter-monthly change figure, and compared it with the measured monthly growth changes of C. lanceolata or P. massoniana forests. For the validation, we compared the differences between the yearly MODIS-NPP product (MOD17A3HGF) [77] (https://lpdaac.usgs.gov/product_search/) (accessed on 16 March 2022) and the model simulated NPP, and based on previous studies [25], we assumed that the mean relative error for MODIS-NPP was 23.58%. The validation data are divided into two distinct periods, the first period is from 2005 to 2013, and the second period is from 2014 to 2021; we calculated the root mean squared error (RMSE, Equation (2)) and mean absolute error (MAE, Equation (3)) based on the ED-2.2 simulation and the MODIS-NPP in the two periods, respectively:

$$\mathrm{RMSE}=\sqrt{\frac{{\displaystyle \sum }\left({\mathrm{Sim}}_{\mathrm{i}}-{\mathrm{Obs}}_{\mathrm{i}}\right)}{\mathrm{N}}}$$

(2)

$$\mathrm{MAE}=\frac{{\displaystyle \sum }\left|{\mathrm{Sim}}_{\mathrm{i}}-{\mathrm{Obs}}_{\mathrm{i}}\right|}{\mathrm{N}}$$

(3)

where Sim_i was NPP [kgC m⁻² yr⁻¹] simulated by ED-2.2; Obs_i was NPP [kgC m⁻² yr⁻¹] of the MODIS product; N was the number of years.

For the verification of monthly NPP, we found that the monthly variation of NPP over the years (2005–2020) decreased slowly after reaching the peak in May (Figure 1), slowed down from July to September, and accelerated after September. For the validation of annual NPP (Figure 2), in the first period (2005–2013), MAE and RMSE were 0.157 and 0.178, respectively. The NPP simulations basically fall in the error range of MODIS-NPP. In the second period (2014–2021), MAE and RMSE are 0.149 and 0.150, respectively. NPP simulation is all falling within the error range of MODIS-NPP. Compared with the first period, the MAE and RMSE in the second period decreased by 5.09% and 15.7%, respectively. The mismatch between ED-2.2 simulated NPP and MODIS-NPP gradually decreases over time although the model generally overestimates NPP compared with MODIS products. The ED2 simulated NPP vs. MODIS-NPP for a separate plot, as shown in Figure S1.

2.2.5. Simulation Scenarios of Disturbance Types and Severities

In the ED-2.2 model, we performed an interannual calculation of NPP for each forest plot to ensure that each plot had been affected by drought and/or warmings of varying severities in July, August, and September. Drought scenarios were based on the Grades of Meteorological Drought [56]; and warming scenarios were based on the projections of possible temperature increases by the IPCC in 2021 and 2050 [27] and some existing studies [11,78,79].

Scenario settings. We set up 10 scenarios of four types (

Table 4. ED-2.2 model scenario settings for baseline, drought, and/or warming.

Type	Scenario	Precipitation Rate [kg m−2 s−1]	Air Temperature [K]
Default	Baseline (B0)	P0	T0
Drought-only scenario	Mild drought (D1)	P0 × 0.6	T0
	Moderate drought (D2)	P0 × 0.4	T0
	Severe drought (D3)	P0 × 0.2	T0
Warming-only scenario	Mild warming (W1)	P0	T0 + 1.0
	Moderate warming (W2)	P0	T0 + 1.5
	Severe warming (W3)	P0	T0 + 2.0
Drought-warming scenario	Mild drought-warming (D1 + W1)	P0 × 0.6	T0 + 1.0
	Moderate drought-warming (D2 + W2)	P0 × 0.4	T0 + 1.5
	Severe drought-warming (D3 + W3)	P0 × 0.2	T0 + 2.0

P₀ and T₀ are the precipitation rate [kg m⁻² s⁻¹] and air temperature [K] under the baseline scenario, respectively.

