Dry-season autotrophic and heterotrophic soil respiration in contrasting agroecological settings

ABSTRACT

In arid Africa, it is unclear how the transition from grassland to cropland affects CO₂ fluxes and whether these effects are climate-dependent. From November 2019 to March 2020 (dry season), we studied autotrophic (Ra) and heterotrophic (Rh) soil respiration. At highlands, grazing land (1.18 ± 0.13 µmol m⁻² s⁻¹) and teff croplands (1.36 ± 0.14 µmol m⁻² s⁻¹) had the lowest Rh, while acacia decurrens plantations (1.94 ± 0.19 µmol m⁻² s⁻¹) were the largest. When compared to teff (2.08 ± 0.20 µmol m⁻² s⁻¹) and khat (1.73 ± 0.24 µmol m⁻² s⁻¹), grazing land in the midland had a lower Rh (1.18 ± 0.13 µmol m⁻² s⁻¹). Ra on teff (1.51 ± 0.27 µmol m⁻² s⁻¹) was substantially (P < 0.05) greater than on grazing land (1.08 ± 0.32 µmol m⁻² s⁻¹) in the highlands. Grazing lands (1.54 ± 0.21 µmol m⁻² s⁻¹) have a considerably (P < 0.05) greater Ra than farmlands (teff, 0.20 ± 0.07 µmol m⁻² s⁻¹; groundnut, 0.09 ± 0.03 µmol m⁻² s⁻¹) in lowlands. Growing teff in the highlands and midlands will result in higher CO2 fluxes into the atmosphere. However, in-depth research is needed to implement effective land management that reduces CO₂ emissions.

KEYWORDS:

• Drought-prone
• land use
• soil moisture
• soil respiration partition
• soil temperature

Introduction

Soil respiration (Rs) or CO₂ flux is the mechanism by which soil releases carbon dioxide (CO₂). Annual global CO₂ fluxes from soil ranged from 68 to 87 Pg C yr⁻¹ (Warner et al. 2019). Rs is the sum of the autotrophic (Ra) and heterotrophic (Rh) respiration of soil organisms and plant roots, respectively. The worldwide terrestrial carbon cycle includes Rh, which is influenced by environmental factors (Bukombe et al. 2021). Rh is also influenced by microbial biomass such as fungi and bacteria (Li et al. 2006). To determine CO₂ emissions, it is necessary to quantify the contribution of Ra and Rh to Rs. The main uncertainty in projecting future climate-carbon feedback is the emission of CO₂ (due to Ra and Rh) in soils. An increase in SOC loss and subsequent climatic feedback could cause the Rh/Rs ratio to increase (Bond-Lamberty et al. 2018). The global response of Rh and Ra to climate change is important in predicting the future behaviour of the terrestrial carbon cycle. In subtropical China, coniferous forests have much lower Rh levels than three other types of forests with a predominance of broadleaf species (Wei et al. 2015). Ra’s vulnerability to temperature increased with land use, unlike that of Rh.

The potential for African soils to act as substantial carbon reserves is quite high (Epule 2015). Farmland modifications and forest destruction are getting worse in the Congo basin and other tropical areas, and this worsening is having an impact on carbon emissions (Verchot et al. 2020). About 60% of the African continent is composed of savannas, which are important contributors to global carbon emissions via land use and fire (Fan et al. 2015). Due to their vast expanse, savannas cause significant changes in global carbon levels that are difficult to understand and predict, explaining global warming. Heterotrophic microorganisms break down SOC, resulting in the release of CO₂ into the air. The overuse of South Africa’s Karoo rangeland is anticipated to lead to changes in ecosystem function, and measuring Rs may help to identify these changes (Mills et al. 2011). Africa’s growing participation in the carbon cycle could, have a major implications for future climate change. To understand how climate change affects the carbon cycling and accumulation, it is necessary to decompose Rs into its components. The Rs composition of soils has been measured using a variety of methods, with varying results from one study to another range of methodologies, and there is variation in the results among studies. However, root excavation experiments are commonly used to study Ra and Rh (Hu et al. 2018).

Soil respiration studies conducted in Ethiopia include those by Fekadu et al. (2023), who reported on the dynamics of soil respiration in land management practices. Michelsen et al. (2004) investigated soil CO₂ emissions during the wet season, and Teklay et al. (2006) studied the kinetics of soil respiration as a function of microbial nutrient status. There has never been a study of Ra and Rh anomalies in different agroecological regimes of different land uses in Ethiopia. Our objectives here were to (1) estimate the magnitude of Ra and Rh in dry months; (2) examine the effects of different land uses and agroecological systems on Ra and Rh; and (3) evaluate the factors contributing to variations in the components of Rs. We hypothesised that Ra and Rh differ significantly across land uses in different agroecologies.

