Effects of Extreme Temperature and Precipitation Events on Daily CO 2 Fluxes in the Tropics

: The effects of anomalous weather conditions (such as extreme temperatures and precipitation) on CO 2 flux variability in different tropical ecosystems were assessed using available reanalysis data, as well as information about daily net CO 2 fluxes from the global FLUXNET database. A working hypothesis of the study suggests that the response of tropical vegetation can differ depending on local geographical conditions and intensity of temperature and precipitation anomalies. The results highlighted the large diversity of CO 2 flux responses to the fluctuations of temperature and precipitation in tropical ecosystems that may differ significantly from some previously documented relationships (e.g., higher CO 2 emission under the drier and hotter weather, higher CO 2 uptake under colder and wetter weather conditions). They showed that heavy precipitation mainly leads to the strong intensification of mean daily CO 2 release into the atmosphere at almost all stations and in all types of study biomes. For the majority of considered tropical ecosystems, the intensification of daily CO 2 emission during cold and wet weather was found, whereas the ecosystems were predominantly served as CO 2 sinks from the atmosphere under hot/dry conditions. Such disparate responses suggested that positive and negative temperature and precipitation anomalies influence Gross Primary Production (GPP) and Ecosystem Respiration (ER) rates differently that may result in various responses of Net Ecosystem Exchanges (NEE) of CO 2 to external impacts. Their responses may also depend on various local biotic and abiotic factors, including plant canopy age and structure


Introduction
The observed rapid growth in global air temperature, changes in precipitation, as well as an increase in the frequency and severity of extreme weather events can have a significant impact on the growth and development of plant communities in various geographical regions [1][2][3][4][5]. The extreme weather phenomena such as heat waves, catastrophic droughts and heavy precipitation may disturb the functioning of plant communities that is evidenced in changes in plant transpiration, Gross Primary Production (GPP), Ecosystem Respiration (ER), and Net Ecosystem Exchange (NEE) of CO 2 between land surface and the atmosphere [6][7][8]. Prolonged and intense extreme weather events can significantly reduce ecosystem resilience, biodiversity, and net primary production. Understanding and quantifying the sensitivity of various terrestrial ecosystems to such influences and the primary driving mechanisms is important for assessing climate impacts and developing effective adaptive strategies [9]. The tropics are characterized by the world's greatest biological diversity, with many species interacting in complicated and multifaceted ways [10]. The structure and productivity of plant ecosystems in the tropics are very sensitive to changes in energy and water balance, which could be caused by changes in precipitation and temperature [11][12][13][14][15][16][17]. The largest areas in the humid and semi-humid tropics are covered by forests that play an important role in the global climate system [11,18]. They regulate the exchange of greenhouse gases (GHGs) between the Earth's surface and the atmosphere, perform important water-regulating and water-saving functions, influence the radiation and energy budgets, determine the microclimate of the large territories, absorb and retain atmospheric carbon, sequestrate it in soil, and retain in an inactive state during large time intervals [19][20][21].
To assess possible atmospheric effects on plant ecosystems, the rates of evapotranspiration, GPP, ER, and NEE (determined as the difference between ER and GPP) can be considered as universal characteristics determining the living conditions and functioning of plant communities in different geographical areas [22,23]. The global GHG flux monitoring network based on the eddy covariance method includes, in the present time, over 1000 active and historical flux stations, with only a small part of them situated in tropical regions [24]. Around 25% of the stations have an operating period of more than 3 years. The GHG monitoring stations provide very useful and comparable information on the spatial and temporal variability of the fluxes, as well as the flux sensitivity to varying atmospheric conditions. It can be expected that the aggregated analysis of global GHG fluxes and meteorological conditions can provide new knowledge on the response of different terrestrial ecosystems in different geographical regions to atmospheric influences.
To assess the impact of anomalous climate and weather events (heat waves, frosts, droughts, floods, etc.) on GHG fluxes in various ecosystems in the tropics and midlatitudes over the last decades, numerous studies have been carried out [8,14,16,[24][25][26][27][28][29][30][31][32][33][34]. It was demonstrated that droughts and heat waves likely result in increased emission of CO 2 to the atmosphere, whereas the cold and wet conditions are mainly associated with the CO 2 uptake intensification [8]. Several studies have highlighted the precipitation deficit as a crucial factor leading to the abrupt decline of GPP [8,16,32]. However, the response of CO 2 exchange to the weather conditions is rather uncertain: the high temperature often causes the increased CO 2 uptake by terrestrial ecosystems [33], but, at the same time, leads to the intensification of soil CO 2 emission [34]. Despite a number of experimental and modeling studies of anomalous climate and weather events, there are still limited studies on the regional and global synthesis of the terrestrial ecosystem responses to atmospheric impacts. Given the diversity of possible feedbacks between terrestrial ecosystems and the atmosphere, we may suppose that new comprehensive knowledge of the plant-atmosphere interaction is essential for a better understanding of the changes in ecosystem functioning, productivity, and stability, as well as for predicting future climate change.
