A Deciduous Forest’s CO₂ Exchange Within the Mixed-Humid Climate of Kentucky, USA

Abstract

Forests play a crucial role in carbon cycling, contributing significantly to global carbon cycling and climate change mitigation, but their capture strength is sensitive to the climatic zone in which they operate and its adjoining environmental stressors. This research investigated the carbon dynamics of a typical deciduous forest, the Daniel Boone National Forest (DBNF), in the Mixed-Humid climate of Kentucky, USA, employing the Eddy Covariance technique to quantify temporal CO₂ exchanges from 2016 to 2020 and to assess its controlling biometeorological factors. The study revealed that the DBNF functioned as a carbon sink, sequestering −1515 g C m⁻² in the study period, with a mean annual Net Ecosystem Exchange (NEE) of −303 g C m⁻²yr⁻¹. It exhibited distinct seasonal and daily patterns influenced by ambient sunlight and air temperature. Winter months had the lowest rate of CO₂ uptake (0.0699 g C m⁻² h⁻¹), while summer was the most productive (−0.214 g C m⁻² h⁻¹). Diurnally, carbon uptake peaked past midday and remained a sink overnight, albeit negligibly so. Light and temperature response curves revealed their controlling effect on the DBNF trees’ photosynthesis and respiration. Furthermore, clear seasonality patterns were observed in the control of environmental variables. The DBNF is a carbon sink consistent with other North American deciduous forests.

1. Introduction

Forests, occupying 33.2% of the United States land area, remove an estimated 800 MT of Carbon dioxide (CO₂) annually [1,2,3]. This CO₂ is absorbed from the atmosphere into the forests’ biomasses through photosynthesis, demonstrating their crucial role in controlling global warming and climate change in the United States. The significance of forests and climate interaction has sparked a sustained interest in comprehending and measuring carbon cycling and ecological processes that hold regional importance, especially at fine scales. Due to varying climatic conditions and unrepresentative and inadequate long-term data at the ecosystem scale, significant variations and uncertainties are expected when the interaction is measured in the context of exchange rates of CO₂ in forests. To better study this, there is a need for improved CO₂ flux monitoring of differing climatic zones at various time scales, as suggested by several studies [4,5,6,7], which will be critical in achieving a comprehensive scientific understanding.

CO₂ flux is actively driven by the extent of ecosystem photosynthesis and respiration processes, which in turn are influenced by biophysical factors. In the literature, many biophysical variables have been established in their control of these processes. These include photosynthetically active radiation (PAR) [8], air temperature [9,10], soil temperature [11], and soil moisture, which is responsible for increased photosynthesis; however, when it is denatured, it leads to low carbon uptake [12,13]. Studies [8,14,15,16] have revealed that light, a key component of the photosynthetic process, increases the amount of carbon produced up to the saturation point. Other factors include salinity, which controls the physiological development of trees by modifying the production of carbon through water loss, CO₂ concentration, conductance of stomatal pores, and expansion of leaf surfaces [16].

Nevertheless, an aggregating factor—the prevailing climate—exerts a greater control on the level of carbon productivity of ecosystems operating within it. The length, mixture, and variability of these climatic influences, based on their daily, annual, and seasonal patterns, have a unique influence on the plant phenology and overall forest carbon exchange potential [17,18,19,20]. Several studies have established the responses of vegetation phenology to differing climates and have successfully linked the responses to their carbon exchange [21,22,23]. Changes in climate and the prevailing climate, such as earlier spring starts, warming temperatures, and altered precipitation, may affect leaf emergence, flowering, leaf senescence, and other plant-ecological cycles critical to their survival [22,23].

The Mixed-Humid climatic zone of Kentucky, USA, has unique ecological characteristics and holds immense potential for carbon sequestration. However, the influence of its patterns and controlling influence on ecosystem productivity processes is largely quantitatively understudied. Kentucky has a significant forest coverage (48.8%, 12.4 million acres) and experiences ample precipitation year-round, with high humidity and infrequent frost but a cold winter. The state is humid subtropical and located in a Mixed-Humid transition zone [24,25]. It is characterized by a mix of the northern cool season influenced by the frigid winter winds from the Canadian plains and the southern warm season, making it a unique ecological zone. The deciduous forests in this zone experience cold winter temperatures, which begin in October, marking leaf shedding the onset of acclimation as the trees try to conserve limited nutrients and soil moisture [26], which affect the extent of photosynthesis and respiration. This is then followed by spring months with higher precipitation and the hot summer months of Kentucky [24,25]. The cumulative effect of these changes on the forest causes significant variations in their annual productivity [27,28], and the joint understanding of this alongside the stressors highlights the resilient nature of the deciduous forests in Kentucky. The prevailing climatic conditions there exert substantial control over the CO₂ exchange rate and magnitude of carbon exchanged in North American forests [9,29,30,31]. For instance, with increased spring temperatures, forests in North America have shown an early onset of CO₂ uptake [32,33] and increased NEE and GPP [32,34].

This study, therefore, employed the eddy covariance (EC) method to measure how a deciduous forest exchanges CO₂. While similar studies have been conducted in and around the United States [35,36,37,38,39,40], none have been conducted in the Mixed-Humid climatic zone of Kentucky. This study examined the net ecosystem exchange (NEE) of CO₂ in the Daniel Boone National Forest of Kentucky from 2016 to 2020 across temporal scales (daily, monthly, and annually) to establish its carbon status as a sink/source. It also evaluated coinciding meteorological data to determine the degree of predictability of these factors on ecosystem exchange variables.