). Type 1 was baseline (B0). Type 2 was drought-only scenarios, including mild drought (D1, P₀ × 0.6), moderate drought (D2, P₀ × 0.4), and severe drought (D3, P₀ × 0.2). Type 3 was a warming-only scenario, including mild warming (W1, T₀ + 1.0), moderate warming (W2, T₀ + 1.5), and severe warming (W3, T₀ + 2.0). Type 4 was drought-warming scenarios, including mild (D1 + W1, P₀ × 0.6, and T₀ + 1.0), moderate (D2 + W2, P₀ × 0.4, and T₀ + 1.5), and severe (D3 + W3, P₀ × 0.2, and T₀ + 2.0) (Table 3). We assumed the future drought and/or warming occurred after 5 years (from 2021, i.e., 2026), and we also assumed each climate scenario occurred consecutively in July, August, and September [80,81] in 2026.

In order to compare the difference in coniferous forest NPP between baseline and disturbance scenarios, we used the boxplot which is based on the simulation results of ten plots (including P. massoniana forests and C. lanceolata forests) over the course of a year.

3. Results

3.1. Forest Productivity during and after the Drought and/or Warming Scenarios

In general, NPP showed a significant downward trend in the three months (from July to September) when climate extremes occurred (Figure 3). The average NPP when drought and warming happen was 0.83 kgC m⁻² yr⁻¹, which was significantly lower than that in the B0 scenario (1.05 kgC m⁻² yr⁻¹), decreased by 20.78%. In addition, the average NPPs of the warming and drought scenarios were 0.95 kgC m⁻² yr⁻¹ and 0.96 kgC m⁻² yr⁻¹, respectively, and the change rates were −9.04% and −7.95% (Figure 3). In addition to the W1, W2, W3, D1, and D1 + W1 scenarios, the monthly NPP variation trend for other scenarios is different from that of the B0 scenario. Especially in the D3 and D3 + W3 scenarios, NPP decreased sharply after reaching the peak in May until the end of the disturbance (after September), which was very different from that in the B0 scenario, in which NPP decreased slowly in July after reaching the peak in May. The average monthly NPP for each forest plot is shown in Figure S2.

According to the monthly difference, we calculated the difference and relative rate of change of NPP during (July~September) and after (October~December) the drought and/or warming scenarios (

Table 5. Absolute change and relative change rate of NPP (kgC m⁻² yr⁻¹) during the drought and/or warming scenarios (July~September).

Month	NPP [kgC m−2 yr−1]		Absolute Change [kgC m−2 yr−1]	Relative Rate of Change [%]
	Baseline	Disturbance
July	1.014	0.986 (D1)	−0.028	−2.76
		0.966 (D2)	−0.048	−4.73
		0.935 (D3)	−0.079	−7.79
		0.941 (W1)	−0.073	−7.20
		0.905 (W2)	−0.109	−10.75
		0.869 (W3)	−0.145	−14.30
		0.910 (D1 + W1)	−0.104	−10.26
		0.847 (D2 + W2)	−0.167	−16.47
		0.768 (D3 + W3)	−0.246	−24.26
August	1.048	1.008 (D1)	−0.040	−3.82
		0.970 (D2)	−0.078	−7.44
		0.905 (D3)	−0.143	−13.65
		0.988 (W1)	−0.060	−5.73
		0.958 (W2)	−0.090	−8.59
		0.928 (W3)	−0.120	−11.45
		0.941 (D1 + W1)	−0.107	−10.21
		0.857 (D2 + W2)	−0.191	−18.23
		0.736 (D3 + W3)	−0.312	−29.77
September	1.078	1.055 (D1)	−0.023	−2.13
		1.007 (D2)	−0.071	−6.59
		0.841 (D3)	−0.237	−21.99
		1.023 (W1)	−0.055	−5.10
		0.994 (W2)	−0.084	−7.79
		0.965 (W3)	−0.113	−10.48
		0.992 (D1 + W1)	−0.086	−7.98
		0.887 (D2 + W2)	−0.191	−17.72
		0.526 (D3 + W3)	−0.552	−51.21

and

Table 6. Absolute change and relative change rate of NPP (kgC m⁻² yr⁻¹) after the drought and/or warming scenarios (October~December).