Another very important issue is the different response of ecosystems to atmospheric anomalies of different intensities and durations. There are several studies discussing the extreme threshold definition in regard to terrestrial ecosystem response to weather and climate anomalies [3]. Smith [35] proposed to consider events that simultaneously experience abnormal climatic conditions and natural ecosystems experience strong impacts that exceed normal variability. Frank et al. [3] introduced the term "extreme impact", when a resilience threshold ("extreme response threshold") is passed, placing the ecosystem and associated carbon cycling into an unusual or rare state. Threshold values are usually exceeded when a stressor dose (i.e., cumulative amount defined by stress intensity multiplied by stress duration) reaches a critical level (e.g., during waterlogging, drought, and/or extended periods of exceptionally high or low temperatures), or when the intensity of an extreme climatic event is critically high (e.g., during a storm). Thresholds can be overcome at a plant organ, individual plant, or entire plant community level, resulting in changes in the carbon and water budgets of the entire ecosystem. However, in the above-mentioned study, the specific threshold values for meteorological parameters are not defined. Meanwhile, the threshold values of meteorological parameters associated with the abrupt changes in the rates of evapotranspiration, GPP, ER, and NEE of CO 2 between ecosystems and the atmosphere may differ significantly between geographical regions. In this context, the question of identifying weather extremes in terms of their impact on plant communities, taking into account regional conditions and using different methods for extreme event assessment, needs further investigation.
The main goal of our study is to assess the differences in the response of daily net CO 2 fluxes between different tropical ecosystems and the atmosphere to extreme weather conditions (such as extreme temperatures and precipitation). As a working hypothesis, the study suggests that the daily CO 2 flux response of tropical vegetation can differ depending on local geographical conditions and intensity of temperature and precipitation anomalies. To solve this scientific problem, we are going to use available reanalysis data, as well as information about CO 2 fluxes from the global FLUXNET database [24].

Meteorological and CO 2 Flux Data Sets
To analyze the possible CO 2 flux feedbacks of tropical terrestrial ecosystems on the extreme temperature and precipitation events, 17 monitoring stations of CO 2 fluxes situated on different continents in tropical latitudes with various landscape and climate conditions and providing the open access data were selected. All stations were assigned to 4 main biome types [36] in accordance with ecosystem classification used in the FLUXNET archive: savannas and woody savannas, evergreen and tropical rain forests, dry or seasonal forests (include deciduous needle-leaf forests and tropical seasonal deciduous forests), and wetlands ( Figure 1). riods from 1 to 9 years. The archive included meteorological data, net radiation, CO2 fluxes, and latent and sensible heat fluxes measured using the eddy covariance method [38]. The eddy covariance method allowed for providing direct long-term measurements of turbulent GHG fluxes on an ecosystem scale. All FLUXNET stations used similar standardized equipment and data processing software to produce comparable time-series data on GHG fluxes. The gaps in the CO2 flux time series, caused by equipment and power failures, low wind, strong rainfalls, etc., for all selected flux stations were filled using the REddyProc package [39].

Data Analysis
The daily means were calculated from 3 h and 1 h reanalysis datasets on temperature and precipitation, respectively. In the next step, they were used to determine the To identify the extreme weather conditions, the reanalysis produced by the European center for medium-range weather forecast ERA5 [37] was used. We used the temperature at 2 m above ground with a temporal resolution of 3 h and precipitation amount with an hourly resolution. The spatial resolution of the data sets was 0.25 • × 0.25 • , and the period analyzed was 1991-2021. We believed that the reanalysis data set was more useful for our study, as they did not have gaps and were, therefore, able to provide a continuity of timeseries data and, therefore, better data comparability at different locations. The FLUXNET meteorological data set contained numerous gaps in temperature and precipitation data, ranging from 2 to 37% for selected stations. This made their application for spatial and temporal data analysis challenging. A correlation analysis revealed a strong agreement between the reanalysis and the gap-free FLUXNET data sets for the air temperature. The R-squared values for the temperature data sets ranged mainly from 0.69 (AU-Fog) to 0.96 (AU-Rob) at p < 0.05 (Supplementary Materials Figure S1). For MY-PSO and Br-Sa1 ( Figure S2), it was somewhat less (0.69 at p < 0.05). The agreement between the precipitation rates obtained from reanalysis and monitoring stations was worse due to the high spatiotemporal inhomogeneity of the precipitation fields and the numerous gaps in precipitation measurements at the flux stations. At the same time, it was noteworthy that the correlation between extreme precipitation (exceeding the thresholds) from reanalysis and FLUXNET data for the threshold 95% quantile at stations with a longer period of observation (more than 6 years) remained high, ranging from 0.45 to 0.73 at p < 0.05 ( Figure S3). The latter factor determined the appropriateness of reanalysis precipitation data in our study.
