Modelling Net Ecosystem Exchange in the Biebrza Wetlands using satellite and meteorological data

The Intergovernmental Panel on Climate Change (IPCC) is the leading international body for the assessment of climate change. One of the IPCC's main activities is the preparation of comprehensive Assessment Reports on the state of scientific, technical and socio-economic knowledge on climate change, its causes and potential impacts, and on response strategies in preparing national inventories of Greenhouse Gas (GHG) emissions by source and removals by sinks. Wetlands are very important ecosystems, due to their accumulation of more than 30% soil carbon (Bridgham et al. 2006; Parish 2008; Windecker 2019) and are included in the GHG reports.

Soils of organic origin, including wetlands, cover 3–5% of the land surface (Mitsch & Gosselink 2007; Davidson 2014; Ciężkowski et al. 2020) and accumulate on average as much as 547 Pg of carbon (Pullens et al. 2016). Wetlands are extremely vulnerable to climate change, and hence to changes such as rising temperatures or decreasing rainfall. Due to the large accumulation of carbon in the peat, depending on the thermal and humidity conditions, they can be either a source of GHGs or a place for their accumulation, mainly CO₂ and CH4 (Canadell et al. 2011). It is currently estimated that almost 6% of global anthropogenic CO₂ emissions come from wetlands (IUCN reports).

There is a strong need to combine the research on carbon fluxes at the different peatlands across different regions. The objectives of the Regional Carbon Cycle Assessment and Processes (RECCAP) (Canadell et al. 2011), covering international carbon cycle research, were to establish the mean carbon balance of large regions of the globe at the scale of continents, but there is the need to provide higher spatial resolution and evaluate the regional ‘hot spots’ of interannual variability and trends by fulfilling the growing demand for the capacity and establish a global carbon observation strategy (Ciais et al. 2010).

The Net Ecosystem Exchange (NEE) of CO₂ between the terrestrial ecosystem and the atmosphere reflects the balance between the absorption by vegetation through photosynthetic gross CO₂ assimilation (gross primary production – GPP) and ecosystem respiration (RESP), in which, photosynthesis processes are CO₂ sinks and respiration processes are CO₂ sources (Chojnicki 2010; Canadell et al. 2011).

The commonly used methods of measuring NEE are eddy covariance and Bowen-ratio/energy balance (BREB). Eddy covariance and BREB systems are nonintrusive micrometeorological methods that impose minimal influences on microenvironments of the soil surface compared with chamber-based methods (Dugas 1993). Besides, in many studies of field CO₂ flux measurements, the photosynthetically active radiation (PAR), plant community composition, water table, air and soil temperature have been identified as the main variables controlling CO₂ flux exchanges (Bellisario et al. 1998). Gilmanov (2005) examined the relationship between CO₂ and the remotely sensed Normalized Difference Vegetation Index (NDVI), and other environmental factors. Such relationships allowed the tower measurements to obtain regional-scale estimates of the carbon budget of grasslands in Northern Great Plains to be scaled up. Much attention is thus paid to replace in situ surveys with different products from satellite measurements. For example, in order to measure the wetness of peatlands, models with the input of radar images can be used (Dąbrowska-Zielińska et al. 2018).

The objective of the study carried out at the wetlands of Biebrza was: -

to measure the stream of net ecosystem exchange (NEE) and respiration at 28 measuring points during the growing season, i.e. from April to October, using chamber measurements, and observations at one point using the eddy covariance method

The Biebrza Wetlands are located in Central Europe, in the north-eastern part of Poland. The marsh area is located in the Central European Lowlands, the appearance of which was dominated by the last glacial period. There are many fluvioglacial deposits here, which influence the present landscape, also shaping the water relations and, consequently, the flora and fauna of the region (Okruszko 1990). Under the Biebrza Wetlands, there are impermeable strata – smelting depressions left by the ice sheet that caused the creation of the marshes. The species composition of wetlands is extremely diverse (both in terms of flora and fauna), even unique in Europe – there are more than 70 natural and semi-natural plants, including sedge, sedge-moss, reed communities, and mineral islands. For this reason, this area is guarded by numerous conventions, including Natura 2000 and RAMSAR (since 1995), as well as being placed under Polish protection as a National Park in 1993. The legally protected area covers a total of 59,233 ha, in which 15,547 ha of forests, 18,182 ha of agricultural land and 25,494 ha of wetlands (the most valuable habitats of the park) can be distinguished. The main source of water for the fen is the Biebrza River. It spreads widely at higher water levels, with a small slope where it flows into the larger river – the Narew (Batelaan et al. 2009; Berezowski et al. 2019).

