Pasture Restoration Reduces Runoff and Soil Loss in Karst Landscapes of the Brazilian Cerrado

Abstract

Water erosion is a major driver of soil degradation in the Brazilian Cerrado, intensified by the conversion of natural vegetation into agricultural land. The excessive runoff and sediment generated in poorly covered karst slopes impact the hydrologic cycle of the biome’s sinkholes and underground rivers. This study evaluated the effectiveness of pasture restoration in reducing runoff and soil loss in three experimental farms situated in a vulnerable karst area of Central Brazil. Runoff and soil loss were monitored during three hydrologic years in plots of degraded pasture (DP), restored pasture (RP), and natural savannah (NS), using unbound Gerlach settings. The experiment was carried out on three farms in the Vermelho river basin, which were treated as blocks. The results indicate that pasture restoration reduced runoff by 50% and soil loss by 55–95% when compared to degraded pasture conditions, below on-site erosion tolerance thresholds. Runoff and soil loss in restored pasture (RP) plots fell between DP and NS, though in some cases, soil loss in RP reached levels that are comparable to the natural savannah. Normalized soil loss was highly correlated with runoff (R² = 0.94), allowing for the latter to be used as a proxy of the former. The increased groundwater recharge and reduced sediment yield resulting from pasture restoration improve on- and off-site resilience in vulnerable karst landscapes and could be utilized as a sustainable soil conservation policy.

1. Introduction

Water erosion is a primary driver of soil degradation which is aggravated by the removal of natural vegetation, resulting in decreased land productivity [1,2]. Permanent vegetation is considered to be the only natural component of soil protection [3], stabilizing the soil through its canopy and root system, enhancing infiltration, reducing surface runoff [4] and retaining soil moisture [5].

In addition to other on-site impacts, such as soil degradation and declining agricultural productivity [6], soil erosion also generates off-site effects, such as downstream silting and eutrophication [7], nutrient enrichment of sediments [8,9], and the consequent deterioration of downstream water quality [10,11,12,13].

The Brazilian Cerrado, in Central Brazil, has been increasingly subjected to anthropogenic pressures, with profound impacts on its natural landscape [13], including stressed aquifer systems [14] and excessive soil erosion [15]. More than 50 million hectares of the Cerrado biome are now at high risk of water erosion, a process intensified by a combination of high rainfall erosivity, high soil erodibility, and the conversion of natural vegetation into agricultural land [16], culminating in permanent soil degradation [17]. It was shown that soil loss is positively correlated to and even predicted by surface runoff. However, maintaining permanent soil cover, such as natural vegetation and well-managed pastures, is critical for minimizing runoff generation and reducing soil erosion [18].

Within the Brazilian Cerrado, there are highly vulnerable karst areas, such as the Vermelho river basin, where poorly structured sandy soils, lying over fractured limestones, are severely affected by soil erosion, leading to the silting of sinkholes and subterranean rivers [3].

Pasturelands, which dominate the majority of the Brazilian Cerrado landscapes, have been severely degraded in the last 50 years, largely due to erosion and inadequate soil management [19,20]. Natural restoration of these areas has proven insufficient, as it fails to restore key soil properties and ecosystem functions [19].

In contrast, active reclamation strategies, such as pasture restoration and the establishment of adequate vegetation cover, have significantly reduced erosion and improved ecosystem services [21,22], while enhancing the well-being of rural communities [23]. However, effective pasture restoration requires adaptive management, including regulated stocking rates to allow for regrowth and the creation of appropriate soil cover, and cattle fencing during critical germination and seedling periods [24]. Additionally, effective restoration is achieved when on- and off-site soil loss in the landscape is reduced below the accepted tolerance values, i.e., below 10 Mg ha⁻¹yr⁻¹ and 1.0 Mg ha⁻¹yr⁻¹, respectively [18].

In a pioneer study carried out in a karst area of the Brazilian Cerrado, it was found that active pasture restoration with grasses and shrubs reduced runoff by 50% and soil loss by 80%, with respect to fallow, uncovered areas [3]. However, the research looked at one soil type only and could not be generalized to different landscapes.