Information about net daily CO 2 fluxes (NEE) was taken from the FLUXNET archive [24] that had open access to historical flux data sets. Available datasets covered the periods from 1 to 9 years. The archive included meteorological data, net radiation, CO 2 fluxes, and latent and sensible heat fluxes measured using the eddy covariance method [38]. The eddy covariance method allowed for providing direct long-term measurements of turbulent GHG fluxes on an ecosystem scale. All FLUXNET stations used similar standardized equipment and data processing software to produce comparable time-series data on GHG fluxes. The gaps in the CO 2 flux time series, caused by equipment and power failures, low wind, strong rainfalls, etc., for all selected flux stations were filled using the REddyProc package [39].

Data Analysis
The daily means were calculated from 3 h and 1 h reanalysis datasets on temperature and precipitation, respectively. In the next step, they were used to determine the values of the air temperature and precipitation at the location of the FLUXNET station by averaging 4 adjacent grid points. It allowed for a reasonable approximation of meteorological parameters at selected flux tower locations, presented in continuous time series.
The temperature anomalies were calculated by removing the daily mean, calculated over the period 1991-2021. For the precipitation, the daily anomalies were not calculated as in the region characterized by lack of precipitation during some periods, the daily anomalies were not informative.
The mean daily net ecosystem CO 2 fluxes were calculated by averaging 30 min FLUXNET data sets. Gaps in the CO 2 flux measurement records were filled using standard algorithms provided by the REddyProc online tool, which was based on procedures described by Reichstein et al. [40]. The daily anomalies of CO 2 fluxes were calculated as the difference between mean daily and mean monthly CO 2 fluxes over the available observation period at each station.
The periods with extreme weather conditions were defined using two methods: − as a period when the daily mean temperature anomaly/daily precipitation amount exceeded 95% (90%) quantile (for extreme high temperature and precipitation) or did not reach 5% (10%) quantile (for extreme low temperature and precipitation) of probability density function (PDF); − as a period when the daily mean temperature anomaly exceeded one standard deviation (STD) calculated for each calendar month over the whole time series; − as the precipitation was not normally distributed, the threshold definition based on the STD exceeding was less informative. Therefore, only the quantile thresholds were applied for precipitation.
The PDFs were calculated for every month of the observation periods and averaged over 1991-2021. The normal distribution for the temperature and the Weibull distribution for the precipitation were used that were demonstrated to be the most appropriate for the continents in the tropics [41].
The extreme daily CO 2 flux anomaly was defined as an anomaly greater (lower) than 1 STD (−1 STD). The STD was calculated separately for each calendar month to exclude the influence of the seasonal cycle. The quantile thresholds were not applied to CO 2 fluxes because the PDF of CO 2 fluxes varied significantly between the ecosystem type that complicated the choice of the theoretical distribution to approximate the CO 2 flux data. Moreover, the short time series available at several stations also made it difficult to identify the appropriate theoretical PDF.

Temporal Variability of Daily Temperature, Precipitation, and CO 2 Fluxes in Various Tropical Terrestrial Ecosystems
Analysis of the variability of daily CO 2 fluxes, air temperature, and precipitation, as well as daily CO 2 flux feedbacks to the anomalous temperature and precipitation events in the tropical ecosystems, showed significant variability depending on geographical locations, landscape and climate conditions, and plant species composition.
Firstly, we focused on the temporal variability of the weather conditions (temperature and precipitation) and associated CO 2 flux anomalies. The daily flux data from 17 measurement stations in the tropics for the entire period of measurements were analyzed (see the examples for four selected biomes in the Supplementary Materials). As an example, Figure 2 shows the temporal variability of daily temperature, precipitation, and CO 2 flux anomalies for the measurement station located in the Australian savannas (Dry River, AU-Dry).