Materials and methods

Study sites

Three agroecological systems were used for the research (Figure 1). Districts in which the study’s watersheds were located varied. The Dibatie site (10^o45′30″−10^o47′0″N, 36^o16′0″−36^o18′0″E), the Aba Gerima (11^o38′0″−11^o40′30″N, 37^o29′30″−37^o31′0″E), and the Guder watershed (10^o59′30″−11^o1′0″N, 36^o54′0″−36^o56′0″E) were used to represent the lowlands, midlands, and highlands, respectively (Berihun et al. 2019). These locations exhibit various land uses, annual rainfall patterns, and elevations in the major agroecological systems in the basin (Sultan et al. 2018). The predominant crops, major land uses, soil types, and climatic data for each watershed are listed in Table 1. We collected climate data from the Bahir Dar for the Aba Gerima, the Dangila and Enjibara for the Guder watershed, and the Chagni and Dibatie meteorological stations for the Dibatie watershed.

Figure 1. Study sites.

Table 1. Research site features.

Characteristic	Watershed
	Guder (highlands)	Aba Gerima (midlands)	Dibatie (lowlands)
Altitude (m)	2498–2857	1912–2126	1501–1718
Average annual rainfall (mm)	2462	1629	1377
Average annual temperature (°C)	16–20	18–23	20–25
Main land use ^a	Acacia decurrens plantation, bushland, forest, grazing land, cropland	Cropland, grazing land, bushland, Khat plantation (Catha edulis), and forest	Grazing land, bushland, cropland
Key crop type^b	Potato, wheat, barely, teff	Lupin, finger millet, teff, maize	Maize, finger millet, chili pepper, teff, groundnut
Major soil types^c	Acrisols, Luvisols	Luvisols, Leptosols	Vertisols, Luvisols

Potato (Solanum tuberosum); wheat (Triticum aestivum); maize (Zea mays); finger millet (Eleusine coracana); teff (Eragrostis tef); chili pepper (Capsicum annuum); barley (Hordeum vulgare), and lupin (Lupinus spp); groundnut (Arachis hypogea)

^a Berihun et al. (2019).

^b Mihretie et al. (2022).

^c Getahun (2016).

The growing season in the study area is somewhat different. Sowing everywhere begins in July. Harvest is expected in the midlands and lowlands in November. In the highlands, the harvest season continues into January. The dry season lasts from November to April. When looking at the onset of the dry season (November) in the midland and lowlands, crops are in their maturation stage or ready for harvest. However, this period was the tillering season, the beginning of the data collection, and January is the end of the harvest season in the highlands.

The major land uses in the study sites were teff cropland, plantations, and grazing land in the midlands and highlands. But in the lowlands, teff cropland, groundnut, and grazing lands were the dominant land uses. In the highlands, croplands accounted for roughly 41% of all land usage, followed by acacia plantations at 30% and grazing land at 9.5% (Sultan et al. 2018). The main land use in the midland was arable land, accounting for 51.64%, followed by plantations and grazing land at 8.9 and 3.43%, respectively (Mulualem et al. 2021). Lowland arable land accounted for 54.0% of the total land in 2017, with grazing land the second largest in 1982 (36.4%). However, grazing area decreased by 38.3% between 1982 and 2017, mainly due to arable land (Berihun et al. 2019).

Setup of experimental plots

In order to explore Rh and Ra in the three contrasting watersheds, we established experimental plots with various land uses in 2019. Fifty-four fixed test plots were set up (3 replicates, 3 land uses, 2 variants [with or without roots], and 3 watersheds). We inserted a polyvinyl chloride (PVC) collar in each experimental plot at the commencement of the Rs and Rh measurements, 0.05 meters below the soil’s surface. The PVC collar used to define the area for measurement of Rh and Rs was 0.19 meters in diameter and 0.11 meters long (Fekadu et al. 2023). Experiments were conducted where the slope was nearly uniform in each watershed. The mean elevation in the highlands was 2585, 2586, and 2587 m a.s.l. for grazing land, A. decurrens, and teff cropland, respectively. In the midlands, the average elevations for teff cropland, grazing land, and khat plantations were 1992, 1991, and 1989m a.s.l., respectively. There were no elevation data available for the lowlands study site because data collection had been interrupted due to security problems. The land uses were located at adjacent sites in each watershed. The experimental plots were designed to measure Rs and Rh. One meter of soil was removed from the ground. The soil was then filled in after any roots had been manually removed. To ensure the soil’s stability and the accuracy of the Rh measurements, before placing the PVC collar, we waited 6 months after we had replaced the soil in each hole (Figure 2). To avoid root respiration, we cut plants or any weed without disturbing soils before measuring data at Rh plots. On the same date, we measured soil respiration at the Rs and Rh plots. We subtract Rh value from Rs to get Ra. All plots were 1 meter square. To avoid further variations, the distance between paired plots with and without roots was about 1 meter for each land use.