2. Materials and Methods

This study was conducted in the Daniel Boone National Forest (DNBF) of Kentucky (Figure 1a), a federal forest established in 1937, contained within the 8498.4 km² proclamation boundary, all originating from the Cumberland Plateau of the Appalachians. Covering an expansive 2865.2 km², the DNBF boasts a diverse array of deciduous and evergreen tree species, including American beech (Fagus grandifolia), sugar maple (Acer saccharum), white pine (Pinus strobus), hemlock (Tsuga canadensis), various oak species (Quercus sp.), and hickory (Carya ovata) [41]. The prevailing wind at the site is southeasterly at 3.6–5.7 m s⁻¹ (Figure 1d), and the annual mean temperature throughout the study period (2016–2020) was 13.68 °C with a mean annual precipitation of 1394 mm.

The eddy flux system used for this research was installed on the 30.48 m self-supporting Pea Body Communication tower (Figure 1b) within the DBNF used by the Forest Service near the Redbirds neighborhood, Clay County, Kentucky (37°08′17.16″ N, 83°34′46.82″ W, 743 m above sea level) (Figure 1a). The flux system instrumented on the tower comprised a 3-dimensional sonic anemometer (Gill Wind Master Pro), an open-path infrared gas analyzer (Licor 7200, LICOR, Lincoln, NE, USA) for CO₂ analysis, a net radiator, a humidity sensor, a rain gauge, and CO₂ and biomet data loggers for data collection (Figure 1b). The gas analyzer was oriented to achieve maximum CO₂ fetch from the forest within a catchment area of 5.76 km. Following Kljun et al. [42], footprint analysis showed that ~90% of fluxes are attributable to vegetation within ~600 m of the flux sensor (Figure 1c). Vegetation cover within the catchment area is mostly deciduous (84.24%), followed by mixed forest (2.03%), pasture/hay (2.55%), shrub (2.55%), and grassland/herbaceous (1.28%). The primary height class of trees within the footprint ranged from 13 to 30 m (99.65%) [43]. The tower footprint is characteristic of the entire DBNF forest [41].

The flux tower operates based on measuring the exchange of CO₂ using the eddy covariance method. This method is crucial for understanding the role of local ecosystems in global carbon dynamics [44,45]. It is governed by the following equation:

$$\overline{WC}=\overline{(\overline{w}+w^{\prime })(\overline{c}+c^{\prime })}=\overline{w}\overline{c}+\overline{c^{\prime }{w}^{\prime }}$$

(1)

where w is vertical wind speed, c = trace gas concentration, overbar denotes mean, and ′ indicates the instantaneous turbulence. It works by averaging the product of two variables, which is equivalent to the product of the averages of the means ($\overline{w}\overline{c}$) and the product of the averages of the fluctuating components ($\overline{c^i{w}^i}$) (Equation (1)).

The prevailing wind at the site is southeasterly at 3.6–5.7 m s⁻¹ (Figure 1c). The annual mean temperature throughout the study period (2016–2020) was 13.68 °C with a mean annual precipitation of 1394 mm.

To calculate daily, seasonal, and annual integrated magnitudes of NEE, it is essential to address gaps in the data series. Missing NEE values were addressed based on the duration of the gap, generating values. In instances where daytime 30-min data was absent, the daytime net ecosystem exchange (NEE day) was modeled using an empirically determined light response curve employing a Michelis–Menten rectangular hyperbola (Equation (2)) [46,47,48].

$${NEE}_{day}=R_E+{\displaystyle \frac{\phi \times \alpha \times P_{max}}{\phi \times \alpha +P_{max}}}$$

(2)

where P_max is the maximum ecosystem CO₂ uptake rate (μmol CO₂ m⁻²s⁻¹) at saturated light intensity, R_E is the daytime ecosystem respiration (μmol m⁻² s⁻¹), $\phi$ is photosynthetic photon flux density PPFD (μmol m⁻²s⁻¹), and $\alpha$ is the apparent quantum efficiency (μmol CO₂ m⁻²s⁻¹PPFD⁻¹).

During the night, carbon uptake is negligible in the absence of light; therefore, NEE at night time is assumed to be equivalent to night time ecosystem respiration R_E. Moreover, we assume that the total ecosystem R_E at night results from a combination of both heterotrophic and autotrophic sources [49,50,51]. Then, the missing nighttime R_E values were filled using a respiration model derived from Lloyd and Taylor (1994) [49] (Equation (3)).

$${NEE}_{night}=R_E=a_0^{\left(a_1{T}_s\right)}$$

(3)

Here, a₀ and a₁ (both in °C) represent empirical coefficients fitted for each growing season. a₀ signifies the intercept of R_E soil respiration when T_s is 0 °C, indicating the base respiration rate at 0 °C. Meanwhile, a₁ reflects the sensitivity of R_E across a range of temperatures.

The fitted model is specified below:

$$R_E={\beta }_0{exp}^{{\beta }_1\times T}$$

(4)

where ${\beta }_0$ and ${\beta }_1$ are coefficients, T is temperature.

$$Q_{10}={\left({\displaystyle \frac{R_2}{R_1}}\right)}^{10/(T_2-T_1)}$$

(5)

Q₁₀ is the factor by which the rate of respiration (reaction) increases for every 10 °C increase in temperature.

In both day and night conditions, when data gaps were less than one hour, the 30-min averages of the preceding and following time steps were averaged (i.e., interpolation) and used to fill the missing values. Under these assumptions, daytime R_E is temperature-dependent, and the nighttime R_E vs. T_s relationship can extend to daytime conditions (i.e., nighttime NEE (NEE_night) = R_E). Ultimately, Gross Primary Productivity (GPP) was computed by subtracting NEE from R_E (GPP = R_E − NEE). The gap-filled data was then utilized to segregate NEE into its components and calculate seasonal and annual integrated values of NEE, R_E, and GPP [50,51].