The time-series analysis evidenced that the CO 2 flux anomalies in the savanna and woody savanna areas usually occurred during the wet season. Positive daily CO 2 flux anomalies (up to 4 gC m −2 d −1 ) were usually observed on days with extreme precipitation (Figure 2, Figure S4a of Supplementary Materials) and were associated with high soil moisture, resulting in higher soil respiration rates [42]. This effect may have been associated with the "Birch effect" [43], manifested by a strong CO 2 release following the rewetting of dry soils. Several phenomena connected with the effect were previously found: more nitrogen and carbon are released from soils under wetting and drying cycles than from continuously wet soils; the longer the drying period the greater the degree of decomposition and mineralization during subsequent wetting; enhanced decomposition and the mineralization of soil organic matter decline with time after rewetting [44,45]. A sharp increase in CO 2 uptake by these ecosystems, leading to negative flux anomalies, was observed between a few days and one week after heavy rains (for example the period March-April 2012 at Figure 2). It may have been due to the intensification of photosynthesis in living plants [32]. Similar effects of precipitation on CO 2 uptake were previously detected at several experimental sites in savanna in Australia [46,47] and Africa [48]. The time-series analysis evidenced that the CO2 flux anomalies in the savanna and woody savanna areas usually occurred during the wet season. Positive daily CO2 flux anomalies (up to 4 gC m −2 d −1 ) were usually observed on days with extreme precipitation (Figure 2, Figure S4a of Supplementary Materials) and were associated with high soil moisture, resulting in higher soil respiration rates [42]. This effect may have been associated with the "Birch effect" [43], manifested by a strong CO2 release following the rewetting of dry soils. Several phenomena connected with the effect were previously found: more nitrogen and carbon are released from soils under wetting and drying cycles than from continuously wet soils; the longer the drying period the greater the degree of decomposition and mineralization during subsequent wetting; enhanced decomposition and the mineralization of soil organic matter decline with time after rewetting [44,45]. A sharp increase in CO2 uptake by these ecosystems, leading to negative flux anomalies, was observed between a few days and one week after heavy rains (for example the period March-April 2012 at Figure 2). It may have been due to the intensification of photosynthesis in living plants [32]. Similar effects of precipitation on CO2 Figure 2. The time series of daily net CO 2 flux (NEE) anomaly and daily temperature anomalies (a-c), daily precipitation amount (d,e) at flux monitoring station AU-Dry (Dry River) for the period January-July 2012. The days when the daily CO 2 flux anomalies were greater (lower) than 1 STD of CO 2 time series for this station are marked by red (blue) dots. The red (blue) shading is applied for the periods when the temperature exceeds the upper (lower) threshold: (a) 1 STD (−1 STD); (b) 90% (10%); (c) 95% (5%) PDF quantile. The red shaded column (blue triangle) is applied for the days when the precipitation daily amount exceeds the upper (lower) threshold: (d) 90% (10%); (e) 95% (5%) PDF quantile.
The influence of temperature on CO 2 fluxes in savannas is stronger in a dry season: low temperatures are accompanied by negative CO 2 flux anomalies ( Figure S4b), mainly due to the reduction in soil and plant CO 2 respiration rates, whereas positive flux anomalies (increased CO 2 emission) tend to occur during hot periods ( Figure S4c). However, all these anomalies were significantly lower than CO 2 flux anomalies during extreme precipitation periods.
The tropical seasonal or dry forests were characterized by well manifested seasonal variability in CO 2 fluxes caused by plant phenology, air temperature, and precipitation. Significant negative daily CO 2 flux anomalies (increase in CO 2 uptake up to 4.0 gC m −2 day −1 ) may have been due to photosynthesis intensification after the heavy rainfalls at the onset of the summer monsoon season ( Figure S5a). Another reason for such an effect was the inhibition of soil respiration by strong rainfalls [49]. Similar but weaker (<−2.0 gC m −2 day −1 ) negative CO 2 flux anomalies were also associated with negative temperature anomalies ( Figure 2b). Prolonged heavy rainfalls were often accompanied by positive flux anomalies (increasing CO 2 release) due to soil wetting and solar radiation decrease ( Figure S5c). Such positive daily CO 2 flux anomalies did not exceed 2.6 gC m −2 day −1 . The positive CO 2 flux anomalies occurred also during the dry period in the days with extreme high temperatures, resulting from suppressed photosynthesis and the respiration of vegetation ( Figure S5b March-April-May). These daily anomalies were usually lower than 1.0 gC m −2 day −1 .
Temperature extremes are quite rare in the equatorial climate; therefore, CO 2 flux anomalies in tropical rainforests are primarily related to precipitation variations. The precipitation intensity varies over the year, depending on the seasonal migration of the Intertropical convergence zone, resulting in drier and wetter seasons [31,50,51]. Inter-annual variability of precipitation in the Pacific is also induced by El Niño Southern Oscillation phenomenon [29]. During the wetter period, the positive daily CO 2 flux anomalies (up to 6.2 gC m −2 day −1 ) are observed in days with heavy precipitation due to decreased solar radiation, reducing GPP, and sufficient soil moisture conditions, enhancing soil decomposition and the mineralization processes ( Figure S6a). The analysis also showed that even insignificant precipitation leads to positive CO 2 flux anomalies during dry seasons ( Figure S6a, November-December). The positive CO 2 flux anomalies are also observed during the periods with positive temperature anomalies due to an increase in ER and CO 2 emissions into the atmosphere ( Figure S6b). However, these anomalies are significantly lower than those induced by extreme precipitation and do not exceed 1.0 gC m −2 day −1 .
The CO 2 fluxes in wetlands are primarily dependent on precipitation rather than on anomalous temperatures. Positive daily CO 2 flux anomalies (up to 3.1 gC m −2 day −1 ) tend to occur during heavy precipitation events ( Figure S7), resulting in an increase in CO 2 release into the atmosphere. As annual temperature fluctuations are small, and temperature extremes are rarely exceeded in the wetlands, their influence on the anomalies of CO 2 fluxes is less than that of precipitation.