Figure 2. Preparation of the experimental area and measurement of soil respiration. (a) Excavation of the soil, (b) refilling of the hole, (c) polyvinyl chloride collar inserted into the experimental plot, and (d) measurement of heterotrophic respiration.

Data collection

The Li 8100A, a device for measuring Rs rates, was used to collect data between November 2019 and March 2020. Each month, Rh and Rs are measured between 9:00 and 12:00 am (Fekadu et al. 2023). For measurements using dynamic closed-chamber systems, the chamber had to be placed on the PVC collar on the soil surface for 60–120 s (Rochette et al. 1997). We measured Rh and Rs for 90 s. Since the machine records every second, an average of Rs or Rh was used for each collar. Measurements were subjected to a 90% coefficient of determination of the lower limit prior to analysis (Savage et al. 2008). Along with data on soil respiration (Rs and Rh), the soil’s moisture and temperature were simultaneously monitored. Moisture content was monitored monthly by using a hand auger to collect approximately 200 g of soil from each plot at a depth of 0.05 m, as used for the Rs measurement. Soil samples were kept in an oven at 105°C for 24 h until a stable weight was reached. To calculate the soil moisture (S_m), the difference in mass between the wet and dry soil samples was divided by the mass of the dry sample and multiplied by 100. A soil temperature (S_T) omega probe, module number 6000-09TC (Li-Cor Inc.), was used to estimate S_T at a depth of 0.05 m.

Sampling and statistical analysis

Soil samples were taken from the experimental site at a depth of 20 cm (Rs and Rh plots within 1 m²) for further analysis. After the soil samples were dried in the air at room temperature, they were crushed in a mortar and pestle and sieved through a sieve having a mesh size of 2 mm. Soil pH and texture were studied using soil samples sieved with a 2 mm mesh sieve. For total nitrogen (TN) and soil organic carbon (SOC) analyses, the soil was ground and soil samples were sieved through a 0.5 mm sieve. Soil pH was measured in a supernatant suspension of a mixture of soil and distilled water in a 1:2.5 ratio using a pH meter (Peech 1965). Soil organic carbon (SOC) was measured by the dichromate oxidation method of Walkley and Black (1934). Total nitrogen (TN) was titirimetrically determined using the Kjeldahal method described by Nelson and Sommers (1980). Soil texture was analyzed using a hydrometer (Bouyoucos 1962). The bulk density of the soil was determined by the core method.

A one-way analysis of variance (ANOVA) was used to examine statistical differences in soil properties between major land uses. Least Significant Difference (LSD) tests were used to detect significant differences (P < 0.05) between soil properties for major land uses. The data were checked for normality for Rh and Ra. The normality checks showed that all of the Rh and Ra data were non-normal. In order to analyze the data, Friedman’s non-parametric test was applied. Friedman’s analysis was followed by a Wilcoxon post-hoc analysis (P < 0.05) to separate means. Pearson correlations were also used to study the relationship between Rs components, environmental factors, and soil properties.

The following function was used to estimate Rh or Ra in S_T (Huang et al. 2018): (1) $Rh\left(a\right)=a{e}^{b{S}_T},$(1) where Ra or Rh is the CO₂ flux in µmol m⁻² s⁻¹; S_T – soil temperature (^oC); a and b are the model coefficients.

The value of Q10 (temperature sensitivity) was calculated by substituting parameter b into Equation 2 (Meyer et al. 2018a) (2) $Q10=e^{10\ast b}$(2) Where b is the coefficient obtained from Equation 1

From S_m data, either Rh or Ra might be predicted using the second-order equation below (Peng et al. 2015): (3) $Rh\left(a\right)=c{S}_m^2+b{S}_m+a,$(3) Where, S_m is soil moisture in %; a, b and c are the model coefficients.

A regression model was applied to predict Ra/Rh in S_T and S_m (Cui et al. 2020): (4) $Rh\left(a\right)=a+b{S}_T+c{S}_m,$(4) In the equations above, all variables are as specified

Results

Characteristics of soil

Soil physic-chemical analysis was performed for bare soil (without roots to evaluate Rh) and for soil covered in the relevant grassland, crop, or plantation to evaluate Rs. No significant differences (P > 0.05) were found between soil parameters by land-use type in the highlands at Rh plots. In the midlands TN, clay content varied significantly (P < 0.001) between land uses, and soil pH varied significantly (P < 0.001). In lowlands, sand and silt varied significantly (P < 0.05) across land use. At the same time, pH, BD, and clay content were statistically different at P < 0.01 (see Table 2).

Table 2. Topsoil (0–20 cm) sample characteristics (mean ± SE) in the bare soil (Rh plots) (n = 3).