Due to the large diversity of flora and fauna, as well as the relatively small transformation of the area by humans (first started about 150 years ago and continued for 50 years in the 20th century), the peat bog system on the Biebrza River is characterized by its wide range and its diversity, both in terms of species and depth, and is thus associated with a large carbon capacity. Unfortunately, there has been a gradual degradation of some parts of the Biebrza Wetlands, caused by both climate change and direct human interference (Kleniewska et al. 2009). For this reason, research on the current carbon balance in this area is extremely important, both from the local and pan-European point of view.

The work consists of several stages: collecting field data for the years 2015–2022, data analysis, modelling based on the collected measurement parameters, and the preparation of maps, applying satellite indicators.

Data were collected during the growing season (between April and October) in the years 2015 to 2022. They consisted of meteorological measurements (air temperature, air humidity, PAR), soil parameters (soil moisture and temperature at depths from 5 to 50 cm), type of vegetation and CO₂ flux measurements (chamber and eddy covariance methods). The chamber measurements took place at 28 stations located on different parts of the wetlands, and the EC on one of them (Fig. 1). About 250 chamber measurements were collected at observation points on three types of Biebrza plant communities: grasses, sedges and reeds.

Figure 1

Field measurements were performed every 2 to 3 weeks, and were closely related to the flights of satellites Sentinel-1 and Sentinel-2. At each measurement site, the following measurements were performed: -

Carbon balance using the chamber method (NEE, RESP);

The carbon flux measurements were performed in two ways: the first by the chamber method and the second by the eddy covariance method. CO₂ fluxes are widely measured using the chamber method (Dugas 1993; Livingston & Hutchinson 1995; Jensen et al. 1996). CO₂ fluxes over wetlands have been measured by Chojnicki et al. (2010). In the study, the measurements were carried out simultaneously in two chambers: a transparent chamber and a covered (by dark material) chamber. Both chambers were made of plexiglass. Similar methodology has been carried out by Rychlik & Dąbrowska-Zielińska (2011). In their study, CO₂ was measured in one chamber: 6–10 minutes in a transparent chamber (simulating the daytime) and in the same chamber covered by dark material. A different method of chamber measurements is to measure CO₂ in one chamber all night and then all day (Juszczak et al. 2012). All of the mentioned methods also consist of the measurement of the soil temperature outside the chamber (next to it). Chamber measurements were carried out in the standard manner provided for this method. In this study, changes in CO₂ concentration over vegetation in the closed chamber were observed, as well as the air temperature. A Senseair sensor was used for the measurements. The entire measurement took 16 minutes: 8 minutes in conditions of solar energy influx and 8 minutes after darkening the chamber. On the basis of the first 8 minutes, the Net Ecosystem Exchange was determined. Data collected during the next 8 minutes was used to determine Total Ecosystem Respiration. Gross Primary Production, representing the total CO₂ amount absorbed by vegetation, was calculated as the difference between NEE and RESP. The CO₂ measurements were performed between 9 a.m. and 5 p.m. in June and between 10 a.m. and 4 p.m. in other months. The measurements were not carried out during the rain. It is extremely important to note such conditions, and they will provide information for further analysis. One CO₂ measurement was performed on a test site by the eddy covariance method.

Another popular way to measure CO₂ fluxes is the eddy covariance method, in which the net transport between the surface and atmosphere is calculated by the covariance between the turbulent fluctuations of the vertical wind and the quantity of interest (Foken et al. 2012). Measurements at the EC tower were performed continuously, also from April to October. It should be noted, however, that some measurements in this method are omitted due to numerous disturbances (e.g. too low wind speeds, contamination of sensors, power supply problems at a point distant from the buildings). Therefore, the results from the tower served mainly to refer the EC measurements to the chamber measurements done in the different conditions of the habitat's sites.

Soil moisture measurements were conducted using TRIME-PICO devices. These mobile probes also measure soil conductivity and temperature. The measurements are carried out in the top layer of soil, up to 10 cm in depth. Information on soil moisture is important for estimating soil-water conditions – in particular, the phases of water demand by plants. Water availability is related to pluvial conditions, the ability of the soil to retain water and the water conditions of habitats. These conditions result from how water feeds the habitat, and the relief and hydrogeological conditions of the area. Next to the EC station, there is a station for measuring soil moisture, in stationary mode, at four different depths – 5 cm, 10 cm, 20 cm, 50 cm. Measurements are made at a frequency of one every fifteen minutes. In addition to the soil moisture itself, the soil temperature, soil dielectric constant and soil electrical conductivity are also measured – the latter two parameters have not been used directly in this work.