Before gap-filling was completed, low-quality data was removed from the necessary EddyPro^® (v.7.0.9) output file. The decision was made to remove all values for CO₂ flux, LE, and H, whose quality flag was a “1” or a “2”, to ensure that only the highest quality data was included. Due to the biological exchange capacity of forests in a humid/tropical climate and to ensure that values observed are representative of target vegetation in the zone, empirical decisions were made to remove and gap-fill CO₂ flux values beyond ±40 μmol/m⁻²s⁻¹ [52,53,54]. These removed values were considered above the established threshold [53,54] for temperate deciduous forests, and are largely suspected to be due to instrument and/or measurement error [52]. Also, values for LE beyond 0 to 700 wm⁻², and for H beyond −100 to 700 wm⁻², were filtered out as apparent outliers. This followed findings from energy balance studies [55,56,57] in similar ecosystems, which showed that LE and H fluxes in humid tropical and temperate forests rarely exceed these ranges. The REddyProc software (v.1.3.2) fills gaps in both eddy covariance data and inputted meteorological (biomet) data using similar methods to Falge et al. (2001) [48], with the addition of considerations of co-variation of fluxes with meteorological variables as well as the temporal autocorrelation of fluxes [58].

Flux calculations with insufficient turbulence are indicated by low friction velocity (u*). Minimum friction velocity u* is estimated using a method described in [54]. These are removed and also filled using the methods above. Data partitioning into Gross Primary Productivity (GPP) and ecosystem respiration (Re) was also done by REddyProc using daytime and nighttime estimation methods by Reichstein, 2005 [58] and Lasslop, 2010 [59], respectively.

3. Results

3.1. Weather and Climate

The average annual air temperature of the study site is 13.68 °C and is peculiar to the temperate climate of Kentucky, with the year 2017 having the highest recorded air temperature (13.81 ± 7.1 °C) (Figure 2a). The highest rainfall was recorded in the year 2018, with a total precipitation of 1552 mm. On a monthly basis, as shown in Figure 2b, average air temperature followed seasonal patterns, with the lowest temperature range recorded in January (2.2 ± 1.9 °C) characteristic of winter months and July being the hottest month on average (23.9 ± 20.5 °C).

3.2. Ecosystem Carbon Dioxide Fluxes

The Net Ecosystem Exchange (NEE) of CO₂ over the study period of 5 years ranged from −1.670 to 1.730 g C m⁻² h⁻¹ at the DBNF, with a mean annual NEE of −0.038 g C m⁻² h⁻¹ and −0.927 g C m⁻²dy⁻¹. Ecosystem Respiration (RE) ranged from 0.00 to 0.886 g C m⁻² h⁻¹ with a mean daily rate of 4.205 g C m⁻²dy⁻¹. Gross Primary Productivity (GPP) annual mean was 0.214 g C m⁻² h⁻¹ and 5.132 g C m⁻²dy⁻¹.

The daily NEE rate for winter was 0.069 g C m⁻² h⁻¹, −0.023 g C m⁻² h⁻¹ in spring, 0.034 g C m⁻² h⁻¹ in fall, and highest in summer (−0.214 g C m⁻² h⁻¹). For gross primary productivity rate, summer also has the highest (0.511 g C m⁻² h⁻¹), followed by spring (0.188 g C m⁻² h⁻¹), then fall (0.141 g C m⁻² h⁻¹), while the lowest rate is observed in winter (−0.012 g C m⁻² h⁻¹). The ecosystem respiration rate was the highest in summer (0.297 g C m⁻² h⁻¹), followed by fall (0.175 g C m⁻² h⁻¹), spring (0.165 g C m⁻² h⁻¹), and the lowest rate was also recorded in winter (0.056 g C m⁻² h⁻¹).

Kruskal–Walis One-Way ANOVA revealed that there are significant seasonal differences among the four seasons in the average daily CO₂ exchanges for NEE, RE, and GPP at p < 0.001 with moderate effect sizes (

Table 1. Non-parametric Kruskal–Wallis test for seasonal differences among four distinct seasons at the Daniel Boone National Forest over the study period.

	χ2	df	p	ε2
NEE	6766	3	<0.001	0.0863
RE	33,970	3	<0.001	0.4332
GPP	22,488	3	<0.001	0.2868

Note: χ² = Chi-square statistic that measures the relationship between observed and expected data; df = degrees of freedom; p = p-value, indicating the probability of observing χ² statistic under the null hypothesis at 5% significance level; ε² = epsilon squared indicating effect size, which is a measure of explained variance.

). The Dwass–Steel–Critchlow–Fligner pairwise comparisons revealed that apart from spring–fall RE and GPP insignificant comparisons, all seasons’ comparisons were significantly different in terms of the distribution of NEE, RE, and GPP at p < 0.05.

3.3. Cumulative Ecosystem Exchange

Over the five-year period (2016–2020), the forest biome’s annual carbon dynamics, as revealed by Gross Primary Productivity (GPP), Respiration Expenditure (RE), and Net Ecosystem Exchange (NEE), provide invaluable insights into its role as a carbon sink. The detailed observations, spanning a full year in most cases, showed unique patterns in carbon flux. In 2016 (DOY 1-366), a complete year of observations showcased high carbon sequestration, establishing the forest as an extremely high carbon sink with a GPP of 1756 gCm⁻²yr⁻¹, an RE of 1353 gCm⁻²yr⁻¹, and a resulting NEE of −403 gCm⁻²yr⁻¹ (

Table 2. Gross Primary Productivity (GPP), Ecosystem Respiration (RE), and Net Ecosystem Exchange (NEE) (gCm⁻²yr⁻¹) distribution by year.