To assess runoff and soil erosion under different land-uses, several methods have been used, including standard (bound) runoff plots [3] and erosion pins [25], which are complex and time consuming [26]. Simpler devices, such as unbound Gerlach troughs [26,27] are more practical and inexpensive to install and to operate, although non-cumulative runoff collection is a challenge [24]. Additionally, traditional bounded runoff plots are unnatural erosion monitoring setups, since plot borders of bounded plots force runoff and sediment in a pre-determined direction, affecting the natural flow path [28].

Considering the existing data and knowledge gaps about the theme, the objective of this research was to assess the effectiveness of pasture restoration in different soils, land-use, and landscape settings of the karst landscape of the Brazilian Cerrado, and to evaluate the ecosystem services provided with respect to runoff and erosion abatement, using unbound Gerlach plots, and to evaluate the relationship between them.

2. Materials and Methods

2.1. Study Area

The study area consisted of three cattle farms located at the Vermelho basin, in the Brazilian Cerrado, situated 20–30 km apart (Figure 1).

The selected experimental sites had different soils but similar topographies, land use and management practices (

Table 1. Topography, soils, and land uses of the three experimental sites.

Site	Topography (Slope Grade)	Soil Type	Land Uses
Funil farm	Gentle (9.3%)	Red Orthox (sandy loam)	Pasture and savannah
Progresso farm	Gentle (6.8%)	Psamment (sand)	Pasture and savannah
Tarimba farm	Moderate (11.5%)	Orthent (silt loam)	Pasture and savannah

), spanning the typical landscape variability of the Vermelho river basin.

Because of intrinsic soil and landscape characteristics of the three sites, each one was considered to be an experimental block, following the randomized complete block design (RCBD), with one factor (land use and management) and three treatments (DP, RP, NS). This design is recommended for large experimental units that must be subdivided, grouping them into relatively homogeneous subsets before assigning treatments [29].

Degraded pasture (DP) was characterized by reduced grass cover (0–20% ground cover), planted with Brachiaria sp. and Andropogon sp., high soil compaction [30], assessed by a ring penetrometer, low infiltrability, and low fertility and organic matter. The plot areas of degraded pasture (DP) in the three sites were between 0.5 hectares and 1.0 hectare (

Table 2. Penetration resistance and final infiltration rate of the degraded pastures (DP) in the three experimental sites.

Experimental Site	Penetration Resistance, 0–10 cm (MPa)	Infiltration Rate (cm min−1)
Funil farm	3.33	0.28
Progresso farm	0.78	0.91
Tarimba farm	1.95	0.80

).

Adjacent to the degraded pasture (DP) plots, restored pasture (RP) plots were established on formerly degraded pastureland, with areas varying between 0.5 and 1.0 ha. Pasture restoration was implemented during the dry season preceding the first monitoring year and included disc ploughing (25 cm), liming (2.0 Mg ha⁻¹ of dolomitic limestone) and fertilization (240 kg of 4-14-8 NPK granular fertilizer), herbicide (2 L ha⁻¹ of glyphosate), planting of Massai grass (Panicum maximum v. Massai), and control of leafcutter ants (10 kg ha⁻¹ of sufuramid). One year after planting, the Massai grass cover was suitable for grazing, and cattle were allowed to graze in the three restored plots.

The natural savannah (NS) plots comprised small remnants of original vegetation cover (0.2 to 2.0 ha) within each experimental farm. These sites contained small trees (3 to 5 m), covering approximately 50% of the area, surrounded by shrubs and natural grasses [31].

Within each block (farm), the NS, DP, and RP treatments were installed in the same soil type and had similar slope gradients (Table 1), allowing for the plots to be taken as independent treatments (Figure 2). All plots were fenced to prevent cattle intrusion.

2.2. Runoff Plots

Within each treatment plot, two Gerlach troughs were installed, one on the upper slope area and the other on the lower slope, to collect the upstream runoff and sediment. To avoid the interference of the vegetative cover, a planialtimetric survey of the plot areas was carried out with RTK surveying equipment, using pins installed in a 0.5 × 0.5 mesh, to determine the exact plot drainage area and slope grade (Figure 3). The drainage areas of the runoff plots of the three farms are presented in

Table 3. Runoff plot drainage areas (m²) in each experimental site/treatment.