One of the most important parameters measured in the Biebrza Wetlands were meteorological parameters. The station was located on a flat, homogenous marshland site within the Biebrza National Park, and covered unmanaged sedges with moist organic soil. The following parameters were measured: photosynthetically active radiation (PAR – Fig. 2), air temperature, relative humidity, leaf wetness, wind speed and direction, atmospheric pressure, soil moisture and precipitation.

Figure 2

Figure 2 presents PAR over time and months – showing changes in the length of the day between April and October. The highest PAR is in July, while the lowest is in October. During the day the highest PAR is around noon. It was important to deduce the length of the day for modelling daily GPP (Dąbrowska-Zielińska 2022). The station is also equipped with the eddy covariance system with standard instrumentation, and consists of an open path gas analyzer (Li-Cor LI-7500A), four component net radiation sensors to analyse shortwave and longwave downward and upward radiation, an ultrasonic anemometer (Gill WindMaster 3D) and heat flux plates.

The exchange of plant carbon with the environment takes place in the photosynthesis process and consists of three stages: gross primary production (GPP), respiration (RESP) and net ecosystem exchange (NEE). All three are linked by the equation: (1) NEE=GPP+RESP NEE = GPP + RESP where GPP≤0, RESP≥0. NEE<0 describe the accumulation of CO₂ (Aurela et al. 2001; Zhou et al. 2009; Zhang et al. 2011; Chojnicki 2013; Green et al. 2018; Li et al. 2019; Heiskanen et al. 2021).

Figure 3 illustrates the chamber measurements of NEE, RESP and GPP calculated (eq. 1). The course of monthly averages of NEE, RESP and GPP were observed at 28 sites, on grassland, sedge and reed communities. It can be seen that the ability for both photosynthesis and respiration is larger for grasses. In May, they are the highest for grasses and sedges, and for reeds in June. The net absorption is also highest in May.

Figure 3

The monthly quantity of observations in the period May–September is about 30 for sedges and grasses, and for reeds it is about 10. In April and October, the quantity is about 10 monthly for every habitat.

GPP calculated from (1) as the difference between the measurements from transparent and dark chambers is the instant gross photosynthesis productivity. The chamber measurements took place at different times of day and may not represent the daily values of CO₂ fluxes exactly. The simulation of the daily course of GPP was performed as a function of PAR. The dependency followed the Michaelis-Menten equation which is described, used and modified by many researchers (Dąbrowska-Zielińska 2022).

GPP modelling was performed using the model described by Dąbrowska-Zielińska et al. (2022). Below we present a brief description of the approximations and modelling used.

The set of GPP¯day {\overline {GPP} _{day}} values, being the averages of GPP¯ \overline {GPP} daily simulations for the observation points, was fitted to derived values and calculated from the Sentinel-3 satellite index (NDVI) and surface temperature (T_S). The efficient model was estimated using NDVI and Ts-Ta, where Ta is air temperature from the meteorological station (where R = 0.80, R² = 0.64, RMSE = 1.4 μmol m⁻² s⁻², N = 68), as below: (2) GPPSAT=−0.16⋅exp(4.39⋅NDVIS3−0.004⋅(Ts−Ta)2 {GPP}_{SAT} = - 0.16 \cdot \exp \left({4.39 \cdot {NDVI}_{S3} - 0.004 \cdot {{\left({{T_s} - {T_a}} \right)}^2}} \right. where NDVI characterizes the greenness of vegetation and biomass. The second factor T_s-T_a is related to the humidity conditions. The intensity of CO₂ uptake reflects the dynamics of vegetation growth.

Analogously, the indices from Sentinel-2 representing vegetation greenness and moisture – NDVI and Normalized Difference Infrared Index (NDII) (Hardisky et al. 1983) – was used to estimate the intensity of photosynthesis: (3) GPPSAT=−0.37⋅exp(3.15⋅NDVIS3−1.71⋅(NDIIS2) {GPP}_{SAT} = - 0.37 \cdot \exp \left({3.15 \cdot {NDVI}_{S3} - 1.71 \cdot \left({{NDII}_{S2}} \right)} \right. where R = 0.78, R² = 0.60, RMSE = 1.7 μmol m⁻² s⁻², N = 60. In future calculations, we only used formula (2).

