Distribution of ecosystem service potential in marginal agroecosystems in a mosaic-type landscape under exploratory scenarios

ABSTRACT

Land-use change in agriculturally marginal mosaic-type landscapes involves economic, environmental, and social aspects and can lead to social and economic problems in rural areas. This study focuses on the western part of Vidzeme Uplands, Latvia, which has been heavily impacted by rural depopulation and agricultural land abandonment. The research explores the value of low-intensity agroecosystems in agriculturally marginal landscapes using an ecosystem service approach. Three scenarios were developed: a ‘business as usual’ scenario, a ‘no subsidies’ scenario, and a ‘payments for ecosystem services’ scenario. The ‘business as usual’ scenario shows a pattern of subtle intensification, while the ‘no subsidies’ scenario predicts a significant loss of low-intensity agroecosystems and their services. The ‘payments for ecosystem services’ scenario estimates the safeguarding of essential regulatory ecosystem services, ensuring the potential supply of provisioning services, and maintaining the fragile structure of the marginal mosaic-type landscape.

KEYWORDS:

• Exploratory scenarios
• marginal mosaic-type landscapes
• low-intensity agroecosystems
• ecosystem services
• drivers of land-use change

1. Introduction

In recent decades, ecosystem service (ES) approach has justified its position in land-use and policy planning (Clements et al., 2021; Lam & Conway, 2018; Viglizzo et al., 2012) and has reached the status of ‘actionable knowledge’ (Brunet et al., 2018). The employment of ESs can involve a variety of tools and approaches (Grêt-Regamey et al., 2017), depending on the specific goals and context of the case (Beaumont et al., 2017). The ES approach has been used to explore alternative futures and to develop consensus on eventual land-use policies starting from Millennium Assessment (MEA Ed., 2005) onwards (Brunet et al., 2018; Carpenter et al., 2006; González-García et al., 2020; Kubiszewski et al., 2017; Liu et al., 2017; Nelson et al., 2009). Although the concept of ESs is increasingly being recognized as an important aspect of environmental policy at the EU, to improve its coherence with sectoral EU policies, the integration of the ESs concept in the implementation of existing EU policies at national and regional levels must be furthered (Bouwma et al., 2018).

Among the most important sectors to impact the distribution of ESs is agriculture which provides a wide range of ES (Zhang et al., 2007). Thus, of particular importance is the agricultural land-use change that can heavily affect the future distribution of ESs (Hasan et al., 2020; Metzger et al., 2006). Even though such changes are often related to the dynamics of agricultural market values (Bateman et al., 2013), changes driven by agroecological conditions and locational factors must be considered equally significant in understanding the patterns of ES distribution (Prishchepov et al., 2013; Sroka et al., 2019; Wang et al., 2020). Mosaic-type landscapes – landscapes consisting of highly complex land cover patterns, soil composition, and topography – are particularly vulnerable to agricultural land-use changes (Rendenieks et al., 2017). At a regional level, agriculturally marginal mosaic-type landscapes play a significant role in supplying regulation/maintenance and cultural ESs (Bengtsson et al., 2019). These landscapes, where farming practices are on the verge or even below cost-effectiveness, have developed through centuries of low-intensity farming systems. They are crucial for the conservation of specific European habitats and species (Anderson & Mammides, 2020). Such landscapes are heavily threatened by the abandonment of agricultural land and the expansion of intensive agriculture (Plieninger & Bieling, 2013).

Landscape scenarios can be especially valuable in marginal areas, where the impacts of change can be more severe, and the resources available for adaptation and resilience may be limited. Scenarios assist in creating plausible descriptions of potential future directions of the development of complex systems (Amer et al., 2013). Exploratory scenarios are used to explore potential outcomes or impacts of different management approaches or policy interventions and can examine how different land-use patterns and management practices may influence the distribution of ES potential (Cord et al., 2017) in marginal agroecosystems in mosaic-type landscapes. These scenarios can help decision-makers understand the trade-offs and potential impacts of different land-use choices (Ellis et al., 2019) and can inform the development of more sustainable and resilient land-use strategies (Heinze et al., 2022). Scenarios involving agricultural policy and ESs explore the potential consequences of different policy choices in providing ESs to determine whether new forms of policy instruments are necessary in order to offer insight into how these policies will affect society (van Zanten et al., 2014). These scenarios are used to inform policy decisions and help guide the development of strategies to promote the sustainable use of ESs in the agricultural sector (Qiu et al., 2018).

