5:15 PM - 6:45 PM
[MIS01-P09] Contrasting results obtained from the empirical-model ensembles applied to the forest and grassland ecosystem soil respiration modeling in different weather conditions
Keywords:Soil Respiration, Soil Respiration Modeling, Empirical Soil Respiration Models, SCLISS model, Climate conditions
Soil respiration (SR) modelling is an important alternative to direct chamber measurements (Kurganova et al. 2020) due to the high heterogeneity of measurements and the practical impossibility of allocating sufficient resources to cover large areas (Peltoniemi et al. 2007). At the same time, both modeling using simple empirical models linking SR with temperature (T) and precipitation (P) (Kurganova et al. 2022) and more complex dynamic models based on simulation of processes occurring in the soil (Peltoniemi et al. 2007) are considered.
Both SR measurements and applications of empirical SR models highlight a strong dependency of SR on the temperature (Raich and Potter 1995; Juhász et al. 2022), soil moisture (Raich and Potter 1995; Maier et al. 2010), precipitation (Raich and Potter 1995; Karelin et al 2017; Sukhoveeva and Karelin 2022), change of water level in soil (Pavelka et al. 2016), allocation of above-ground biomass (Reichstein et al 2003), and the amount of soil organic carbon (SOC) stored in soils (Lal 2005; Chen et al 2010; Kivalov et al. 2023).
In this study, we apply the ensemble of the empirical SR models to forest and grassland ecosystems situated on the Entic Podzol soil in Prioksko-Terrasny Nature Biosphere Reserve ≈50 m apart from each other to investigate how the difference in ecosystem properties and their effects on soil respiration can be addressed by the model ensembles. We use 25-year long SR time series collected at these sites by the closed chamber method (Kurganova et al. 2020) and measured once a week together with the soil and air temperatures (Tsoil, Tair). The monthly averaged air temperature (Tair) and precipitation (Prec) were also collected from the Complex Background Monitoring Station. The following 5 years were classified as “wet”—1999, 2003, 2006, 2008, 2020; the following 9 years were “dry”—2002, 2007, 2010, 2011, 2014, 2015, 2018, 2021, 2022; and the remaining 11 years stayed as “normal”—1998, 2000, 2001, 2004, 2005, 2009, 2012, 2013, 2016, 2017, 2019.
The following SR models are considered: T (dependent on T), TP (dependent on T, P), TPP (dependent on T, P, and Pm-1), TPPC (dependent on T, P, and Pm-1, and SOC), and TPPrh (dependent on T, T2, P, and Pm-1 - Raich-Hashimoto type). We also use the modeling of the soil temperature (TlitM, TsoilM) and moisture (MlitM, MsoilM) with the SCLISS model (Bykhovets and Komarov 2002) to see if the modeled values provide an appropriate substitution for the measured temperatures and precipitation.
We found out that two models from the ensemble set – TPPC and TPPrh have the best performances among all the models, which is also consistent with our previous findings (Kivalov et al. 2023). For the model comparison among different annual environmental conditions (norm, dry, wet) and different sources of the temperature (Tsoil, Tair, TlitM, TsoilM) and moisture (Prec, MlitM, MsoilM), we apply three metrics: R2 – determination coefficient, slope – the slope of the comparison between the measured and modelled SR values, and MBE – the mean bias error.
Focusing on the Entic Podzol soils, we found that for the forest SR with the low SOC, the direct Tsoil measurements in combination with SOC provide the better parametrization for the empirical TPPC model than if the only Tair parametrization was used. For the grassland SR with the relatively high SOC, the Tair becomes the governing factor, and the TPPrh model provides the better performance over all the considered models. Such a difference could be explained by both the SOC and temperature differences: the SOC becomes the limiting factor in the poor forest soil bringing forward the TPPC models, and the high Tsoil ≈ Tair values become the limiting factor for the grassland soil favoring for the TPPrh model.