Month	NPP (kgC m−2 yr−1)		Absolute Change (kgC m−2 yr−1)	Relative Rate of Change (%)
	Baseline	Disturbance
October	1.034	1.006 (D1)	−0.028	−2.71
		0.894 (D2)	−0.140	−13.54
		0.469 (D3)	−0.565	−54.64
		1.032 (W1)	−0.002	−0.19
		1.031 (W2)	−0.003	−0.29
		1.031 (W3)	−0.003	−0.29
		0.993 (D1 + W1)	−0.041	−3.97
		0.796 (D2 + W2)	−0.238	−23.02
		0.329 (D3 + W3)	−0.705	−68.18
November	0.750	0.741 (D1)	−0.009	−1.20
		0.718 (D2)	−0.032	−4.27
		0.689 (D3)	−0.061	−8.13
		0.749 (W1)	−0.001	−0.13
		0.749 (W2)	−0.001	−0.13
		0.748 (W3)	−0.002	−0.27
		0.736 (D1 + W1)	−0.014	−1.87
		0.707 (D2 + W2)	−0.043	−5.73
		0.669 (D3 + W3)	−0.081	−10.80
December	0.720	0.716 (D1)	−0.004	−0.56
		0.708 (D2)	−0.012	−1.67
		0.695 (D3)	−0.025	−3.47
		0.719 (W1)	−0.001	−0.14
		0.718 (W2)	−0.002	−0.28
		0.718 (W3)	−0.002	−0.28
		0.714 (D1 + W1)	−0.006	−0.83
		0.704 (D2 + W2)	−0.016	−2.22
		0.681 (D3 + W3)	−0.039	−5.42

). NPP varies over time in different scenarios. In the first month (July) after the disturbance, NPP under drought, warming, and drought-warming scenarios showed a significant downward trend, and NPP in the drought-warming scenario decreased by 17.00% on average. NPP under the warming scenario decreased by 10.75% on average; NPP declined by an average of 5.09% under drought conditions. In August, NPP decreased by 19.40% on average under drought-warming scenarios; NPP under the warming scenario decreased by 8.59% on average; NPP declined by an average of 8.30% under drought conditions. In September, NPP decreased by 25.64% on average under drought-warming scenarios; NPP under the warming scenario decreased by 7.79% on average; and NPP declined by an average of 10.24% under drought conditions. In general, within three months of the setting scenario, the decline rate of NPP under the drought-warming scenario and drought scenario gradually increased, while the decline rate of NPP under the warming scenario gradually slowed down. In the following October, the three scenarios maintained this trend and returned to normal levels in November and December (Table 6). NPP decreased more seriously under the combined drought-warming scenario, with an average decrease of 31.72%, than the drought-only and warming-only scenarios in October.

3.2. Relationships between Available Water, Canopy Temperature, and Forest Productivity under the Disturbance Scenarios

We explored the relationship between soil available water and NPP under drought scenarios from July to October (Figure 4). The results showed that the mean value of soil available water from boxplots under different drought scenarios (i.e., D1, D2, and D3) gradually decreased as the drought level increase (from mild drought to severe drought). In addition, the mean value of NPP from boxplots under different drought scenarios gradually decreased with the increase in drought level. As a result, NPP and soil available moisture changed in the same direction with increasing drought severities.

We investigated the relationship between canopy temperature and NPP under warming conditions from July to October (Figure 5). The results showed that the mean value of canopy temperature from boxplots under different warming scenarios (i.e., W1, W2, and W3) increased as warming level increase (from mild warming to severe warming). Furthermore, the mean NPP from boxplots under different warming scenarios decreased with the increase in warming level. Consequently, NPP and canopy temperature changed in the opposite direction with increasing warming severities. We also found a nonlinear relationship between NPP and canopy temperature (Figure 5). With the increase in canopy temperature, NPP showed a trend of first increasing and then decreasing.