The presented study aims to operationalize the ES approach through explorative scenario building, thus revealing the multi-functional value of low-intensity agroecosystems in agriculturally marginal landscapes. In the presented paper, we assess agricultural land-use change, study factors behind this change, and use those factors to model future land-use change. We employ Viva Grass tool (Villoslada et al., 2018; Vinogradovs et al., 2020) – a tool aimed at enhancing the maintenance of ESs provided by low-intensity agroecosystems – to construct a plausible distribution of ES potential under alternative futures with varied agricultural land uses. We build two opposed exploratory scenarios and one ‘business as usual’ scenario. Subsequently, we compare and substantiate ES potential distribution under different land-use scenarios and draw conclusions for agricultural land-use planning and policy recommendations.

2. Materials and methods

2.1. Study area

The study area (Figure 1) is located in the western part of Vidzeme Uplands, Latvia, in the landscape of undulated relief and mixed land uses mainly consisting of coniferous (with some deciduous mixture) forests, agricultural lands on poor, eroded soils, as well wetlands and lakes (Table 1). There are no significant settlements in the area; about one-third of population lives in villages (average 500 inhabitants), and the rest is evenly spread throughout the region and accommodated in small farmsteads. After the collapse of the collective farming system, the area underwent a significant change in agricultural production regime, shifting back to small-scale subsistence farming. It was followed by gradual adaptation to a capitalistic market-driven rural economy that eventually led to abandoned farmland (Vinogradovs et al., 2018), population aging, and migration (Table 2).

Figure 1. Study area.

Table 1. Land-cover/land-use distribution in the study area.

Land cover/land use	km^2	%
Residential, industrial	3.43	0.5
Agricultural	273.35	38.2
Forests	421.85	59.0
Wetlands	1.70	0.2
Water bodies	14.60	2.0
Total	714.93

Table 2. Demographic dynamics in the study area.

	Year
	2000	2019	Change
Number of inhabitants	7369	5122	−2247
The proportion of inhabitants aged >65 years	14.3	21.8	34.4%
Population density	10.3	7.2	−30.5%

After joining the EU and CAP, farmland abandonment gradually diminished, and even some reclamation of abandoned land in favourable agricultural conditions and transfer to larger farms and cooperatives appeared (Nikodemus et al., 2010; Vanwambeke et al., 2012).

2.2. Assessment of agricultural land-use change and underlying factors

We carried out assessment of land-use change in agricultural land in two steps. Firstly, a complete land-use map was created for the situation in 2014, combining several available databases (Department of Land Amelioration, M. of A. E. R. of L., 2018; Ministry of Agriculture, L, 2020a, 2020b; State Land Survey, R. of L, 2020) and an extensive field survey (Vinogradovs et al., 2018). Secondly, we updated the land-use map for the situation in 2018 using up-to-date IACS-LPIS data from Rural Support Service. We assessed agricultural land-use change for land-use categories (Table 3). We carried out change detection by using overlying agricultural land-use maps in GIS software and detected changes visited in the field survey in the autumn of 2019. Detected changes were sorted in ‘from – to’ categories matrix, and the area of each change was calculated. For example, if a plot was classified as ‘permanent grassland’ in 2014 and then changed to ‘winter wheat’ in 2018, it indicates an intensification of agricultural land use has occurred. Conversely, if a plot transitioned from ‘cultivated grassland’ in 2014 to ‘permanent grassland’ in 2018, this suggests a disintensification of agricultural land use. The comparison between 2014 and 2018 was utilized in our study to design land-use change patterns within the ‘business as usual’ scenario, taking into account the period of relative stability in agricultural policies.

Table 3. Description of categories of agricultural land uses.

Agricultural land-use category	Description
Permanent grassland (low-input agroecosystem)	Rarely seeded or fertilized, grasses usually grow naturally and contain both natural vegetation and cultivated species. The grass is cut not more than two times a season or used as pasture.
Semi-natural grassland (low-input agroecosystem)	Semi-natural grasslands contain high levels of biodiversity and are used as low-intensity pastures or hay fields (one late cut per season).
Cultivated grassland (intensive use agroecosystem)	Seeded (often a monoculture – Festuca sp., Phleum sp., Dactylis sp.), fertilized, and plowed at least once in 5 years. The grass is cut several (max 3–4) times per year.
Arable land (intensive-use agroecosystem)	Intensively managed farmland used for crop production, plowed at least one time (usually 2–3 times) in the season and fertilized.
Abandoned land	Former agricultural land not used for any agricultural practice and left to natural succession (afforestation) mostly consists of grass vegetation with a thick layer of thatch and sparse cover of shrubs and trees.