The Statistics of Relationships between Extreme Weather Conditions and Daily CO 2 Flux Anomalies
To quantify the relationships between extreme temperature/precipitation and CO 2 fluxes, we calculated the percentage of the days when the temperature (precipitation) and CO 2 flux thresholds were simultaneously exceeded from the total number of the days when one of the characteristics (temperature or precipitation or CO 2 flux anomalies) exceed the threshold values. For example, we considered the number of days when the temperature anomaly exceeded the threshold (95%, 90% quantile or 1 STD) or when it was lower than the threshold (5%, 10% quantile or −1 STD) during the analyzed period as 100%. Within this sample, we then calculated the percentage of days when the CO 2 flux anomalies occurred simultaneously with temperature extremes: higher than 1 STD (max), lower than −1 STD (min), and the anomaly was less than STD (norm). This percentage is presented in Figures 3a and 4a for extreme high temperatures and in Figures 3c and 4c for extreme low temperatures. The same procedure was applied to precipitation, where the 95 (90)% and 5 (10)% thresholds were considered (Figures 3b,d and 4b,d). As a second step, we verified the inverse relationship, i.e., considered the total number of the days, when CO 2 flux anomalies exceeded 1 STD as 100% and, within this sample, calculated the number of the days when temperature (daily precipitation) anomalies simultaneously exceeded the thresholds (Figures 5a,c and 6a,c for CO 2 flux anomalies > 1 STD, Figures 5b,d and 6b,d for CO 2 flux anomalies <−1 STD).

Relationships between Daily Air Temperature and CO2 Flux Anomalies
The response of CO2 fluxes to extreme temperatures varied significantly between ecosystems and even within one biome. It may have been due to the stronger dependence of CO2 fluxes of tropical ecosystems on the precipitation than on temperature variations, as was evidenced by the earlier analysis of the time series.
At two savanna stations (AU-Dry and AU-RDF), the extreme high temperatures resulted in increased CO2 emissions (Figure 3a) and extreme low temperatures-in increased CO2 uptake (Figure 3c). The inverse relationship was also significant (with lower Figure 6. The same as Figure 5 but for tropical evergreen, deciduous needle leaf, tropical rain, and dry or seasonal forests.

Relationships between Daily Precipitation and CO 2 Flux Anomalies
Analysis showed that abundant precipitation mainly leads to increased CO 2 emissions (Figure 3b) in savannas: during 25-37% of the days with extreme precipitation (exceeding 95% quantile), the positive anomalies of CO 2 fluxes exceeded 1 STD. This trend was confirmed by the inverse relationship: 18-22% of the days with CO 2 flux anomalies higher than 1 STD corresponded to the days with extreme (>95% quantile) precipitation (Figure 5b).
This was due to the watering of the soil horizons and was a consequence of the increased activity of soil biota resulting in increased CO 2 emissions from the soil surface. An exception to this rule was the station Au-RDF located in the north of Australia, where the extreme precipitation almost equally entailed strong positive and negative CO 2 flux anomalies (Figure 3b). This may have been due to limited soil organic carbon stocks, leading to reduced soil respiration, as well as the rapid response of the photosynthetic apparatus of plants to increased water availability in the root zone. Noteworthy, this station was very close to the AU-Dry station, where the increase in CO 2 emissions was only connected with abundant precipitation (Figure 1).
Precipitation deficit in savannas led to an increase in CO 2 uptake on the African continent: negative CO 2 flux anomalies exceeding 1 STD occurred within 28-100% of days, with daily rainfall below 5% (Figure 3d). Half of the days with CO 2 flux anomalies lower than −1 STD (50%) coincided with precipitation amounts lower than 5% quantile in Sahel (SN-Dhr). At the other stations, the percentage was much lower (Figure 5d). On the other continents, the relationship "increased CO 2 uptake/precipitation deficit" was not evident, and the CO 2 flux anomalies did not exceed the STD during the days with the lack of precipitation. The increased CO 2 uptakes under dry conditions may have resulted from the combined effects of the reduced ER and non-significantly changed GPP in savanna ecosystems due to their adaptation to drought conditions. The water supply from deeper water layers may have been a key factor influencing the high rate of plant photosynthesis. This was despite the fact that the seasonal GPP variations in semi-arid ecosystems were mainly controlled by near-surface soil water [52,53].
An increase in CO 2 release in evergreen broadleaf forests was associated with extreme precipitation over 20-30% of days (Figure 4b), resulting from soil wetting and intensive soil respiration. This was the main feature of all analyzed flux stations located in evergreen tropical rainforests on different continents. A lack of precipitation (Figure 4d) led to an increase in CO 2 uptake (16-23% of days). The latter may have been due to the increase in GPP values at high solar radiation (during the days without precipitation and clouds) and optimal soil moisture conditions that remained in evergreen forests even in the absence of precipitation. The inverse relationship was significant for positive CO 2 flux anomalies that often were accompanied by heavy rainfall (Figure 6b) but was not significant for negative anomalies that were observed with no relevance to the precipitation extremes (Figure 6d). In tropical rainforests (e.g., flux station in South East Asia-BNS), the relationships between precipitation and CO 2 fluxes were less evident, and the extreme precipitation usually did not correspond to the anomalies of CO 2 fluxes.