Site	Land use	SOC	TN	pH	BD	Sand	Silt	Clay	STC
		(%)	(%)	1:2.5(H2O)	(g cm^−3)	(%)	(%)	(%)
Guder	TC	2.61 ± 0.66	0.13 ± 0.03	5.28 ± 0.1	1.45 ± 0.04	31 ± 4.4	32 ± 5.7	36 ± 8.8	CL
	GL	2.25 ± 0.81	0.10 ± 0.02	5.18 ± 0.11	1.39 ± 0.02	27 ± 0.67	21 ± 3.1	52 ± 3.5	C
	AP	1.47 ± 0.56	0.11 ± 0.04	5.34 ± 0.16	1.37 ± 0.06	24 ± 7.0	20 ± 4.4	56 ± 5.9	C
	P-value	ns	ns	ns	ns	ns	ns	ns
Aba Gerima	TC	1.51 ± 0.31	0.13 ± 0.01^a	5.67 ± 0.06^b	1.41 ± 0.11	43 ± 4.7	27 ± 1.3	30 ± 4.2^a	CL
	GL	0.68 ± 0.12	0.07 ± 0.05^b	5.97 ± 0.07^a	1.47 ± 0.07	59 ± 9.8	26 ± 7.2	15 ± 2.7^b	SL
	KP	1.55 ± 0.27	0.07 ± 0.04^b	5.14 ± 0.05^c	1.38 ± 0.03	42 ± 1.2	24 ± 1.2	34 ± 1.2^a	CL
	P-value	ns	**	***	ns	ns	ns	**
Dibatie	TC	2.38 ± 0.45	0.12 ± 0.02	5.44 ± 0.10^b	1.29 ± 0.03^a	14 ± 0.1^b	14 ± 1.8b^c	72 ± 1.8^a	C
	GC	1.90 ± 0.20	0.11 ± 0.02	5.25 ± 0.02^b	1.41 ± 0.06^a	17 ± 2.4^ab	12 ± 1.8^c	71 ± 1.2^a	C
	GL	2.70 ± 0.08	0.16 ± 0.01	5.83 ± 0.08^a	1.13 ± 0.02^b	22 ± 1.2^a	21 ± 1.8^a	57 ± 2.9^b	C
	P-value	ns	ns	**	**	*	*	**

Within a column, treatments that don’t contain the same letter vary considerably from one another. ***, **, * indicate that the p-values of ANOVA are significant at P < 0.001, P < 0.01, and P < 0.05, respectively. ns, not significant; SBD, soil bulk density; TN, total nitrogen; SOC, soil organic carbon; STC, soil texture class; C, clay, CL, clay loam; SL, sandy loam; GL, grazing land; TC, teff cropland; KP, khat plantation; AP, Acacia decurrens plantation; GC, groundnut cropland

One-way ANOVA results for Rs plots showed that in highland soil, pH was statistically (P < 0.05) lower in teff cropland or Acacia decurrens plantations than in grazing land (Table S1). In the midland and lowlands Rs plots, pH varied significantly (P < 0.05) across land use, where SOC, TN, sand, and clay were statistically different at P < 0.01 (Table S1).

Monthly variations in Rh and Ra in different agroecological systems and with different land uses

On grazing land, the lowest Ra value was recorded in the midlands in January, and the highest was recorded in the highlands in November. In December, the lowlands on teff had the smallest mean of Ra and the highest in the highlands in November (Figures 3(a)–5(a)). The lowest Rh value on grazing land was recorded in March in the lowlands, followed by the midlands in March and the highlands in November. The highest Rh value on grazing land was recorded in the lowlands in January, followed by the midlands in January and the highlands in December. The lowest and highest Rh values on teff cropland were found in the lowlands in March and the midlands in January, respectively (Figures 3(b)–5(b)).

Figure 3. Monthly soil respiration rates at Guder for various land uses: Ra, autotrophic soil respiration (a); Rh, heterotrophic soil respiration (b); Rs, soil respiration (c); and Rh/Rs (d). Note: TC, teff cropland; GL, grazing land; AP, acacia decurrens plantation.

Figure 4. Monthly soil respiration rates at Aba Gerima for various land uses: Ra, autotrophic soil respiration (a); Rh, heterotrophic soil respiration (b); Rs, soil respiration (c); and Rh/Rs (d). Note: TC, teff cropland; GL, grazing land; KP, khat plantation.

Figure 5. Monthly soil respiration rates at Dibatie for various land uses: Ra, autotrophic soil respiration (a); Rh, heterotrophic soil respiration (b); Rs, soil respiration (c); and Rh/Rs (d). Note: TC, teff cropland; GC, groundnut cropland; GL, grazing land.