When comparing the performances over the modeled with SCLISS temperatures (TlitM, TsoilM) and moistures (MlitM, MsoilM) data, we see that for both forest and grassland, the parametrization with the litter TlitM, and MlitM provides better results in all R2, slope, and MBE. These findings can identify the main source of the soil respiration as the topsoil or litter layer rather than the deeper soil layers. In wet conditions, using the litter modeled values (TlitM, MlitM) provide the best results for both TPPC and TPPrh models, while in normal and dry conditions the litter modelled values are slightly worser than the air parametrized values. This makes the parametrization with the SCLISS model for the litter temperatures and moistures the viable alternative, when the direct soil observations are not available on site.
Both SR measurements and applications of empirical SR models highlight a strong dependency of SR on the temperature (Raich and Potter 1995; Juhász et al. 2022), soil moisture (Raich and Potter 1995; Maier et al. 2010), precipitation (Raich and Potter 1995; Karelin et al 2017; Sukhoveeva and Karelin 2022), change of water level in soil (Pavelka et al. 2016), allocation of above-ground biomass (Reichstein et al 2003), and the amount of soil organic carbon (SOC) stored in soils (Lal 2005; Chen et al 2010; Kivalov et al. 2023).
In this study, we apply the ensemble of the empirical SR models to forest and grassland ecosystems situated on the Entic Podzol soil in Prioksko-Terrasny Nature Biosphere Reserve ≈50 m apart from each other to investigate how the difference in ecosystem properties and their effects on soil respiration can be addressed by the model ensembles. We use 25-year long SR time series collected at these sites by the closed chamber method (Kurganova et al. 2020) and measured once a week together with the soil and air temperatures (Tsoil, Tair). The monthly averaged air temperature (Tair) and precipitation (Prec) were also collected from the Complex Background Monitoring Station. The following 5 years were classified as “wet”—1999, 2003, 2006, 2008, 2020; the following 9 years were “dry”—2002, 2007, 2010, 2011, 2014, 2015, 2018, 2021, 2022; and the remaining 11 years stayed as “normal”—1998, 2000, 2001, 2004, 2005, 2009, 2012, 2013, 2016, 2017, 2019.
The following SR models are considered: T (dependent on T), TP (dependent on T, P), TPP (dependent on T, P, and Pm-1), TPPC (dependent on T, P, and Pm-1, and SOC), and TPPrh (dependent on T, T2, P, and Pm-1 - Raich-Hashimoto type). We also use the modeling of the soil temperature (TlitM, TsoilM) and moisture (MlitM, MsoilM) with the SCLISS model (Bykhovets and Komarov 2002) to see if the modeled values provide an appropriate substitution for the measured temperatures and precipitation.
We found out that two models from the ensemble set – TPPC and TPPrh have the best performances among all the models, which is also consistent with our previous findings (Kivalov et al. 2023). For the model comparison among different annual environmental conditions (norm, dry, wet) and different sources of the temperature (Tsoil, Tair, TlitM, TsoilM) and moisture (Prec, MlitM, MsoilM), we apply three metrics: R2 – determination coefficient, slope – the slope of the comparison between the measured and modelled SR values, and MBE – the mean bias error.
Focusing on the Entic Podzol soils, we found that for the forest SR with the low SOC, the direct Tsoil measurements in combination with SOC provide the better parametrization for the empirical TPPC model than if the only Tair parametrization was used. For the grassland SR with the relatively high SOC, the Tair becomes the governing factor, and the TPPrh model provides the better performance over all the considered models. Such a difference could be explained by both the SOC and temperature differences: the SOC becomes the limiting factor in the poor forest soil bringing forward the TPPC models, and the high Tsoil ≈ Tair values become the limiting factor for the grassland soil favoring for the TPPrh model.
When comparing the performances over the modeled with SCLISS temperatures (TlitM, TsoilM) and moistures (MlitM, MsoilM) data, we see that for both forest and grassland, the parametrization with the litter TlitM, and MlitM provides better results in all R2, slope, and MBE. These findings can identify the main source of the soil respiration as the topsoil or litter layer rather than the deeper soil layers. In wet conditions, using the litter modeled values (TlitM, MlitM) provide the best results for both TPPC and TPPrh models, while in normal and dry conditions the litter modelled values are slightly worser than the air parametrized values. This makes the parametrization with the SCLISS model for the litter temperatures and moistures the viable alternative, when the direct soil observations are not available on site.