The grasslands and permanent wetlands were characterized by the response of CO 2 fluxes to the precipitation variability, similar to rainforests (Figure 3b,d): the strong CO 2 release resulting from heavy rainfall (up to 33% of the days) and increased CO 2 uptake associated with the lack of precipitation (up to 25% of the days). This type of relationship also occurred in the seasonal tropical deciduous forest, where up to 40% of days with extreme precipitation were characterized by extremely positive CO 2 flux anomalies (Figure 4b), and up to 46% of the days with low precipitation coincided with strong negative CO 2 flux anomalies (Figure 4d). However, in the other type of seasonal tropical forest-deciduous needle-leaf forest-there was no evidence of any relationship between precipitation and CO 2 fluxes, with an almost equal percentage of the days with positive and negative flux anomalies accompanying the lack of precipitation. The heaviest precipitation was mostly associated with increased CO 2 emission, but the percentage was low (up to 19%). The differences in the observed relationship were mainly influenced by the differences in ER and GPP and their responses to changing soil moisture conditions caused by both precipitation anomalies and local soil moisture conditions, as well as it was affected by various specific biotic and abiotic factors [54].

Relationships between Daily Air Temperature and CO 2 Flux Anomalies
The response of CO 2 fluxes to extreme temperatures varied significantly between ecosystems and even within one biome. It may have been due to the stronger dependence of CO 2 fluxes of tropical ecosystems on the precipitation than on temperature variations, as was evidenced by the earlier analysis of the time series.
At two savanna stations (AU-Dry and AU-RDF), the extreme high temperatures resulted in increased CO 2 emissions (Figure 3a) and extreme low temperatures-in increased CO 2 uptake (Figure 3c). The inverse relationship was also significant (with lower percentage): the positive CO 2 flux anomalies were more likely during the hot days ( Figure 5a) and negative anomalies-during the days with temperature declines (Figure 5c). At the station AU-Ade, the opposite response was detected: minimum temperatures were associated with increased CO 2 emission (up to 38% of the days) and maximum temperatures, in turn, with CO 2 uptake (up to 30% of the days). In Sahel (SN-Dhr), the hot days were characterized by prevailing increased CO 2 uptake. In Congo (CG-Tch), there was no evident relationship between air temperature and CO 2 flux variations, which was likely the result of the dominant influence of precipitation conditions (as evidenced by the high percentage of precipitation extremes). At least in Sudan (SD-Dem), the negative CO 2 flux anomalies dominated during the period of observations, with no relevance to the temperature changes.
In the evergreen forests, the temperature decreases mainly coincided with high CO 2 emissions (Figure 4c). Similar relationships were found in seasonal or dry forests (an exclusion for MKL, where the positive and negative CO 2 flux anomalies were equally likely during the periods of low temperature) and in wetlands (Figure 3c). However, these tendencies may also have been related to precipitation changes: as mentioned above, the abundant precipitation results in increased net CO 2 releases due to lower GPP and higher ER. At the same time, the heavy precipitation was accompanied by increased cloudiness. Cumulonimbus clouds with a high cloud optical depth dominated in the equatorial latitudes. This led to a significant reduction in total solar radiation, a corresponding decrease in temperature under cloudy conditions, and a large angle of incidence of sunlight. However, in Australian evergreen and tropical rain forests, the positive CO 2 flux anomalies occurred, on the contrary, during the days with high temperature (Figure 4a) and negative CO 2 flux anomalies-during the days with low temperature (Figure 4c).
In comparison to precipitation, when the maximum percentage of simultaneous extremes in weather conditions and CO 2 fluxes was observed for the 95% quantile threshold, for temperature, the maximum percentage was observed for the 90% quantile of PDF.

The Correlation between Daily CO 2 Flux Anomalies and Temperature/Precipitation Extremes
The correlations between extreme air temperature (precipitation) and CO 2 flux anomalies confirmed the results mentioned above (Table 1). Correlations were calculated for the time series, consisting of the days when the extreme event threshold was exceeded: 5%/95% (P Q95) and 10%/90% (P Q90) quantiles for precipitation and 5%/95% (T Q95) and 10%/90% (T Q90) quantiles with ±1 STD (T STD) for temperature. The number of samples used for correlation analysis is shown in Table S1 of the Supplementary Materials.
The extreme precipitation was positively correlated with CO 2 flux anomalies, i.e., the abundant precipitation was associated with the strong CO 2 release and weak precipitation with increased CO 2 uptake. The statistically significant (p < 0.05) correlation for 95% quantile threshold was found in the Australian and Congo savannas, the evergreen forests in Australia (the maximum correlation-0.81) and Malaysia, and the deciduous needle-leaf forest in Brazil. The statistically significant correlation for the 90% quantile threshold was found only in two stations (Au-Rob and MY-PSO) in the evergreen forests. At two stations, the MKL in monsoon forests in Thailand and the GH-Ank in evergreen forests in Ghana, a negative correlation between precipitation and CO 2 fluxes was found, but it was not statistically significant. The extreme temperatures had a positive and negative correlation with CO 2 flux anomalies. A positive correlation was found in the savanna in Sudan and the evergreen forests in Australia (R = 0.39 for 95% quantile) and Ghana. In these ecosystems, the high temperatures corresponded to increased CO 2 releases, and a decrease in temperature led to higher CO 2 uptake. Most ecosystems were characterized by a negative correlation between temperature and CO 2 flux anomalies, i.e., positive CO 2 flux anomalies occurred during periods of low temperature and negative anomalies during rising temperatures. The maximum correlation (R = −0.51, p < 0.05) was found in the Brazilian needle-leaf forest.