In the highland catchment, the lowest average Ra CO₂ fluxes at grazing land (0.32 ± 0.11µmol m⁻² s⁻¹) and teff (0.78 ± 0.16 µmol m⁻² s⁻¹) were measured in March. Conversely, the lowest average flux of Ra (0.36 ± 0.11µmol m⁻² s⁻¹) in the A. decurrens was observed in February. In November, teff (2.81 ± 0.41µmol m⁻² s⁻¹) recorded the highest average Ra at highland, preceded by grazing land (2.42 ± 1.32 µ mol m⁻² s⁻¹) and A. decurrens (1.82 ± 0.75 µmol m⁻² s⁻¹) (Figure 3(a)). In midland, the mean Ra value varied from 0.12 ± 0.07 (January) to 2.29 ± 0.54 (December) for grazing land; from 0.85 ± 0.22 (March) to 2.17 ± 0.15 (November) for teff; and from 0.1 ± 0.05 (March) to1.62 ± 0.22 µmol m⁻² s⁻¹ (December) for khat (Figure 4(a)). The lowest average Ra CO₂ fluxes in lowland were measured in March for land use grazing and (0.92 ± 0.29) and groundnut (0.03 ± 0.01 µmol m⁻² s⁻¹). On teff, the lowest Ra was recorded in December. The greatest Ra values were discovered on grazing land in February, teff in March, and groundnut in December (Figure 5(a)).

The lowest mean Rh CO₂ fluxes at the highland site were found at grazing land in November, teff in March, and A. decurrens in March. The highest mean grazing land and A. decurrens values were observed in December, whereas CO₂ flux on teff cropland was highest in November (Figure 3(b)). At the midland site, the lowest Rh CO₂ flux values were in decreasing order: khat in March, teff in December, and grazingland in March. Similarly, the peak Rh CO₂ flux values with the same decreasing trend were recorded at khat in November, teff in January, and grazing land in January (Figure 4(b)). Rh CO₂ flux values in lowlands were generally lower than in the other agroecologies. In grazing land, Rh (µmol m⁻² s⁻¹) ranged from 0.33 ± 0.11 in March to 3.18 ± 0.64 in January. The lowest and highest values of Rh CO₂ flux at teff cropland were in March and December, respectively. For groundnut, the months of January and March had the most and the least Rh CO₂ flux levels, respectively (Figure 5(b)).

Examination of the mean Rh as a percentage of Rs at highland showed that A. decurrens (71%) had the highest percentage, followed by grazing land (58%) and teff cropland (49%) (Figure 3(d)). At midland, the lowest mean Rh/Rs percentage (57%) was observed on grazing land and the highest (71% on khat) (Figure 4(d)). At lowland, groundnut (83%) and teff (77%) had higher mean Rh/Rs percentages than grazing land (46%) (Figure 5(d)).

Agroecological systems and land use influences in Ra and Rh

The mean Rh value of grazing land was higher in the lowlands than in the other two areas, whereas it was the same in the highlands and midlands (Table 3). For teff cropland, the highest mean rank Rh value was recorded in the midlands and then the highlands, followed by the lowlands. From the highlands through the midlands to the lowlands, the overall mean Ra value exhibited a rising tendency for grazing land and a steady decline for teff. The mean Ra and Rh values for plantations were greater in the highlands than in the midlands (Table 3). The maximum Rh value in the highlands was found in A. decurrens, which was accompanied by teff and grazing land. In the midlands, Rh was higher on teff and khat than on grazing land. In lowland areas, no significant difference in Rh value was observed; however, grazing land had the highest Rh values, preceded by teff and groundnut. Furthermore, in the highlands, the Ra value on teff was markedly (P < 0.05) greater than on grazing land. In the midlands, the Ra measurements at different land uses did not differ significantly from each other. However, the highest Ra values were found for teff (1.40 ± 0.26), GL (1.16 ± 0.28), and khat (0.78 ± 0.17). In the lowlands, grazing land had a statistically (P < 0.05) higher Ra than teff and groundnut (Table 3).

Table 3. ANOVA results for Ra and Rh (µmol m⁻² s⁻¹) at various land uses in different agroecological systems (n = 3 for each land use for each month).

Site	Land use	Rh			Ra
		Average ± SE	Median	MR	Average ± SE	Median	MR
Guder	TC	1.36 ± 0.14	1.18	1.73^b	1.51 ± 0.27	1.14	2.57^a
	GL	1.18 ± 0.13	1.11	1.67^b	1.08 ± 0.32	0.72	1.60^b
	AP	1.94 ± 0.19	2.01	2.60^a	0.88 ± 0.22	0.59	1.83^ab
	χ^2	8.13			7.76
	df	2			2
	P-value	0.017			0.021
Aba Gerima	TC	2.08 ± 0.20	1.94	2.47^a	1.40 ± 0.26	1.28	2.40^a
	GL	1.18 ± 0.14	1.25	1.53^b	1.16 ± 0.28	1.08	1.93^a
	KP	1.73 ± 0.24	1.3	2.00^a	0.78 ± 0.17	0.5	1.67^a
	χ^2	6.53			4.13
	df	2			2
	P-value	0.038			0.127
Dibatie	TC	0.80 ± 0.13	0.69	2.08^a	0.20 ± 0.07	0.09	1.67^b
	GC	0.45 ± 0.07	0.41	1.50^a	0.09 ± 0.03	0.08	1.42^b
	GL	1.62 ± 0.34	1.51	2.42^a	1.54 ± 0.21	1.28	2.92^a
	χ^2	5.17			15.5
	df	2			2
	P-value	0.076			0.0001

Land use in the same column with different letters is significantly different (P < 0.05). SE = standard error; MR = mean rank. Note: TC, teff cropland; GC, groundnut cropland; GL, grazing land; AP, acacia decurrens plantation; KP, khat plantation; Ra, autotrophic soil respiration; Rh, heterotrophic soil respiration.