Year	NEE	RE	GPP	RE/GPP	Status
2016	−403	1353	1756	0.771	Sink
2017	−364	1429	1793	0.797	Sink
2018	−144	1259	1403	0.897	Sink
2019	−254	1617	1871	0.864	Sink
2020	−350	1212	1562	0.776	Sink
Total (Mean)	−1515 (−303)	6870 (1374)	8385 (1677)		Sink

). Subsequent years, including 2017 (DOY 1-365) and 2019 (DOY 1-365), maintained consistent carbon sink characteristics, with 2019 standing out as the most productive in photosynthetic carbon production. Despite fewer observation days in 2020 (DOY 1-256, 338–366), the forest sustained its sink status, highlighting its resilience. On average, the forest upheld its role as a carbon sink, with a mean GPP of 1677 gCm⁻²yr⁻¹, a mean RE of 1374 gCm⁻²yr⁻¹, and a mean NEE of –303 gCm⁻²yr⁻¹. Cumulative totals over the five years underscore the consistent carbon sink nature of the forest biome, and the calculated standard deviation and standard error metrics enhance our understanding of the reliability of these measurements.

Daily sums of RE, GPP, and NEE were plotted by calendar year (Figure 3). Aside from the missing values that occurred as a result of machine failure, inclement weather, and unfilled data, all the years under consideration showed expected patterns. The daily values for GPP, NEE, and RE showed a minimal difference, revolving around 0 in the winter months, peaking in the summer, and decreasing towards the end of the year. The forest is a carbon source in the winter months as NEE is above GPP but changes to a sink as the year progresses, with the highest carbon sequestration occurring in the summer days. Both GPP and RE move in the same direction in all the years under consideration. The cumulative GPP, NEE, and RE values were 1677 g C m⁻² yr⁻¹, −303 g C m⁻² yr⁻¹, and 1374 g C m⁻² yr⁻¹, respectively (Figure 4a). Although the GPP curve was higher than RE, both cumulatively kept increasing throughout the study period. The NEE followed a similar but negative trend, indicating that as the year goes by, the DBNF gains more carbon from the atmosphere than it loses, hence being a net carbon sink. It is also noteworthy that the fluctuating nature of the curves represents the seasonality of the observations of these critical CO₂ exchange variables. Figure 4b shows the variability and central tendency of observed fluxes. The range for GPP was widest followed by RE, with NEE showing the tightest dispersion. Also, the median GPP was higher than that of RE, which shows that the DBNF was more productive as the negative median NEE confirms this. The non-zero and non-positive upper whisker line of NEE confirmed that the site was never a carbon source in any of the years under study.

3.4. Seasonal Distribution of Gross Primary Productivity (GPP), Ecosystem Respiration (RE), and Net Ecosystem Exchange (NEE)

Examining the year’s carbon dynamics reveals distinctive variations in Net Ecosystem Exchange (NEE), Ecosystem Respiration (RE), and Gross Primary Productivity (GPP) monthly, as shown in Figure 5. From January to April, DBNF consistently had carbon source NEE rates (0.024 to 0.034 g C m⁻² month⁻¹).

In January, the ecosystem experiences minimal RE (0.025 g C m⁻² month⁻¹), highlighting a period of low carbon use, while GPP is also at a low rate (0.001 g C m⁻² month⁻¹). February sees an uptick in RE (0.034 g C m⁻² month⁻¹) and GPP (0.009 g C m⁻² month⁻¹), setting the stage for higher carbon exchange. March presents a unique scenario as GPP rate becomes slightly negative (−0.004 g C m⁻² month⁻¹) and RE reduces (0.030 g C m⁻² month⁻¹), representing an 11.8% fall. As spring progresses, April becomes noteworthy, having higher RE (0.079 g C m⁻² month⁻¹) and GPP (0.049 g C m⁻² month⁻¹), signifying higher metabolic activity and carbon assimilation. May marks a shift, with NEE becoming negative, indicating a carbon sink period despite high RE and GPP values.

In the summer months, June is a significant carbon sink period, with the highest NEE rate (−0.127 g C m⁻² month⁻¹), accompanied by RE (0.132 g C m⁻² month⁻¹), showcasing substantial metabolic activity. July is a crucial period for the DBNF with the highest RE rate (0.174 g C m⁻² month⁻¹), GPP (0.287 g C m⁻² month⁻¹), and, consequently, NEE (−0.113 g C m⁻² month⁻¹), indicating the peak of summer carbon dynamics. August sustains high RE and GPP values, maintaining the carbon sink status. September witnesses a decrease in NEE (−0.031 g C m⁻² month⁻¹), indicating a reduced carbon assimilation rate. October signifies a shift towards a carbon source period, with RE (0.106 g C m⁻² month⁻¹) and NEE (0.033 g C m⁻² month⁻¹) remaining relatively high. November sees a further carbon loss with a positive NEE rate (0.052 g C m⁻² month⁻¹). December records the lowest RE rate (0.027 g C m⁻² month⁻¹), GPP (−0.025 g C m⁻² month⁻¹), and NEE (0.052 g C m⁻² month⁻¹), marking the least active period in carbon exchange. The variability of the monthly and total values of these exchanges reveals how sensitive they are to seasonal stressors.