Site	Degraded Pasture		Restored Pasture		Natural Savannah
	a	b	a	b	a	b
	--------------------------------- (m2) -----------------------------------------------
Funil farm	17.0	20.7	18.4	21.2	21.0	19.0
Progresso farm	12.7	22.3	15.1	21.6	20.0	22.0
Tarimba farm	28.5	22.7	20.9	20.8	20.0	16.0

.

At the lower end of each runoff plot, a metallic Gerlach trough [24] was installed flush with the soil surface to ensure the proper collection of runoff and sediment. A 1 L Nalgene^® bottle was connected to the trough’s downstream outlet by a plastic hose to collect the runoff samples (Figure 4).

2.3. Rainfall Erosivity

At each experimental site, a WMO-type stainless steel rain gauge (ϕ = 0.15 m) was installed, and precipitation volumes were recorded on a daily basis, concurrently with the runoff and sediment sampling. The rain gauge had a minimum resolution of 0.2 mm, and it was placed at a safe distance from buildings and trees, to avoid rainfall interception.

Rainfall erosivity in each farm was calculated on a monthly and yearly basis, using a Fournier-type equation [32] developed for Central Brazil:

$$\mathrm{R}=12.59{\sum }_{\mathrm{i}=1}^{12}{\left({\displaystyle \frac{{\mathrm{M}}_{\mathrm{i}}^2}{\mathrm{P}}}\right)}^{0.60}$$

(1)

where R (MJ mm ha⁻¹ h⁻¹) = site annual rainfall erosivity; M_i (mm) = monthly precipitation; and P (mm) = annual precipitation.

2.4. Runoff Sampling and Laboratory Analyses

Gerlach troughs were used to collect the runoff and sediment from the plots, since they avoid the flow restriction of the traditional bound plots, being easier to install and to operate [26].

The 1 L runoff bottles were collected whenever rainfall exceeded 10 mm, since this is the threshold for runoff generation [31], which was confirmed empirically in a previous study in a nearby experiment [3]. In the laboratory, the contents of the Gerlach bottles were transferred to beakers, decanted with alum, oven-dried at 100 °C for 48 h (Figure 5), and weighed to determine the runoff volume and sediment concentration.

2.5. Runoff Volume and Soil Loss

Because the Gerlach bottles collected only 1 L samples of the generated runoff and sediment during each significant rainfall event, hydrological calculations were required to estimate the total runoff and the subsequent sediment yield from the plots. The NRCS equation [33] was used to obtain the total runoff in each event:

Q = (P − 0.2 S)²/(P + 0.8 S)

(2)

and

S = (25,400/CN) − 254

(3)

where Q = surface runoff volume (mm), P = precipitation volume (mm), S = abstraction factor (dimensionless), and CN = NRCS curve number (dimensionless). CN values for the experimental sites were derived from observed and calibrated data that were collected at 22.1 m-long bounded runoff plots of a previous study conducted at the Tarimba farm [3].

Soil loss in each significant rainfall event (P > 10 mm) was calculated as the product of sediment concentration obtained from the 1 L samples and the runoff volume estimated from Equations (2) and (3), namely:

SL = F_c C_s Q/100

(4)

where SL (Mg ha⁻¹) = soil loss in each event in the Gerlach plot; F_C (dimensionless) = correction factor for the 22.1 m-long (observed) runoff plots of the Tarimba farm; C_S (g/L) = sediment concentration in the 1 L sampling bottle; Q (L) = Gerlach plot runoff.

The correction factor F_C was obtained by dividing the observed 22.1 m bounded plot soil loss of the Tarimba farm from a previous study [3] by the corresponding Gerlach plot soil loss (Equation (4)), assuming that, due to plot setup and sediment collection conditions, a difference could arise [34]. The same correction factor was used in the other two farms, since no corresponding bounded plots existed. However, due to the proximity of the three experimental areas and their similar climate and landscape characteristics, this assumption was hydrologically acceptable [35].