The excretion of CO₂ by the ecosystem into the atmosphere comes from the respiration of living organisms, both in plants and in the soil. There are many factors contributing to this phenomenon, the main ones being air temperature, soil temperature, soil moisture and the respiratory capacity of growing plants. The commonly used formula is from Lloyd & Taylor (1994), based on air temperature measurements, taking into account the specificity of vegetation and the phases of their development through physical and empirical parameters. In this paper, we investigate the possibility of using indicators from satellite imagery to estimate the respiration in order to obtain spatial differentiation of CO₂ flux exchange. We formulated an ecosystem respiration model using the factors that can be obtained for the resolution of the satellite pixel for the Biebrza area from the SAR satellite images from Sentinel-1, Sentinel-3 and Terra MODIS. In order to obtain a common spatial resolution for satellite data, a resampling was performed. Terra MODIS products that are provided with 1000 m resolution and Sentinel-1 images with 10 m resolution were resampled to 300 m (similar to Sentinel-3).

The soil parameters obtained from ground measurements in observation points are the soil moisture and soil temperature at a depth of 10 cm. Their diversification for the extracted habitats of the Biebrza area is illustrated in the graphs below (Fig. 4). There are significant differences between the soil moisture: sedges grow in the more humid area, and water remains for most of the season in the reeds. The distribution of the soil temperature for sedges and grasses is similar; for reeds, the high moisture translates into lower soil temperatures throughout the growing season.

Figure 4

In May, the NDVI for the grass habitat is higher than for sedges, which reflects the more intensive vegetation (Fig. 5). In June, the situation is the reverse – sedges achieve the higher dynamic of vegetation, grasses enter the senescence phase. The significant difference in LST (Land Surface Temperature, defined also as Ts from satellite data) in May and June is related to the distribution of soil moisture in the habitats (Fig. 5).

Figure 5

The vegetation growth differed during the season for 2015–2022. The mean NDVI and surface temperature from the MODIS satellite differed for the different months and type of vegetation (sedges, reeds, grasses). The air temperature (T_a) was derived from the IGiK station set up at the Biebrza Wetlands. The station was located on a flat, homogenous marshland site within the Biebrza National Park. Acquisitions from the Terra MODIS satellite (Land Surface Temperature – T_s) over Poland mainly occur between 10:00 and 12:00 local time every day.

The exponential dependency of ecosystem respiration with air temperature T_AIR measured in black chambers (Fig. 6), as well as with satellite-derived NDVI index (Fig. 7) was investigated. However, the dependency of soil moisture SM has a polynomial character (Fig. 8). The highest respiration took place where the soil moisture was in the 40–50% range. If it is above 80% in sedges and grasses, it has an inhibitory effect.

Figure 6

Figure 7

Figure 8

Figure 9

In the analysis, the factor (1-NDVI)·T_SOIL was estimated as statistically significant. This represents the interaction between the soil and vegetation in the respiration context: the higher the NDVI (high biomass) the lower the soil, which explains the variance of ecosystem respiration. This suggests that the impact of the soil temperature on respiration is dependent on the vegetation biomass covering the soil.

The measurements were taken in the hours between 7:00 and 15:30 UTC. The model was estimated for the log (RESP) dependency on the environment's variables: SM – soil moisture, NDVI from Terra Modis, T_SOIL – soil temperature, T_{AIR_obs} – air temperature at the moment of chamber CO₂ measurement. The result is the multifactor exponential equation: (4) RESP¯=4.44⋅exp(−0.004⋅(SM/10)2+0.0255⋅TAIR+0.039⋅(1−NDVI)⋅TSOIL) \overline {RESP} = 4.44 \cdot \exp \left({- 0.004 \cdot {{\left({SM/10} \right)}^2} + 0.0255 \cdot {T_{AIR}} + 0.039 \cdot \left({1 - NDVI} \right) \cdot {T_{SOIL}}} \right)

The goodness-of-fit statistics of the log-linear regression are R = 0.81, RMSE = 1.2 μmol CO₂ m⁻² s⁻¹, N = 105. There is mutual independence of the equation components R² = 0.02, so the tolerance is on the level R² = 0.98.

The extrapolation of equation (4) for the nighttime temperatures gives the night values of ecosystem respirations.

The application of equation (4) with soil moisture SM, soil temperature T_soil, daily air temperature T_air, and NDVI allows the daily mean respiration RESP_day to be obtained. The mean night respiration RESP_night was calculated taking the mean night air temperature. Assuming the lognormal fits the distributions, the day respiration is higher than that of the night by about 1.5 μmol CO₂ m⁻² s⁻¹ on average (Fig. 10). The number of points was 177.