4. Discussion

4.1. Model Performance in the Subtropical Coniferous Forests

The goodness of fit for the DBH-height allometric equation after model calibration was better than the default value in ED-2.2 (Table 3), indicating that the ED-2.2 model can reliably simulate tree height and related tree biomass (e.g., leaf, sapwood, heartwood, bark, and root). Although ED-2.2 models are commonly used in tropical forests such as the Amazon rainforest [49,82], they are rarely used to model the dynamics and response of subtropical coniferous forests to disturbances in southern China [65]. Consequently, we refitted the DBH-height allometric equation based on forest inventory data and corrected the sensitive physiological and ecological parameter to localize the parameter in ED-2.2 as much as possible (i.e., SLA).

The ED-2.2 model qualitatively captured the seasonality of NPP in subtropical coniferous forests (Figure 1), indicating that the model can be used to simulate the monthly productivity dynamics of subtropical coniferous forests in southern China. The results of the model for monthly productivity changes in C. lanceolata and P. massoniana forests are highly similar to those of another study on the monthly growth of C. lanceolata forests in subtropical China [83].

The mismatch between simulated annual NPP and MODIS-NPP gradually decreased over time (Figure 2), which means that we can use ED-2.2 to accurately predict the future productivity dynamics of subtropical coniferous forests. This study found that ED-2.2 performed better in simulating NPP in the second period (2014–2021) compared with the first period (2005–2013) of model validation. Although ED-2.2 overestimated annual NPP compared with MODIS products, the error between model simulation and remotely sensed observation gradually decreases over time. Another study using the ED2 model to predict changes in the carbon flux of a Puerto Rican tropical forest under climate change scenarios showed similar results [84]. The mismatch in evaluating NPP might be derived from the uncertainty of remotely sensed products, and when soil water stress was present, MODIS data have the potential to underestimate vegetation production [85].

4.2. Resistance and Resilience of the Subtropical Coniferous Forests to Seasonal Drought and/or Warming

In the study, the effect of drought on NPP was greater than that of warming (Figure 3 and Table 5), indicating that the resistance of subtropical coniferous forests to warming is stronger compared with drought. Our result was similar to another ED2 model simulation study from a Puerto Rican tropical forest where a drought-only scenario reduced NPP more than warming-only disturbance [84]. Although forests are potentially susceptible to all types of extreme events, globally, drought is considered the most widespread factor affecting the carbon balance [15]. For instance, during the European 2003 heatwave, a precipitation (and soil moisture) deficit rather than temperature was the main factor reducing carbon fluxes in the temperate and Mediterranean forest ecosystems [86,87].

The decline rate of NPP gradually increased and decreased with time (from July to October) under the drought and warming scenarios, respectively (Figure 3 and Table 6), implying that the resilience of subtropical coniferous forests to warming is stronger compared with drought. Our results showed that drought, rather than warming, had a significant legacy effect on NPP, which was largely due to changes in soil moisture caused by drought. A recent study of European forests showed that even if climatic conditions improve, vegetation production will remain in decline because drought will result in substantial and long-lasting damage to trees [2]. Warming rarely shows only extreme temperatures, usually together by subsequent climate extremes, namely “compound events” [15]. During the combined drought-warming event, stomatal closure induced inhibition of plant photosynthesis therefore limit the carbon absorption of forest ecosystems [88]. This explains why there is no legacy effect after the warming-only scenario in this study. Furthermore, warming and drought often have a synergistic effect; and the early onset of growth and the growth restriction caused by drought are the key factors influencing how warming affects vegetation productivity [89].