To allow for a direct comparison of the runoff and soil loss between the three experimental sites [18], surface runoff and soil loss were normalized by the precipitation volume and rainfall erosivity, respectively:

$${\mathrm{Q}}_{\mathrm{n}}={\displaystyle \frac{\mathrm{Q}}{\mathrm{P}}}$$

(5)

$${\mathrm{S}\mathrm{L}}_{\mathrm{n}}={\displaystyle \frac{\mathrm{S}\mathrm{L}}{\mathrm{R}}}$$

(6)

where Q_n = normalized runoff; Q (mm) = runoff volume; P (mm) = rainfall volume; SL_n = normalized soil loss (Mg MJ⁻¹ mm⁻¹ h); SL (Mg ha⁻¹) = soil loss; and R (MJ mm ha⁻¹ h⁻¹).

To assess the relationship between soil loss and runoff of permanent (NS and RP) and non-permanent (DP) soil covers, normalized means of annual runoff (Q_n) and normalized annual soil loss (SL_n) were plotted in a log-log plot, and the corresponding adjusted function and p-value were obtained [18] and compared with soil loss tolerance thresholds.

Runoff and soil loss were monitored during three consecutive hydrologic years (2022/23, 2023/24, 2024/25), coinciding with the rainy season in the Brazilian Cerrado, which extends from November to May [18]. This three year monitoring period enabled the evaluation of restoration progress over time and accounted for intrinsic inter-annual climatic variability.

2.6. Statistical Analyses

The treatment means and their significance were calculated using ANOVA and the Tukey-HSD multiple comparison test [36] at the 0.05 significance level. Previously, the plot data were tested for normality using the Kolmogorov–Smirnov test, with the R-Studio^® package v. 2025.09.0.

3. Results

3.1. Precipitation and Erosivity

The monthly and annual precipitation volumes, with corresponding rainfall erosivities, are presented in

Table 4. Monthly and annual rainfall volumes (P) and rainfall erosivities (R) of the three experimental sites, during the three hydrologic years.

Month/Year	Funil Farm		Progresso Farm		Tarimba Farm
	P (mm)	R (MJ mm ha−1 h−1)	P (mm)	R (MJ mm ha−1 h−1)	P (mm)	R (MJ mm ha−1 h−1)
Aug/22	0.0	0.0	0.0	0.0	0.0	0.0
Sep/22	0.0	0.0	0.0	0.0	0.0	0.0
Oct/22	0.0	0.0	0.0	0.0	0.0	0.0
Nov/22	252.7	155.9	150.0	84.2	242.2	116.8
Dec/22	264.0	164.3	295.5	190.6	214.5	100.8
Jan/23	176.2	100.9	166.5	95.4	225.7	107.2
Feb/23	27.7	10.9	34.5	14.3	406.4	218.0
Mar/23	108.7	56.4	105.0	54.7	126.7	53.5
Apr/23	155.1	86.5	212.2	127.9	244.5	118.1
May/23	0.0	0.0	0.0	0.0	0.0	0.0
Jun/23	0.0	0.0	0.0	0.0	0.0	0.0
Jul/23	0.0	0.0	0.0	0.0	0.0	0.0
Total	984.5	7236.5	963.61	7140.9	1460.0	8994.0
Aug/23	0.0	0.0	0.0	0.0	19.3	77.1
Sep/23	12.6	4.1	49.7	19.5	0.0	0.0
Oct/23	29.2	11.5	27.5	9.6	98.4	43.6
Nov/23	98.3	49.5	84.6	37.1	69.7	28.8
Dec/23	188.5	108.5	125.2	59.6	152.2	73.9
Jan/24	170.6	96.2	228.5	123.0	183.9	92.8
Feb/24	255.3	156.4	303.7	173.3	333.8	190.5
Mar/24	136.3	73.3	165.2	83.2	173.5	86.5
Apr/24	108.2	55.6	208.0	109.8	199.4	102.3
May/24	0.0	0.0	0.0	0.0	0.0	0.0
Jun/24	0.0	0.0	0.0	0.0	0.0	0.0
Jul/24	0.0	0.0	0.0	0.0	0.0	0.0
Total	999.0	6988.5	1192.5	7743.1	1231.9	7871.4
Aug/24	0.0	0.0	0.0	0.0	0.0	0.0
Sep/24	0.0	0.0	0.0	0.0	0.0	0.0
Oct/24	283.6	161.0	156.1	99.3	261.6	124.1
Nov/24	231.0	125.7	276.6	132.2	293.1	142.3
Dec/24	167.8	85.5	481.1	257.6	271.5	129.8
Jan/25	204.8	108.7	206.3	92.8	286.5	138.5
Feb/25	105.2	48.7	122.6	49.5	129.7	53.2
Mar/25	56.8	23.2	181.4	79.4	164.5	70.9
Apr/25	125.7	60.3	126.4	51.4	105.9	41.7
May/25	0.0	0.0	0.0	0.0	0.0	0.0
Jun/25	0.0	0.0	0.0	0.0	27.1	8.1
Jul/25	0.0	0.0	0.0	0.0	0.0	0.0
Total	1174.9	7719.6	1550.5	9181.7	1540.0	8920.8