Combined Effect of Temperature and Precipitation Extremes on Daily CO 2 Fluxes
In the second step, we examined the combined effect of temperature and precipitation on CO 2 fluxes (Figure 7). The most significant impact on daily CO 2 fluxes had the aggregated effect of cold/wet (CW) conditions, i.e., extremely high precipitation and extreme low temperatures (Figure 7b). In all considered stations, the CW conditions were mostly associated with positive CO 2 flux anomalies: 20-100% of the days with CW conditions corresponded to CO 2 flux anomalies exceeding 1 STD. Maximum simultaneous occurrences of CW and positive CO 2 flux anomalies were observed in evergreen forests, whereas they were rare in the tropical rain forests and tropical seasonal deciduous forests. The increase in the release of CO 2 into the atmosphere under CW conditions likely resulted from the combined effect of GPP and ER changes: cold, wet, and cloudy weather promoted reduced plant photosynthesis and GPP; cold weather led to a decrease in ER; and increased soil water content, SWC (at SWC < field capacity), may have, in turn, resulted in ER increase. The surface water logging under abundant precipitation may have resulted in reduced GPP because of hypoxia and increased heterotrophic soil respiration [55,56]. The effect of cold/dry (CD) anomalous weather events on CO2 fluxes (during more than 20% of the days, the simultaneous CD and CO2 flux anomalies were observed) was detected in several ecosystems in western Africa, northern Australia, the Malaysian peninsula, and Southern Africa. Negative anomalies prevailed, although there was one positive (AU-Fog). The effect of CD anomalies on CO2 fluxes in other studied ecosystems The effect of hot/wet (HW) anomalies on daily CO 2 fluxes was manifested only in savannas and woody savannas, and they could lead to both positive and negative CO 2 flux anomalies. Negative anomalies could be associated with an increase in savanna greenness and GPP, especially after the dry season [57]. Positive CO 2 flux anomalies were mainly the result of higher rates of heterotrophic and autotrophic respiration under elevated temperatures [58] and sufficient soil water supply [42]. It was notable that the high positive CO 2 flux anomalies occurred during hot/wet conditions with almost the same probability as during cold/wet periods in the savannas (Figure 7a,b). This confirmed the hypothesis of the dominant influence of precipitation and soil moisture content on CO 2 fluxes in the semiarid tropical regions.
The hot/dry (HD) weather conditions had a maximum impact on the CO 2 fluxes in the tropical forests (evergreen, seasonal deciduous, and deciduous needle-leaf). During 22-100% of the days with extremely high temperatures and extremely low precipitations, increased CO 2 uptakes (anomaly exceeds −1 STD) were found. These may have been the result of the very high GPP of tropical forests under sufficient soil moisture conditions. The reduced precipitation, in this case, was not a limiting factor in the optimal plant water supply. A similar effect of the hot/dry weather conditions on CO 2 fluxes occurred in the savanna in Congo (Figure 7c). However, in the Australian forest savanna (station AU-Dry), 30% of the days with HD conditions corresponded to the abnormally high CO 2 release. It was an interesting result because, according to an analysis of Chen et al. [59], the soil CO 2 efflux in the northern Australian savanna during the most part of the dry season was much lower than the CO 2 efflux during the wet season. The very high CO 2 release under hot/dry weather conditions may have been due to the contribution of the forest canopy (reduced GPP and high ER) into the total CO 2 flux [60].
The effect of cold/dry (CD) anomalous weather events on CO 2 fluxes (during more than 20% of the days, the simultaneous CD and CO 2 flux anomalies were observed) was detected in several ecosystems in western Africa, northern Australia, the Malaysian peninsula, and Southern Africa. Negative anomalies prevailed, although there was one positive (AU-Fog). The effect of CD anomalies on CO 2 fluxes in other studied ecosystems was very small. The maximum negative CO 2 flux anomaly was found in Sahel (SN-Dhr) and the deciduous broadleaf forest in western Zambia in Africa (ZM-Mon) and was evidently connected with seasonal variations in GPP [61] and the adaptation of woody landscapes to dry conditions [62]. The positive CO 2 flux anomaly in a seasonal wetland in the wet-dry tropics of Northern Australia (AU-Fog) could be associated with variation in soil water content and groundwater depth, influencing GPP and ER [63]. The reduction in GPP under cold/dry anomalies while maintaining CO 2 emissions from the wetted peat may have been the main reason for this trend.