Figure 10

The evaluation of model (4) was done on the nighttime values using the medians as the statistical criteria. Figure 10 presents the comparison of the distributions of the night respiration modelled by eq. (4) and the measurements taken at the EC tower.

The median night modelled observations is 6.7 μmol CO₂ m⁻² s⁻¹; the median night tower observations is 6.5 μmol CO₂ m⁻² s⁻¹. In this sense, the model generates nighttime RESP values at the observation points in Biebrza that are consistent with those measured by the EC tower.

Net CO₂ exchange as the result of simulated GPP and modelled RESP

The sum of the two fluxes, photosynthesis and respiration, where the participation of respiration is weighted with the length of day and night, approximate the value of the diurnal net CO₂ exchange. Equation (5) describes it for simulated GPP and modelled RESP. (5) NEE¯diurnal=GPP¯day+α⋅RESP¯day(TAIR_day)+(1−α)⋅RESP¯night(TAIR_night) {\overline {NEE} _{diurnal}} = {\overline {GPP} _{day}} + \alpha \cdot {\overline {RESP} _{day}}\left({{T_{AIR\_day}}} \right) + \left({1 - \alpha} \right) \cdot {\overline {RESP} _{night}}\left({{T_{AIR\_night}}} \right)

Where α=(day length)/(diurnal length) and length is expressed in any time units; T_{AIR_day} – mean of day air temperature; T_{AIR_night} – mean of night air temperature.

In equation (5), GPP¯day≤0 {\overline {GPP} _{day}}\, \le 0 ; RESP¯day {\overline {RESP} _{day}} , RESP¯night≥0 {\overline {RESP} _{night}}\, \ge 0 ; NEE¯day {\overline {NEE} _{day}} is negative when the uptake of CO₂ outperforms the release; it is positive otherwise.

Figure 11 presents the comparison between the distribution of the daytime EC tower CO₂ flux and daytime NEE modelled by (eq. 5), where RESP_night is omitted, as in the formula below: (5a) NEE¯day=GPP¯day+α⋅RESP¯day(TAIR_day) {\overline {NEE} _{day}}\, = {\overline {GPP} _{day}}\, + \alpha \cdot {\overline {RESP} _{day}}\,\left({{T_{AIR\_day}}} \right)

Figure 11

The means and standard deviation of the above distributions are −4.0 and 5.0 μmol CO₂ m⁻² s⁻¹ for modelled daytime NEE and −4.1 and 5.5 μmol CO₂ m⁻² s⁻¹ for tower daytime NEE. In agreement with the test, the difference between these means is not statistically significant. In this sense, the model generates the daytime NEE values at the observation points of Biebrza, which are consistent with those measured by the EC tower.

Satellite model of soil moisture in the Biebrza Wetlands

Soil moisture for the Biebrza Wetlands was developed on the basis of continuous measurements in 2015–2017 and published in Dąbrowska-Zielińska et al. (2018). The Water Cloud Model is the method based on the dependence of radar backscattering in relation to the soil moisture under the vegetation. It is widely applied using SAR data in combination with satellite vegetation descriptors, calculated from the optical images. In the model, the S-1 cross ratios s°_VH-VV and s°_VV/VH (backscattering coefficients) were applied as the vegetation proxy. This allows the soil moisture in spatial resolution, having one Sentinel-1 image only, to be extrapolated. The equation for the retrieval of soil moisture, citing from Dąbrowska-Zielińska et al. (2018) takes the form: (6) SMSAT=(σ0VH+18.9+0.14⋅(1−T2)⋅cosθ⋅(σ0VH−σ0VV)2)/(0.33⋅T2) {SM}_{SAT} = \left({{\sigma ^0}_{VH} + 18.9 + 0.14 \cdot \left({1 - {T^2}} \right) \cdot \cos \theta \cdot {{\left({{\sigma ^0}_{VH} - {\sigma ^0}_{VV}} \right)}^2}} \right)/\left({0.33 \cdot {T^2}} \right) where t² = exp(−2s°_VV/VH/cos(q)). Factor t², called the attenuation, is connected with the changes of plant density during the vegetation season; q is the projection angle of the Sentinel-1 satellite. The range of soil moisture found in the wetland area is about 30–100%. The precision obtained was 10%.

The maps on Figure 12 have been chosen to illustrate the results of calculating the Biebrza Wetlands soil moisture in the wet time (May 2019) and dry time (April 2020).