4.3. Limitations from Soil Moisture and Canopy Temperature on Productivity

With the increase in drought severity, the average values of available water and NPP together showed a downward trend (Figure 4), suggesting that seasonal drought limited forest production through decreasing soil moisture, which impacts on plant photosynthesis [85]. In the northern temperate and boreal forests of Canada and Amazon tropical forests, seasonal drought has been found to lead to a decline in vegetation productivity and change the regional carbon cycle [3,16,42,48]. Water stored in trees may momentarily alleviate the negative impact of short-term drought [88,90], but long-term drought may ultimately lead to tree hydraulic failure, which is one of the physiological reasons for explaining the death of trees [88,91,92]. Furthermore, available water for the soil layer can limit plant photosynthesis and productivity built into the ED-2.2 model [48]. The stomatal conductance equation [93] fails to characterize the soil moisture limitation of photosynthesis, however, that may be critical for the terrestrial ecosystems with seasonally dry stress [48]. The ED-2.2 model designed a wilting factor to restrain plant production with decreasing soil available moisture for filling in the gaps in the Leuning (1995) equation. The ED-2.2 model also defined soil water availability for photosynthesis, producing a gradual change from stomatal opening to closing when available water is close to the wilting point [48].

With the increase in warming severity, the average values of canopy temperature increased, but NPP decreased (Figure 5), indicating that seasonal warming limited forest production through increasing canopy air space (CAS) temperature, which also impacts on plant photosynthesis [10]. Warming will lead to a decline in forest vegetation productivity [89,94,95]. Extreme temperatures can impact on various plant physiological processes (e.g., photosynthesis and respiration) [96]. Warming impacts on photosynthetic and respiratory activities of plants through increasing chlorophyllase activity and decreasing photosynthetic pigment [97,98], resulting in limited plant production [96,99]. Moreover, temperature for the CAS layer can limit plant photosynthesis and productivity built-in the ED-2.2 model. The maximum attainable carboxylation under ribulose-1,5-biphosphate (RuBP)-saturated conditions for CAS follows a modified Michaelis–Menten equation [48], in which the Michaelis constants depend on temperature, whereas the maximum carboxylation rate is expressed by a modified temperature-dependent function, explaining the rapid decline in production under the extreme temperatures [46,100].

4.4. Limitations and Uncertainties

Firstly, we only took into account a single drought and/or warming event in the simulation. However, these extremes will occur more frequently in the future [15,27]. Further research could consider impacts with various frequencies on the productivity. Secondly, we assumed that seasonal extremes would occur after 5 years because we mainly focused on the impacts on the seasonal variation of NPP without considering the interannual effects. However, the year of disturbance in the future was often uncertain. We also assumed that these extremes would occur in July, August, and September according to the history records in the study area [80,101]; however, they may occur earlier (June) or later (October) [33]. Thirdly, in future work, the Predictive Ecosystem Analyzer (PEcAn) toolbox could be used to constrain PFT-dependent parameters for reducing the ED-2.2 uncertainties through Bayesian model calibration and emulation with the aid of more field measurements [84,102,103,104]. PEcAn can perform ensemble analysis for producing a confidence interval on the model output. Fourthly, we combined monthly verification and annual validation for NPP in the study; however, the fast timescale NPP-related variables (e.g., monthly GPP) can be used to quantificationally evaluate the model performance once the eddy flux tower is available in the future [48].

5. Conclusions

The productivity of subtropical coniferous forests was examined in this study using the process-based ED-2.2 model to simulate the effects of varied severities of seasonal drought and/or warming. The findings demonstrate that the DBH-height allometric equation’s fitting accuracy is superior to that of the default equation in ED-2.2 after model calibration. The ED-2.2 model successfully represented the seasonal variable characteristics of NPP in subtropical coniferous forests. Over time, there was less of a discrepancy between MODIS-NPP and the simulated yearly NPP. Seasonal warming has a smaller impact on NPP than seasonal dryness. The NPP reduction rate changed over time (July–October) in response to seasonal drought and warming situations, respectively. With an average loss of 31.72%, the combined warming-drought scenario in October caused the NPP to decline more dramatically compared with either the drought-only or warming-only scenario alone. Seasonal drought does have a large legacy effect on the productive recovery of subtropical coniferous forests. With the increase in warming severity, the average values of canopy temperature increased, but NPP decreased. Seasonal drought and/or warming limit forest production through decreasing soil moisture and/or increasing canopy air space temperature, which impact on plant photosynthesis and productivity, respectively.