It should be noted that the response of CO 2 fluxes to weather anomalies evidenced in our study was somewhat different from the previously documented relationships. Zscheischler et al. [8] evidenced that positive extremes in CO 2 fluxes were associated with dry and hot conditions in tropical forests. We detected the same CO 2 flux responses in Australian forest savannas, whereas higher CO 2 emission occurred during periods of extreme positive temperature anomalies in several types of evergreen and seasonal forests. However, the opposite feedback was revealed for most of the considered monitoring stations: intense CO 2 emissions, associated with cold/wet and hot/dry weather conditions, were accompanied by strong CO 2 uptakes. This indicated that CO 2 flux changes were not related to the temperature/precipitation fluctuations in a straightforward manner, i.e., positive and negative temperatures and precipitation oscillations differently influenced GPP and ER rates that may have resulted in various responses of CO 2 fluxes to external impacts. The response may have depended on various local biotic and abiotic factors, including plant canopy age and structure, biodiversity, plant plasticity, soil organic carbon, soil water availability, surface topography, solar radiation fluctuation, etc.

Conclusions
The analysis of the temporal variability of daily temperature, precipitation, and CO 2 flux anomalies, as well as their relationships, highlighted the large diversity of CO 2 flux responses to the fluctuations of temperature and precipitation in tropical ecosystems.
The heavy precipitation mainly led to a strong intensification of CO 2 release into the atmosphere due to increased soil moisture and intensified microbial activity, enhanced the decomposition and mineralization of soil organic matter ("Birch effect"), and increased autotrophic plant respiration. These types of relationships have been observed at almost all stations in all types of study biomes. Within a few days after heavy rainfall, CO 2 emission gradually declined with time after rewetting, and the ecosystems began to serve as a CO 2 sink from the atmosphere, mainly due to the intensification of plant photosynthesis under optimal soil moisture conditions.
The precipitation deficit contributed to the negative anomalies (higher uptake) of CO 2 fluxes (with, however, almost half weaker relationship than for heavy precipitation) due to reduced ER and high GPP rates under sufficient soil moisture supply.
The influence of temperature fluctuations on CO 2 fluxes was more pronounced during the dryer period, associated with a lack of precipitation. In evergreen forests, seasonal or dry forests, wetlands, and some savannas, the low temperatures coincided with higher CO 2 emissions. During the rainy season, the temperature effect was closely related to changes in precipitation: the abundant precipitation was accompanied by a strong cloudiness and, therefore, a lower temperature. At the same time, extreme precipitations implied an increase in CO 2 emissions.
In some savannas, Australian evergreen forests, and seasonal or dry forests, high temperatures contributed to higher CO 2 emissions, whereas low temperatures assisted in higher CO 2 uptake. Higher temperatures usually led to higher ER rates if there was a lack of the limiting influence of other external factors such as soil moisture availability. This type of relationship was likely typical of the dry season, when the influence of precipitation was negligible. The extreme high temperatures stressed the photosynthesis, reduced GPP rate, and led to a higher release of CO 2 into the atmosphere.
A comparison of the relationship between daily CO 2 flux, precipitation, and temperature anomalies showed that the sensitivity of CO 2 fluxes to precipitation anomalies was stronger than that of the temperature change. This may have been explained by relatively small temperature fluctuations in the tropics in comparison with the precipitation. In the geographical regions with strong seasonal variations in precipitation conditions, the stronger dependence of CO 2 fluxes on precipitation than on temperature may have been due to the overall soil moisture deficiencies in these ecosystems during dry periods.
The strongest combined effect of temperature and precipitation on the daily CO 2 fluxes was detected under cold/wet weather conditions, resulting in higher CO 2 emissions almost in all considered tropical ecosystems. It may be interpreted as the effect of reduced GPP due to low temperature and reduced incoming solar radiation, as well as more intensive soil respiration due to abundant precipitation. The opposite weather conditions (hot/dry) were mostly associated with increased CO 2 uptake. That was also not an evident result, as a lack of precipitation and extremely hot temperatures often led to the suppression of GPP and positive CO 2 flux anomalies. In the tropical forests, however, the prevailing sufficient soil moisture conditions made precipitation not a limiting factor for GPP, whereas high temperatures associated with high solar radiation may cause the intensification of photosynthesis.
Notably, the anomalies of CO 2 fluxes often did not coincide with extreme precipitation or temperature anomalies, indicating the strong influence of various abiotic and biotic factors on ecosystem functioning, manifesting differently for individual plant communities.
The optimal threshold for determining the strongest relationship between the weather extremes and CO 2 fluxes was different for temperature and precipitation: it was a 95% (5%) quantile for precipitation and a 90% (10%) quantile for temperature. These were likely due to a smoother distribution of temperature as compared to precipitation.
Taking into account the revealed relationships, the results obtained require further multifaceted studies, involving a greater number of monitoring stations and a longer time series of observations of GHG fluxes.