3.5. Diurnal Pattern of Gross Primary Productivity (GPP), Ecosystem Respiration (RE), and Net Ecosystem Exchange (NEE)

The diurnal pattern of ecosystem exchange at the Daniel Boone National Forest, as reflected in Figure 6, reveals the dynamic diurnal carbon dynamics over 30-min intervals. The average rate at which carbon is fixed into plant biomass is −0.019 g C m⁻² 30-min⁻¹. In the early morning, from 12:00 a.m. to 6:30 a.m., positive Net Ecosystem Exchange (NEE) values indicate a carbon source, with the forest losing more carbon than it sequesters. Particularly at 6:30 a.m., the forest experiences a substantial loss, as respiration surpasses photosynthetic gain. Notably, the morning period from 7:00 a.m. to 9:00 a.m. witnesses the lowest carbon production fixation into tree biomass, emphasizing a phase of minimal growth.

As the day progresses, a significant shift occurs at 11:00 a.m., where the NEE rate becomes negative, signifying the forest is acting as a carbon sink. This transition continues until 3:00 a.m., with the highest rate of carbon sequestration observed at 3:00 p.m. (−0.151 g C m⁻² 30-min⁻¹), marking the most productive period of the day. Ecosystem respiration peaks around 7:30 p.m. (0.097 g C m⁻² 30-min⁻¹) and dominates the rest of the night, indicating increased carbon use, which is essential for maintaining structural balance and system functions.

This diurnal pattern, though showing the characteristic U-shape, offers crucial insights into the intricate balance between carbon sequestration and release, and a comprehensive understanding of the forest’s diurnal carbon exchange dynamics.

3.6. Environmental Control of Carbon Dioxide Exchange

3.6.1. Monthly Light Response Curves (Michelis–Menten Curves)

The Michaelis–Menten (light response) model was used to assess the response of Net Ecosystem Exchange (NEE) to light intensity (PPFD) for a typical year in the study period. 2016 was selected because it had a complete dataset, with average temperature and precipitation typical for the area under study. The maximum photosynthetic rate at saturated light (A_max) was lowest in the winter months and peaked in the growing season, with July (3.205 g C m⁻² s⁻¹) being the highest (

Table 3. Monthly curve fit parameters for light response curve (2016). A_max is the maximum ecosystem C uptake rate (g C m⁻² s⁻¹) at saturated light intensity, R_e is the daytime ecosystem respiration (g m⁻² s⁻¹), PPFD is photosynthetic photon flux density PPFD (μmol m⁻² s⁻¹), and ∝ is the apparent quantum efficiency (g C m⁻² s⁻¹ PPFD⁻¹).

Month	Amax	α	Re (gCm−2)	Total NEE (gCm−2)	Mean NEE (gCm−2)	Mean PPFD (µMol/m2)	n	R2 (%)
January	0.000	0.627	0.017	8.812	0.017	494.87	509	0.00
February	0.000	0.588	0.002	1.341	0.002	505.95	556	0.00
March	0.000	0.744	0.025	16.672	0.025	680.87	677	43.01
April	1.511	0.617	1.507	4.093	0.006	839.59	726	38.18
May	0.583	0.001	0.044	−169.581	−0.213	792.08	796	87.29
June	2.656	1.386	3.431	−175.644	−0.216	986.26	814	45.07
July	3.205	1.304	2.947	−191.990	−0.234	865.02	822	39.21
August	1.736	0.676	1.548	−134.826	−0.174	816.94	776	39.50
September	1.246	0.699	1.106	−91.443	−0.134	859.99	682	44.73
October	0.222	1.769	0.175	−30.087	−0.047	693.53	639	46.59
November	0.672	1.286	0.691	10.995	0.020	557.66	542	45.81
December	0.000	0.572	0.018	8.943	0.018	380.27	498	49.87

). The Amax pattern also coincided with the apparent quantum efficiency (∝) distribution and the amount of carbon expended for respiration. The months with high ∝ and A_max showed a clear link with the amount of photosynthetically active light, with mean PPDD highest at those months. In particular, June had the highest PPFD (986.26 µMol/m²), the highest RE rate (3.431 gC m⁻²s⁻¹) (Figure 7), and the highest quantum efficiency (1.386), and also high total NEE (−175.644 gCm⁻²). This indicates that the seasonal availability of light and its intensity variability influences the amount of carbon uptake and carbon usage for the forest ecosystem.

3.6.2. Monthly Temperature Response Curves for the Year 2016

The dependence and reaction response of respiration on air temperature was assessed following the exponential temperature-response curve of the Lloyd and Taylor [49] approach. The fitted model is specified below.

$$R={\beta }_0 {exp}^{{\beta }_1\times T}$$

(6)

where ${\beta }_0$ and ${\beta }_1$ are coefficients and T is temperature.

$$Q_{10}={\left({\displaystyle \frac{R_2}{R_1}}\right)}^{10/(T_2-T_1)}$$

(7)

Q₁₀ is the factor by which the rate of respiration (reaction) increases for every 10 °C increase in temperature.

The fitted parameters of the temperature response curves for respiration in the study area for the year 2016 (Equation (6)) are shown in

Table 4. Monthly curve-fit parameters for temperature response curves for 2016.