Factors influencing Ra and Rh

In the highland study area, there were no statistical relationships found between soil variables and Rh, although there were trends toward positive correlations of Rh with SOC, clay content, and soil pH (Table S2). At midland, Rh was significantly correlated (r = 0.75*) with clay content. SOC and TN also showed non-significant positive relations with Rh (r = 0.04 and 0.68, respectively). A significant correlation was observed between Rh and the pH of the soil (r = 0.77*), silt (r = 0.78*), and clay (r = −0.83*) in the lowland study area. SOC and TN also demonstrated a non-significant trend toward positive association with Rh (r = 0.47 and 0.60, respectively) (Table S2).

Rh and S_m strongly associated in highland study site (r = 0.42**); however, S_T did not significantly connect with Rh (r = 0.05) (Table S2). In the midland study area, neither S_m nor S_T was significantly correlated with Rh. In the lowland study area, S_m was not substantially connected with Rh, while S_T was (r = −0.33*) (Table S4). Rh strongly (P < 0.05) associated exponentially with S_T across all research sites. S_m was significantly associated with Rh in quadratic function at highland but not at midland or lowland (Table 4). Additionally, only at the highland location did S_m and S_T show a strong connection with Rh.

Table 4. Interaction between environmental variables (S_T and S_m) and Rh.

Site	$Rh=a{S}_T+b{S}_m+c$ (Multiple regression model)
	a	B	c	R^2
Guder	0.003	0.023	0.71	0.18*
Aba Gerima	−0.01	4.80	2.51	0.16
Dibatie	−0.06	0.01	2.51	0.14

Rh, heterotrophic soil respiration (µmol m⁻² s⁻¹), S_T is the soil temperature in degree Celsius; S_m is the soil moisture in percent, and a, b, c are the model coefficients. For each plot and month, n = 3 for each site and month.

At the midland and highland sites, no conclusive relationships between soil variables and Ra were observed. Meanwhile, Ra was significantly (P < 0.05) related to pH, sand, silt, and clay content at the lowland research site (Table S3). In the highland study site, Ra was substantially (P < 0.01) correlated with S_m and S_T (Table S4). While S_m (r = 0.35**) was substantially connected with Ra in midland, S_T (r = 0.18) had a non-significant tendency toward a positive connection with Ra. In lowland, no significant (P > 0.05) correlation was found between Ra and S_m or S_T (Table S4).

There was a significant (P < 0.05) exponential relationship between S_T and Ra at all study sites. In the quadratic function, Ra was strongly (P < 0.05) correlated to S_m in highland but not in midland or lowland. The interaction between S_m and S_T was also strongly (P < 0.05) associated with Ra in midland and highland (Table 5).

Table 5. Multiple regression models between Ra and environmental variables (S_T and S_m).

Site	$Ra=a{S}_T+b{S}_m+c$ (Multiple regression model)
	a	b	c	R^2
Guder	0.05	0.04	–1.45	0.39*
Aba Gerima	0.03	0.03	–0.55	0.17*
Dibatie	0.01	0.01	0.33	0.01

Ra is autotrophic soil respiration (µmol m⁻² s⁻¹). Note: the footnotes of Table 4 provide explanations for all variables. Every month, n = 3 per plot.

Discussion

Effects of land-use type on soil properties in various agroecologies

The SOC in the midlands increased when other land uses replaced grazing land, but the increase was not significant. In the highlands, land-use change increased the SOC on teff but decreased it on A. decurrens. In the lowlands, a land-use shift from grazing land to other land uses decreased the SOC. It is possible that these variations in SOC are related to variations in land-management practices. In the midlands on grazing land, we noticed overgrazing and more soil degradation than with other land uses. In the case of the highland watershed, we noticed good land management practices on grazing land that were being used in a controlled manner. However, on the lowlands, the grasses are somewhat thorny, and it is possible that this made the pastures resistant to overgrazing. Rh values increased with a change from GL to other land uses in the highlands and midlands. In other land uses, there are different inputs that contribute to organic matter in the soil. Research done by Hao et al. (2010), which revealed that a decline in SOC was the primary cause for low soil respiration, confirms this finding. Conversely, on the lowlands, the Rh values decreased with a change from grazing land to other land uses (see Table 3).