Figure 12

Maps of CO₂ exchange flows obtained from satellite indicators

Equation (4), with satellite version of environmental factors NDVI and SM, takes the following form: (7) RESPSAT(TAIR)=4.44⋅exp(−0.004⋅(SMSAT/10)2+0.04⋅TSOIL⋅(1−NDVISAT)+0.026⋅TAIR) \matrix{{{RESP}_{SAT}\left({{T_{AIR}}} \right) = 4.44 \cdot \exp \left({- 0.004 \cdot {{\left({{SM}_{SAT}/10} \right)}^2} +} \right.} \cr {\left. {0.04 \cdot {T_{SOIL}} \cdot \left({1 - {NDVI}_{SAT}} \right) + 0.026 \cdot {T_{AIR}}} \right)} \cr}

The equation for implementing NEE, applying the satellite indices and agree with the method described in formulas (2)–(5), takes the form: (8) NEESAT=GPPSAT+a⋅RESPSAT(TAIR_day)+(1−a)⋅RESPSAT(TAIR_night) {NEE}_{SAT} = {GPP}_{SAT} + {\rm{a}} \cdot {RESP}_{SAT}\left({{T_{AIR\_day}}} \right) + \left({1 - {\rm{a}}} \right) \cdot {RESP}_{SAT}\left({{T_{AIR\_night}}} \right)

Where: α=(day length)/(diurnal length) and length is expressed in any time units.

Figure 13 presents the spatial distribution of the NEE (8) and RESP (7) for the study area of Biebrza. The maps are based on Sentinel-1 and Sentinel-3 data for 31 May 2019. The soil moisture during the period of 27–31 May was characterized as high (Fig. 12), and there was high vegetation growth (before the first grass cut). The values of the NEE range from −0.3 μmol m⁻² s⁻¹ to −21.8 μmol m⁻² s⁻¹. The highest average NEE was observed for grass habitats (−15.8 μmol m⁻² s⁻¹), while the lowest average uptake was noted for sedge habitats (−13.3 μmol m⁻² s⁻¹). The reed habitats are characterized by average values of −14.9 μmol m⁻² s⁻¹. Average RESP values are within the range of 6.9 μmol m⁻² s⁻¹ for sedge habitats and 7.8 μmol m⁻² s⁻¹ for grass habitats.

Figure 13

This study describes the method of determining NEE using satellite indicators such as NDVI, LST and NDII, and meteorological environmental parameters Tair, Tsoil, for soil moisture (SM), the model based on Sentinel-1 data. The model (eq. 8) gives the spatial distribution of daily NEE for each pixel covering three types of vegetation habitats at Biebrza wetland area. The differentiation of phenological development and moisture conditions between the habitats translates into CO₂ exchange; the distribution of the daily NEE values generated by the model in the observation points is consistent with those measured at the data EC tower. Two fluxes making up the CO₂ exchange, GPP and RESP, were estimated. The precision of the daily GPP model, using MODIS NDVI and Ts-Ta as the parameters, is 1.4 μmol m⁻² s⁻²; using Sentinel-2 NDVI and NDII is 1.7 μmol m⁻² s⁻². The precision of the instant RESP model, using NDVI and soil temperature, air temperature and soil moisture, is 1.2 μmol m⁻² s⁻².

The results show that the indices derived from satellites have potential as predictors for the wetlands CO₂ exchange. The first stage of the work involved field measurements and collecting satellite data of the area of interest. In the next step, the RESP and NEE models were developed, using the previously used soil moisture models (Dąbrowska-Zielińska et al. 2018) and the gross primary production model (Dąbrowska-Zielińska et al. 2022). It is clearly visible that the highest streams (above −18 μmol m⁻² s⁻¹) are obtained for areas with humidity above 80%. In contrast, in areas with less than 60% humidity and non-reed vegetation, the NEE flux is around −6 μmol m⁻² s⁻¹. This is fully in line with our knowledge about the functioning of peat, which absorbs large amounts of carbon when properly hydrated (usually in May) and releases carbon dioxide when dry. In addition, the type of vegetation plays a huge role – sedges have a lower capacity for CO₂ sequestration than reeds and grasses. The research shows that the changes in peatland moisture may strongly affect NEE.

Due to global warming, there is a great danger of carbon dioxide release. It is important to monitor the hydrological and the Net Ecosystem Exchange systematically by applying the satellite data and elaborated in this article models.

Keeping good water conditions in wetlands is an important mitigation against the effects of climate change, as wetlands being a carbon reservoir is an object of special concern for the global environment.