The initial set of key drivers (Table 4) that underlay and explain ongoing changes in marginal areas were verified in the previous studies (Vinogradovs et al., 2016, 2018) and successfully applied for analysis in other agriculturally marginal locations (Lieskovský et al., 2015).

Table 4. The initial set of variables prepared for modelling.

Driver	Description	Source
Area	The area of continuous agricultural land-use patch – a field (ha)	GIS calculation
Area/perimeter	Index of compactness of continuous agricultural land-use patch	GIS calculation
Soil texture	Granulometric group (sand, loamy sand, sandy loam, clay, peat)	Digital soil maps
Land quality	Quality of agricultural land described as 100-point max grades	Digital soil maps
Slope	The steepness of the slope (degrees)	Digital elevation model
Drainage	Presence of drainage system (dummy 0/1)	Digital maps of drainage systems
Distance to road	Distance to the closest paved road from the edge of the field (m)	GIS calculation
Distance to farm	Distance to the closest paved farm from the edge of the field (m)	GIS calculation

2.3. Statistical analysis

We performed two separate model sessions to estimate the probability of agricultural intensification and deintensification. At first, all quantitative variables before analysis were centred (subtracted mean) and scaled (divided by standard deviation after centering) due to measurement values and amplitude differences. We used classical main effects (Nelder & Wedderburn, 1972) binary regression analysis with logistic link function and selected the best generalization through evaluation of all the possible independent variable combinations. An assessment was performed in R library ‘MuMIn’ (Bartoń, 2019) by Akaike information criterion following an information-theoretic approach (Burnham & Anderson, 2004). To account for spatial autocorrelation, we performed two autologistic binary regression models, including distance-weighted (calculated in R library ‘spdep’ (Bivand & Wong, 2018)) or distance-bin defined (binwidth automatically selected, R library ‘spatialEco’ Evans, 2018) autoregressive covariate in regression analysis. We selected the best autologistic model by the Akaike information criterion. Log-odds were also expressed as odds ratio values for easier interpretation of results. Explained variance was calculated using the McFadden method of pseudo-r-squared for generalized linear models as implemented in the R library ‘pscl’ (Jackman, 2010).

In the deintensification model set, we found a strong correlation (r = .72) between area and area-perimeter; therefore, we did not allow this variable pair simultaneously in a single model. All the remaining correlation coefficients were below .5 in both model sessions. We found no multicollinearity as variance inflation factor values were below 3 in each of the best models. We conducted all statistical analyses in R 3.6.2 (Team, R. C, 2019).

2.4. Scenario building

To explore plausible futures of agricultural land-use change, we designed three exploratory scenarios – first, ‘business as usual’ scenario was constructed based on probabilities derived from statistical analysis of ongoing agricultural land-use change. Two other scenarios delineate extreme situations: a) there are no subsidies for agriculture (‘no subsidies’ scenario), and b) agriculture is subsidized according to regulatory ESs (‘payments for services’ scenario). The latter outlines the situation where minimizing trade-offs among ES and enhancing hotspots of ES supply are heavily subsidized. Both extreme scenarios were developed in the expert/stakeholder panel discussion based on revealed drivers of agricultural land-use change and land-use sustainability analysis (OECD, 2019) optimization capabilities (Nipers, 2019) reports. The expert panel comprised six scientists specializing in landscape ecology, soil science, agriculture, and ecology, along with three rural advisors and six farmers from the study area. Rural advisors are specialists with territorial responsibilities in study area, providing education and guidance to farmers on various aspects of agriculture and farm development. Among the farmers in the panel, half were engaged in dairy farming, while the other half focused on meat cattle farming. The expert panel discussed and developed an agricultural land-use change matrix, where land-use changes were expressed as changes of land-use categories.

2.5. Assessment of impact on ES potential

We assessed the potential of ESs, trade-offs among services, and ‘hot/coldspot’ analysis using Viva Grass tool (Villoslada et al., 2018). Viva Grass tool uses a matrix approach for the assessment of five provisioning services and eight regulating services relevant to the agroecosystem. ES assessment category is a combination of agricultural land use (Table 3), slope steepness (plain, gentle, steep), and soil fertility classes (low, medium, high, organic) – thus, creating 50 possible agricultural land-use classes (i.e. ‘semi-natural grassland on plain relief, organic soil’). Viva Grass tool identifies a substantial trade-off between ES bundles ‘production’ and ‘habitats’ and synergy among ESs belonging to one bundle (Table 5). Viva Grass tool distinguishes a ‘coldspot of ES potential’ as a spatial unit providing a significant number of ESs at ‘low’ or ‘very low’ values and a hotspot as a spatial unit providing ESs at ‘high’ or ‘very high’ values (Vinogradovs et al., 2020).