Monthly variations in Ra and Rh in different agroecological systems and under different land uses

Monthly variations in Rh values were observed according to land use in each study area. From November through March, all sites showed a steady decline for Rh (see Figures 3(b)–5(b)). At the three study sites, the lowest Rh values were measured in March. The main reason for the variation in Rh values could be the different S_m contents in different months under different land uses. In highland, a strong association between Rh values and S_m levels was observed (Table S2). One study reported that CO₂ emissions were highest in the months with the highest S_m (Valente et al. 2021). Additionally, moist soil conditions improve soil microbial physiological processes and root respiration, which increases Rs (Takahashi et al. 2011). One of the most significant ecosystem factors influencing production and carbon cycling in terrestrial ecosystems is S_m (Moyano et al. 2012).

Ra varied across months and under various land uses. There is a statistically significant association between S_m and Ra in midland and highland. Additionally; S_T at highland had a significant relationship with Ra (Table S2). The study by Herbst et al. (2016) found that S_T did not account for a large portion of the fluctuation in Ra.

Impact of land use across different agroecological systems on Ra and Rh

In the highlands, we found a significantly (χ²= 8.13, df = 2, P = 0.017) higher Rh value with A. decurrens than with the teff and grazing land uses. This implies that land-use conversion from grazing land or teff to A. decurrens contributes to an increase in Rh in the highland study areas. A possible reason for this could be S_m level (Hu et al. 2018). The average S_m for the A. decurrens land use was 34.44%, which was higher than the values for the teff (28.16%) and grazing land (29.46%), according to our measurements. At the highland research site, S_m was also substantially connected with Rh (r = 0.42*) (Table S4). However, this site’s soil characteristics and Rh did not significantly correlate (Table S2). In the midlands, the teff and khat land uses had significantly (χ²= 6.53, df = 2, P = 0.038) higher Rh values than that of the grazing land (see Table 3). At the midland watershed, land-use change from grazing to khat or teff cropland contributed to a higher Rh. The most probable reason could be changes in the soil properties. On teff and khat, the clay content was found to be considerably (P < 0.01) higher than on grazing land. In addition, the soil texture varied. The soil texture class for teff and khat was clay loam, whereas that for grazing was sandy loam (see Table 2).

At midland, there was a substantial association between clay content and Rh (r = 0.75*) (Table S2). The amount of clay in the soil affects the amount of moisture. The addition of clay to sandy soils improves the water retention capacity by two to six times (Tahir and Marschner 2016). The primary element influencing soil CO₂ emissions is the fluctuation in S_m (Valente et al. 2021). Rh is also affected by the presence of substrate and the soil environment. In environments with high substrate availability, the potential for carbon sequestration is lower and carbon release is greater (Aguilos et al. 2013). Soil Rh determines whether an ecosystem serves as a carbon sink or source. S_m influences Rs (Meyer et al. 2018b), and soil properties affect the response of Rh to moisture (Shahbaz et al. 2022).

Ra was substantially (P < 0.05) greater on teff than on grazing land at the highland site (Table 3). According to the ANOVA findings at highland, the soil pH was substantially (P < 0.05) high on teff than on grazing land (Table S1). The pH of the soil affects its biochemical processes and fundamentally affects soil behaviour (Neina 2019). In our study, S_m (r = 0.56**) and S_T (r = 0.39**) were significantly associated with Ra at the highland site (see Table S4), and at lowland, we observed significantly (P < 0.05) higher Ra values on grazing land than on groundnut cropland. The reasons for this could be related to soil characteristics. The ANOVA results depicted that, at lowland, TN and SOC were considerably (P < 0.05) greater on grazing than on groundnut cropland. Moreover, pH of soil was substantially (P < 0.01) greater on grazingland compared to other land uses (Table S1). Autotrophic soil processes are closely coupled to carbon production by photosynthesis (Schindlbacher et al. 2009). A statistical association was found between Ra and soil pH, sand content, and silt content at Dibatie (see Table S3).

Factors contributing to Ra and Rh

Both Rh and Ra increased exponentially (P < 0.05) with S_T across all study sites (Figures 6 and 7). Moreover, we found a significant linear relationship between Rh and S_T (R²= 0.11) at lowlands (Figure 6(c)), and a significant linear relationship (R²= 0.15) between Ra and S_T (Figure 7(a)). Rh levels increase with soil temperature. Soil microbial respiration is also temperature dependent. Higher temperatures can accelerate Rh, enhancing the activation of enzymes that degrade organic matter in the soil (Blagodatskaya et al. 2016). The main causes of this destruction are fungi and bacteria (Balogh et al. 2015). The components of soil respiration (Rh and Ra) are less dependent on the temperature in the midlands and lowlands than in the highlands. The reason is that the Q10 value is relatively high in the highlands, followed by the midland and lowlands in that order (Figures 6(a–c) and 7(a–c)).