, for the three hydrologic years.

3.2. Runoff

The calibrated runoff curve numbers (CN) of Equations (2) and (3) of the three sites are presented in Figure 6.

Table 5. Mean normalized runoff (Q/P) across treatments and farms (N = 2). Different letters within the same farm indicate significant differences in Tukey’s HSD test (p < 0.05).

Site	Treatment	Normalized Runoff
Funil farm	DP	0.207 a
	RP	0.116 b
	NS	0.046 c
Progresso farm	DP	0.283 a
	RP	0.133 b
	NS	0.061 b
Tarimba farm	DP	0.242 a
	RP	0.134 ab
	NS	0.055 b

presents the mean annual normalized runoff (mean of two replicates) for the Funil, Progresso, and Tarimba farms for the three consecutive years. At the Funil farm, normalized runoff differed significantly among all treatments, with degraded pasture (DP) showing the highest mean. At the Progresso farm, the restored pasture (RP) and natural savannah (NS) were statistically similar, both exhibiting significantly lower runoff than DP. At the Tarimba Farm, RP did not differ significantly from either DP or NS, indicating an intermediate response.

The box plots of the normalized mean annual runoff are presented in Figure 7. In all cases, NS exhibited the lowest values with minimal variability, while DP consistently showed the highest runoff. RP, in turn, occupied an intermediate position between the other treatments.

3.3. Soil Loss

Table 6. Mean normalized soil loss for the three sites. Different letters within the same farm indicate significant differences in Tukey’s HSD test (p < 0.05).

Site	Treatment	Normalized Soil Loss (Mg hr−1 MJ mm)
Funil farm	DP	0.0060 a
	RP	0.0020 b
	NS	0.0005 b
Progresso farm	DP	6.05 × 10−4 a
	RP	2.80 × 10−4 ab
	NS	6.61 × 10−4 b
Tarimba farm	DP	0.0070 a
	RP	0.0010 b
	NS	0.0004 b

indicates that, overall, soil loss with RP and NS were statistically similar, with both showing lower values than DP. In the Progresso farm, RP showed an intermediate position, not differing significantly from either DP or NS.

The box plots of normalized soil loss are presented in Figure 8 for the three sites. In all cases, NS exhibited the lowest values with minimal variability, while DP consistently showed the highest soil loss. RP, on the other hand, occupied an intermediate position.

3.4. Relationship Between Runoff and Soil Loss and On- and Off-Site Tolerance

The relationship between normalized annual runoff and annual soil loss is presented in Figure 9. In this figure, a good fit (R² = 0.94, p < 0.01) was obtained for a power function, similarly to that of non-karst landscapes of the Brazilian savannah [18].

Additionally, the soil loss of all permanently covered plots (NS and RP) fell below the on-site soil loss tolerance, as opposed to the degraded pasture plots (non-covered plots).

4. Discussion

4.1. Precipitation and Rainfall Erosivity

The results of Table 4 indicate that there was spatial (inter-block) and temporal (inter-annual) variability in rainfall volume and erosivity, which are key drivers of runoff and soil loss, respectively, as reported previously [18]. The relationship between rainfall and erosivity has also been widely documented, with several studies reporting strong correlations between them [37]. Since these sources of variability were minimized through the normalization of runoff and soil loss, respectively (Equations (5) and (6)), they allowed for the unbiased comparison of runoff and soil loss means between treatments and years.

4.2. Runoff

According to Figure 6, CN decreased from degraded pasture (DP) to natural savannah (NS), with restored pasture in between. The differences arise from the diverse soil infiltrability in the three farms, which affect CN [32].