	β0	β1	Q10 raw data	Q10 modeled	Total RE (gCm−2)	Mean RE (gCm−2)	Mean Tair (°C)	R2 (%)	n
Jan	0.010	0.060	2.296	1.825	8.837	0.009	−2.06	92.67	977
Feb	0.011	0.010	1.914	1.108	8.110	0.010	2.71	99.82	836
Mar	0.010	0.039	1.513	1.475	12.313	0.015	8.49	98.05	811
Apr	0.018	0.073	3.083	2.083	30.468	0.043	10.81	91.90	714
May	0.056	0.034	1.474	1.410	62.458	0.090	13.54	99.31	693
Jun	0.026	0.061	1.766	1.834	53.504	0.085	19.14	97.12	626
Jul	0.048	0.064	2.223	1.895	126.243	0.190	21.43	94.10	666
Aug	0.037	0.066	2.121	1.944	111.418	0.156	21.75	98.80	712
Sep	0.052	0.044	1.598	1.556	86.482	0.114	17.53	98.74	758
Oct	0.051	0.037	1.665	1.447	69.696	0.082	12.47	98.57	849
Nov	0.019	0.060	1.799	1.817	26.964	0.030	6.22	95.23	898
Dec	0.014	0.037	2.190	1.445	16.092	0.016	2.77	97.35	990

and Figure 8. The intercept variable (${\beta }_0$), which indicates the respiration rate at 0 °C, showed that for all the months, respiration occurred even at low temperatures. The ${\beta }_0$ improves as the season shifts towards the middle of the year. The slope parameters (${\beta }_1$) followed a similar pattern as ${\beta }_0$. The monthly temperature coefficients (Q₁₀) shown in Equation (7) demonstrate that respiration rates almost doubled (1.65 ± 0.29) with every 10 °C increase.

The annual Q₁₀ analysis also revealed a similar pattern, with the temperature coefficient averaging 2.1 ± 0.12 (

Table 5. Curve fit parameters for temperature response curve for all years.

	Q10 modeled	Total RE (gCm−2)	Mean RE (gCm−2)	Mean Tair (°C)	Total GPP	Mean GPP	n
2016	2.697	612.57	0.064	10.30	1483.40	0.185	9530
2017	1.881	674.82	0.071	10.67	1416.17	0.176	9473
2018	2.002	620.09	0.092	10.19	1212.59	0.226	6770
2019	1.992	774.86	0.081	16.94	1592.45	0.199	9519
2020	2.046	533.33	0.073	16.12	921.51	0.185	7267

). This means that, on average annually, the forest respiration rate doubles every 10-degree temperature increase.

3.6.3. Correlational Analysis

Pearson correlation analysis for the ecosystem exchange variables revealed that latent heat (LE) and evapotranspiration (ET) had the strongest negative correlation to NEE (−58% and −48% respectively), with the same (LE, 61%, ET, 48%) being true for GPP, but positively correlated (

Table 6. Pearson correlation between GPP, RE, and NEE. Blue and red color gradients show the intensity of positive and negative correlations, respectively.

	NEE	RE	GPP
LE	−0.58 **	0.32 **	0.61 **
VPD	−0.16 **	0.19 **	0.21 **
Tair	−0.30 **	0.60 **	0.48 **
ET	−0.48 **	0.24 **	0.50 **
PPFD	−0.39 **	0.25 **	0.42 **
H	−0.22 **	0.00	0.19 **

** Correlation is significant at the 0.01 level (2-tailed).

). However, air temperature (Tair, 60%) had the highest positive correlation to RE. All significant variables for NEE, RE, and GPP were significant at the 1% level.

3.6.4. Seasonal Regression (Four Seasons)

In the winter months, correlation analysis showed—see

Table 7. (a–d) Pearson correlation between flux variables and biomet variables for (a) winter months (December, January, and February); (b) spring months (March-May); and (c) summer months (June–August); and (d) fall months (September–November) for the study period.

Season		NEE	Reco	GPP
(a) winter	LE	0.19 **	0.02 **	−0.19 **
	VPD	0.06 **	0.24 **	−0.01
	Tair	0.04 **	0.49 **	0.06 **
	ET	0.27 **	0.03 **	−0.26 **
	PPFD	−0.08 **	0.09 **	0.10 **
	H	−0.01	−0.01	0.01
(b) spring	LE	−0.51 **	0.35 **	0.55 **
	VPD	−0.06 **	0.22 **	0.13 **
	Tair	−0.21 **	0.57 **	0.36 **
	ET	−0.38 **	0.27 **	0.42 **
	PPFD	−0.29 **	0.19 **	0.31 **
	H	−0.20 **	−0.01	0.17 **
(c) summer	LE	−0.67 **	0.08 **	0.65 **
	VPD	−0.25 **	0.16 **	0.29 **
	Tair	−0.28 **	0.23 **	0.34 **
	ET	−0.59 **	0.07 **	0.58 **
	PPFD	−0.43 **	0.12 **	0.45 **
	H	−0.58 **	0.06 **	0.57 **
(d) fall	LE	−0.47 **	0.21 **	0.49 **
	VPD	−0.30 **	0.12 **	0.30 **
	Tair	−0.28 **	0.45 **	0.44 **
	ET	−0.41 **	0.16 **	0.42 **
	PPFD	−0.37 **	0.17 **	0.39 **
	H	−0.24 **	−0.01	0.20 **

** Correlation is significant at the 0.01 level (2-tailed). Blue and red color gradients show the intensity of positive and negative correlations, respectively.

(a)—that Net Ecosystem Exchange (NEE) has weak negative correlations with photosynthetic photon flux density and specific heat, while the relationship is positive but weak with latent heat, vapor density, and air temperature. A moderately significant and positive correlation was witnessed between Ecosystem Respiration (RE) and Tair (49%), while the rest were weak: VPD (24%), PPFD (9%), ET (3%), and LE (2%). For Gross Primary Production (GPP), a significantly weak negative correlation was observed with ET (−26%) and LE (−19%), while Tair (6%) had a significantly positive but weak correlation with GPP. All significant variables for NEE, RE, and GPP were significant at the 1% level.

For the spring months, NEE negatively correlates with LE (−51%), ET (−38%), PPFD (−29%), Tair (−21%), and H (−20%) (Table 7(b)). However, the reverse is the case for RE and GPP. For RE, Tair (57%) has a moderate correlation, followed by LE (35%), then ET (27%), and then VPD (22%). GPP was moderately correlated with LE (55%), ET (42%), Tair (36%), and PPFD (31%).