Figure 6. Relationship between Rh, heterotrophic soil respiration (µmol m⁻² s⁻¹) with soil temperature (^Oc) and soil moisture (%) at the three study sites.

Figure 7. Relationship between Ra, autotrophic soil respiration (µmol m⁻² s⁻¹) with soil temperature (^Oc) and soil moisture (%) at the three study sites.

S_m made significantly contributions to Rh and Ra in both linear and quadratic functions at the highland study site (see Figures 6(d) and 7(d), Table S4). One explanation for this is that the highland research site has a long rainy season, and the soils there may thus contain enough moisture for soil organisms. Rs increases with S_m, indicating that microbial activity is most by S_m (Thomas et al. 2011). It is generally accepted that S_m and S_T affect Rh and Ra (Wei et al. 2015). Ra is more sensitive to S_T than Rh, but Rh is more sensitive to S_m (Chen et al. 2017).

Multiple regression in the analysis showed that the combination of S_m and S_T significantly affected Ra and Rh in the highlands (Tables 4 and 5). Our finding is supported by the study (Cai et al. 2021). No significant correlation was observed in the lowlands. Our study is supported by Hinko-Najera et al., (2015), who reported that the effects of S_m and S_T on the Rs component were not noticeable in lowlands.

In conclusion, Rs components varied on monthly basis across agroecologies and land use. The Ra values of teff cropland at highland and midland were higher than those of other land uses, whereas at lowland they were lower than those on grazing land. Similarly, the Rh value of teff cropland at midland was higher than other land uses; in highland Rh value was higher in acacia plantation than other land uses. These results suggest that growing teff in the relatively humid regions (highland and midland) will result in higher CO₂ fluxes into the atmosphere. S_m was significantly associated (P < 0.01) with Ra and Rh at highland, but only S_m was statistically (P < 0.05) related to Ra at midland. However, in the lowlands, S_m was not statistically associated with components of Rs. S_T and Ra only strongly (P < 0.01) associated in the highlands. Thus, the components of Rs at highlands continue to depend more on S_T and S_m than at lowlands. An exponential function for S_T (R²= 0.24*), a second-order polynomial function for S_m (R²= 0.34*), and multiple regression for both S_T and S_m (R² = 0.39*) explain the greater variation in Rh at highlands than any other agroecology. In general, different factors contributed to the variations in the components of Rs at the study sites. Consequently, more studies are needed to determine the effects of different land uses on Rh and Ra in order to predict future Rs components and suggest appropriate land management options.

Acknowledgements

We acknowledged the Japan Science and Technology/Japan International Cooperation Agency for funding this research through the Japan Science and Technology Research Partnership for Sustainable Development (SATREPS, grant number JPMJSA1601). We would also like to express our gratitude to the SATREPS project team at Bahir Dar University, Ethiopia, for their excellent collaboration in the laboratory and fieldwork.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data analyzed during the current study are available from the corresponding author on rational request.

Additional information

Funding

This study was funded by the Science and Technology Research Partnership for Sustainable Development (SATREPS, grant number JPMJSA1601), and the Japan Science and Technology Agency (JST) / Japan International Cooperation Agency (JICA).

Notes on contributors

Genetu Fekadu

Genetu Fekadu (M.Sc.) is a PhD candidate at Bahir Dar University in the soil science program and a lecturer at the University of Gondar, Ethiopia.

Enyew Adgo

Enyew Adgo (PhD) is a Professor and researcher at Bahir Dar University, Ethiopia.

Derege Tsegaye Meshesha

Derege Tsegaye Meshesha (PhD) is an Associate Professor and researcher at Bahir Dar University, Ethiopia.

Atsushi Tsunekawa

Atsushi Tsunekawa (PhD) is a Professor and researcher at Tottori University, Japan.

Nigussie Haregeweyn

Nigussie Haregeweyn (PhD) is a Professor and a researcher at Tottori University, Japan.

Fei Peng

Fei Peng (PhD) is an Associate Professor and a researcher at the Chinese Academy of Sciences, China.

Mitsuru Tsubo

Mitsuru Tsubo (Professor) is a Professor and researcher at Tottori University, Japan.

Asaminew Tassew

Asaminew Tassew (PhD) is an Associate Professor and researcher at Bahir Dar University, Ethiopia.

Gizachew Ayalew Tiruneh

Gizachew Ayalew Tiruneh (PhD) is an Assistant Professor and lecturer at Debre Tabor University, Ethiopia.

Nigus Tadesse

Nigus Tadesse (M.Sc.) is a PhD student at Bahir Dar University, College of Agriculture and Environmental Sciences, Ethiopia.

Tsugiyuki Masunaga

Tsugiyuki Masunaga (Professor) is a Professor and researcher at Shimanie University, Japan.