At all sites, runoff from the restored pasture (RP) plots was reduced by at least 50% compared with degraded pasture (DP). The runoff in RP at all experimental sites was situated between DP and NS. This finding indicates that, within a short period, the hydrological behavior of the restored pasture approached that of the natural savannah. Previous research [38] showed that pasture restoration is indeed an interesting conservation practice, since it reduces the runoff, improving water quality in the process.

As shown in Figure 7, normalized runoff was similar in the three farms, ranging from 0.05 to 0.3, despite the pedological variation among them, with the observed differences being associated with the land-use treatments. As indicated in Figure 7, reduced runoff was associated with improved soil cover [18], which enhances water infiltration and mitigates soil compaction [24]. Additionally, the tillage, fertilization, and improved rooting of the restored pastures may have contributed to the increase in infiltration and reduction in runoff [3]. Runoff reduction was more pronounced during the first hydrologic year because of the rapid soil cover in RP, and it equilibrated in the subsequent years, as expected.

4.3. Soil Loss

Soil loss in RP was significantly lower than in DP, with reductions ranging from 55% to 95%, depending on the site analyzed. Comparable reductions were also reported in a nearby restored savannah site established with native grasses and shrubs in bounded 22.1 m plots [39]. Soil loss in the RP plots fell between those of DP and NS in all sites. These findings suggest that pasture restoration can be nearly as effective as, and in some cases comparable to, the natural savannah in reducing soil erosion, only a few years after restoration.

The observed reductions could be attributed to increased ground cover and improved soil structure in RP, which together mitigate the erosion processes. Enhanced soil tilth reduces rill erosion [18] while improved grass cover diminishes sheet erosion by decreasing raindrop impact and runoff [37]. These findings underscore the long-term importance of restoration measures for soil conservation [18] and sustainable land management.

Since the three experimental farms, with their inherent pedological and landscape characteristics, were treated as blocks to avoid unnecessary variability in the treatments [29], a comparison between their soil loss data was irrelevant in the present study. However, soil texture, depth, and natural fertility could have played a role in the restoration process, which was not accounted for.

4.4. Practical Implications of the Study

The significant reductions in runoff and soil loss after 3 years of restoration indicate that degraded pastures, which dominate both karst and non-karst areas of the Brazilian Cerrado, are hydrologically effective, and could be implemented as part of a financial compensation policy based on the hydrologic services provided.

Because of the high correlation between soil loss and runoff (Figure 9), the latter could be used as a proxy to the former, which is more difficult to obtain. Furthermore, soil loss in the permanently covered plots (NS and RP) were below the on-site soil loss tolerance threshold, indicating that the restored pasture is effective for increasing landscape stability and the sustainability of karst areas [3].

Finally, the simple and inexpensive Gerlach troughs were effective in assessing runoff and soil loss, being suitable for developing countries with little research infrastructure, and being able to provide robust data for the establishment of sound soil conservation policies.

4.5. Methodological Limitations

The main methodological limitation of the experiment was the non-integrative runoff sampling of the Gerlach troughs, since only episodic sediment concentration samples were obtained, as opposed to the cumulative sampling of traditional runoff plots [27]. This was solved by the indirect obtention of storm runoff volume, via the NRCS method. However, the high correlation found between the normalized runoff and soil loss (Figure 9) indicates that the indirect method was appropriate.

Also, the correction factor (Fc) of Equation (4), obtained for one of the farms, was assumed to be the same for the two others, despite the local landscape differences. However, considering the proximity of the three experimental sites, and the fact that Fc was very close to 1.0, this evidence reinforces the validity of the assumption [35].

5. Conclusions

Pasture restoration reduced runoff by over 50% and soil loss between 55 and 95%, reaching levels that were comparable to the natural savannah, during a period of three hydrologic years, indicating that pasture restoration and permanent soil cover is an effective hydrologic measure, generating significant hydrologic services both on- and off-site, particularly in vulnerable karst areas. The good fit between normalized runoff and soil loss allows for the utilization of the former as a proxy of the latter. Additionally, simple and inexpensive Gerlach troughs proved to be useful in the assessment of soil loss of natural, degraded, and restored karst areas, and were suitable in the assessment of active restoration services in developing countries.