In the summer months, correlation analysis revealed that NEE has statistically significant negative correlations with all the biomet variables, with moderate relationships witnessed for latent heat (LE, −67%), evapotranspiration (ET, −59%), PPFD (−43%), and specific heat (H, −58%) (Table 7(c)). For RE, all of the correlations are positive and significant, but weak, with the highest being Tair (23%). Likewise, GPP positively and significantly correlate with all the included biomet variables, with the highest being LE (65%), indicating a strong correlation, while ET, H, and PPFD were 58%, 57%, and 45%), respectively.

As shown in Table 7(d), correlation analysis revealed that NEE has statistically significant negative correlations with all the biomet variables (LE, −47%, ET, −41%, PPFD, 37%, VPD, −30%, Tair, −28%, and H, −24%), though mostly moderately so. For RE, Tair (45%) had the highest significance and was positively correlated. All biomet variables were positive and significant for GPP, with the highest being LE (49%), followed by Tair (44%), ET (42%), PPFD (39%), and VPD (30%).

4. Discussion

4.1. Ecosystem Scale Exchanges

Annually, the study site was a carbon sink throughout all the years studied, with a mean annual NEE of −303 gCm⁻²yr⁻¹. The year 2016 had an average air temperature of 13.74 °C, the third highest in the study period, and with the least precipitation, 1261.22 mm, had the highest amount of carbon (−403 gCm⁻²yr⁻¹) sequestered. The driest year (2017) had the second highest amount of carbon sequestered (−364 gCm⁻²yr⁻¹). This indicates that the optimum temperature and moderate precipitation of the higher latitude region may be the driver of carbon sequestration of deciduous forests in the region. This aligns with the positions of other studies of the region [35,36,40], which observed variations from −142 to −364 gCm⁻²yr⁻¹, with the exception of a study carried out at a cold (−3.2 °C) boreal forest in Canada, which was a carbon source (7 gCm⁻²yr⁻¹). This confirms that different forests have different NEE rates.

The year 2018 (the wettest year) had the lowest carbon sequestered (−144 gCm⁻²yr⁻¹). A plausible reason for this may be that the year had more cloud cover due to higher precipitation than other years. This would decrease the amount of ambient light penetrating the forest ecosystem and therefore light use efficiency, leading to lower photosynthetic productivity [35,60].

Further investigation of the literature across several regions also indicates that carbon sequestration potential is greatly modulated by the age of the forest, with very young forests—0 to 20 years—having higher NEE [61] compared to mature forests like that of this study (20–100 years), while old growth forests—above 100 years—have the lowest NEE rates [37,39]. Young forests’ trees’ rate of photosynthesis exceeds that of respiration comparatively, and therefore, more carbon is converted into energy and stored, hence resulting in higher carbon sequestration. However, as the forest grows, the NEE rate declines as the balance of gross photosynthetic productivity (GPP) to carbon respiratory use tilts to support more ecosystem respiration to support the tree structure. Also, respiration increases because of forest disturbances as the tree grows [62]. Therefore, the old growth trees expend more carbon to maintain their stands and store less carbon. Likewise, the extent of the interannual variability in NEE at similar colder regions can be influenced by forest management. Actively managed sites, as in a study in Iceland and Denmark [63], demonstrate that managed deciduous broadleaf trees can be a consistent carbon sink (−100 to −150 gCm⁻²yr⁻¹), even at low temperatures (

Table 8. Comparison of the annual NEE, RE, and GPP of deciduous forests worldwide.

Country	Forest Type	NEE (g C yr−1)	GPP (g C yr−1)	RE (g C yr−1)	Study Years	Location	Reference
USA	Mostly Deciduous	−303	1535	1374	2016–2020	37°8′17.16″ N, 83°34′46.82″ W	This study
“	Deciduous broadleaf	−250	1400	1150	-	42°32′24″ N, 72°10′12″ W	[40]
“	Mixed deciduous	−198	1297	1099	1993–2000	42.5 N, 72.2 W	[35]
“	“	−247	1400	1153	1992–2004	42.538 N, 72.171 W	[40]
Canada	Mixed deciduous & coniferous	−142	1118	976	1996–2003	44°19′ N, 79°56′ W	[39]
Brazil	Dry tropical	−168.7	414.7	246	2014	6°34′42″ S, 37°15′05″ W	[11]
“	“	−145	334	189	2015	“	“
Iceland	Managed deciduous broadleaf	−100	710	610	-	63°49′48″ N, 20°13′12″ W	[63]
Denmark	“	−150	1190	1040	-	55°28′48″ N, 11°38′24″ E	“
UK	Deciduous	−130	2110	1980	2007–2009	51°46′12″ N, 1°19′48″ W	[71]
“	Deciduous broadleaf	−486	2034	1548	1999–2010	51°07′12″ N, 0°51′00″ W	[72]
Germany	Deciduous broadleaf	−490	1560	1070	-	51°04′12″ N, 10°27′00″ E	[60]
Italy	Deciduous broadleaf	−660	1300	640	-	41°52′12″ N, 13°35′24″ E	[63]
Denmark	European beech	−157	1727	1570	1996–2009	55°29′13′′ N, 11°38′45′′ E	[73]
Japan	Deciduous broadleaf	−261	1118	857	2000	42°58′48″ N, 141°22′48″ E	[74]
“	“	−258	-	-	1999–2001	42°40′12″ N, 141°36′00″ E	[75]
“	“	−207	948	741	-	36°07′48″ N, 137°25′12″ E	[76]
“	“	−357	-	-	1997	35°52′12″ N, 139°28′48″ E	[77]
“	Deciduous	−236	978	742	1994–2002	36°08′ N, 137°25′ E	[78]

“ refers to some observation/item as above.

). Another reason for witnessing this annual NEE status is the seasonal pattern of forest carbon sequestration. In our study, a clear seasonal pattern of air temperature was observed. In the frigid winter months, ambient temperature and sunlight are low; hence, photosynthetic conversion is limited. Most deciduous trees in the area enter a dormancy period in this season by shedding their leaves to prevent water loss and using stored-up carbohydrates as an energy source [64,65,66]. The mean RE (117.62 g C m⁻² season⁻¹) was higher than the GPP (−25.76 g C m⁻² season⁻¹), leading to more carbon loss. During this period, more of the carbon stored in the previous season is expended. As the year progresses towards spring and summer seasons, the NEE rate also improves, peaking in summer (−468.16 g C m⁻² season⁻¹), where GPP is almost twice the ecosystem respiration on average. It is important to point out that within the summer months, NEE reached its peak (−0.254 g C m⁻² h⁻¹) in June (21.9 °C), before the hottest month, July (23.9 °C). This shows the level of heat stress the trees can withstand, as trees shut down at above-threshold temperatures, affecting both photosynthesis and carbon sequestration [67]. As temperature increases beyond an optimum point, cellular respiration increases, reducing carbon sequestration potential. However, as seen at the study site, the plant closes its stomata when air temperature becomes excessive [68,69,70] to conserve water for plant use, which invariably limits its CO₂ absorption and, eventually, causes GPP to start to decline (July). As summer temperatures dwindle, so too does the photosynthetic rate.

Diurnal exchange patterns can also explain carbon dynamics. On a diurnal basis, the DBNF sequesters carbon towards noon as respiration was more than photosynthesis in the early hours of the day, causing carbon release. In the absence of sunlight, respiration dominates, with an average rate of 0.172 g C m⁻² h⁻¹, which extends into the early morning. However, as light intensity increases, photosynthesis dominates, and carbon uptake picks up [35,48].

4.2. Controlling Variables

The curve fitting of the Michelis–Menten light response curves and the temperature-response curve confirmed the light and air temperature’s limitation on the extent of ecosystem processes, as critically acclaimed in the literature. This study’s findings revealed that light largely influenced the extent of NEE through gross productivity. This is true because photosynthesis occurs in the presence of light. It has been established that in the study area, this governs the year-round carbon productivity, although it is variable. The winter months had the lowest photosynthetic rates compared to the highly productive summer months. Also, light use efficiency for the deciduous forest was poor in the winter, as opposed to higher efficiencies in other months, except for May.

While we did not quantitatively assess the impact of cloud cover on photosynthesis in this study, the observed pattern in May 2016 aligns with established knowledge that cloud cover diffuses incoming shortwave radiation from reaching the trees, which is essential for photosynthesis [79,80]. We suspect excessive cloud cover due to higher precipitation in May of 2016 (217.97 mm) could have been responsible for the lower efficiency recorded. We acknowledge that the specific contribution of cloud cover to the observed patterns was not the focus of this study. Future studies should be directed at disentangling the effects of cloud cover, precipitation, and other climatic variables on ecosystem carbon exchange in mixed-humid/humid tropical climates like Kentucky to provide a more comprehensive understanding of these dynamics. However, canopy photosynthesis improved (−169.581 gCm⁻²) as more water was available in the soil for growth. This finding supports the argument [81] that moisture in the soil improves photosynthesis.

The curve fitting of the 2016 temperature response curves revealed the great influence air temperature has on respiration. Year-round respiration was non-zero, showing that metabolic processes occur all the time at the study area, even during the cold winter months, although at a lower rate. When examined, an exponential relationship was inferred from the response in terms of respiration to the temperature increase. The temperature sensitivity showed that respiration doubles with every 10-degree temperature increase for the deciduous trees of the Daniel Boone National Forest.

Correlational analysis revealed similar established patterns for light (PPFD) and air temperature (Tair) and their effect on NEE, RE, and PPFD. The correlations of other variables were examined, but biological significance cannot be inferred since their correlations are minimal and significant.

5. Conclusions

The study site has consistently been a carbon sink, with the highest amount of carbon sequestered in 2016, a year characterized by an average air temperature of 13.74 °C and the highest growing season precipitation of the studied years. This is because the total precipitation in the growing season 2016 (May–June, 666.72 mm) was higher than for the other years studied. This suggests that the higher latitude region’s optimum temperature and moderate precipitation may drive carbon sequestration in deciduous forests. However, the wettest year, 2018, had the lowest carbon sequestration, likely due to increased cloud cover reducing ambient light and photosynthetic productivity.

Seasonal patterns further impact carbon sequestration, with lower photosynthetic conversion in the chilly winter months and peak NEE rates in the summer, particularly in June, before the onset of heat stress. This demonstrates the resilience of trees in withstanding above-threshold temperatures, which affect photosynthesis and carbon sequestration.

Consistent with other climatic zones, the extent of NEE and GPP is driven by light (photosynthetically active radiation), while air temperature mainly influences ecosystem respiration. Seasonality exists in the extent of their influence, while other factors like precipitation and air dryness may account for variations not accounted for by these variables

This study, therefore, concludes that the study site at the Daniel Boone National Forest, which is mostly deciduous, is a carbon sink with a distinct seasonal and daily pattern highly influenced by ambient sunlight, air temperature, precipitation, and cloud cover. Being an aging forest, carefully planned management practices would help maintain this carbon status for longer.