Table 5. Ecosystem service supply potential bundles.

Bundle	Ecosystem services
Habitats	Maintaining habitats, global climate control, pollination, herbs for medicine
Production	Reared animals and their outputs, biomass-based energy sources, fodder, cultivated crops
Soils	Chemical conditions of freshwaters, bioremediation, filtration/storage/accumulation

We assessed spatial distributions of ESs under different scenarios using five semi-quantitative classes (from ‘very low’ to ‘very high’) in three ES supply bundles.

3. Results

3.1. Agricultural land-use change

Between 2014 and 2018, more than 10% of agricultural land has undergone agricultural land-use change (Table 6). For further analysis, the two most significant vectors were distinguished – intensification (containing land-use change categories ‘from permanent grassland to cultivated grassland’ and ‘from permanent grassland to arable land’) and disintensification (containing land-use change category ‘from cultivated grasslands to permanent grasslands’). The initial decision to include the land-use change category ‘from arable land to cultivated grassland’ into the disintensification vector and land-use change category ‘from cultivated grassland to arable land’ into the intensification vector was abandoned after the field survey as de facto land uses in both change categories were the same – mixture of grasses and legumes cultivated for fodder and designation to either class was done by a proportion of grassland species.

Table 6. Land-use change matrix; changes for further analysis are marked in bold.

	To cultivated grassland	To permanent grassland	To semi-natural grassland	To arable land	To abandoned land	Total
From cultivated grassland		1491.3	0.8	180	2.9	1675
From permanent grassland	300.1		56.9	480.3	12.8	850.1
From semi-natural grassland	0	21.3		0	0.2	21.5
From arable land	160.2	53.2	0		0.4	213.8
From abandonedland	14.3	49.8	2.9	8.1		75.1
Total	474.6	1615.6	60.6	668.4	16.3	2835.5

3.2. Results of statistical analysis

Intensification. Within the intensification model session, 128 model structures were compared. The best model AIC was 277.78 values lower than the null model. The addition of inverse-distance weighted auto-covariate further decreased AIC by 41.66 values, suggesting better generalization. Binned auto-covariate with binwidth of 1740 m further reduced AIC by 55.01 from a classical binary logistic model and by 13.35 from the inverse-distance weighted model. The variance explained in models with binned spatial autoregression structure is 20.17% (Table 7). The results of the autologistic binary model show that increasing land quality and decreasing area/perimeter ratio of the field form increase the probability of intensification. Decreasing the distance to paved roads and distance to the farm is essential from the intensification model, even though showing larger p values.

Table 7. Results of autologistic binary model for agricultural land-use change ‘intensification’.

Variable	Coef	Std.Error	Wald	Prob	OR	p value
Intercept	−3.9138	0.2590	−15.1096	1.3988e-51	0.0200	<.0001
Area	−7.0283	0.8374	−8.3922	4.7712e-17	0.0009	<.0001
Area/perimeter	1.1810	0.1263	9.3501	8.7486e-21	3.2577	<.0001
Distance to road	−0.1644	0.0769	−2.1374	3.2563e-02	0.8484	0.0326
Distance to farm	−0.2346	0.0897	−2.6133	8.9649e-03	0.7909	0.009
Slope (plain)	0.3088	0.1963	1.572823	1.1575e-01	1.3618	0.1158
Land quality	0.4906	0.0743	6.600109	4.1085e-11	1.6333	<.0001

Disintensification. Within the disintensification model session, 96 model structures were compared. The best model AIC was 84.59 values lower than the null model. The addition of inverse-distance weighted auto-covariate further decreased AIC by 65.239, suggesting better generalization. Binned auto-covariate with binwidth of 2082.662 m further reduced AIC by 121.784 from a classical binary logistic model and by 56.544 from the inverse-distance weighted model. Variance explained in models with binned spatial autoregression structure is 21.74% (Table 8). The results of the autologistic binary model show that decreasing land quality and the area of the field increases the probability of disintensification.

Table 8. Results of autologistic binary model for agricultural land-use change ‘disintensification’.

Variable	Coef	StdError	Wald	Prob	OR	p value
Intercept	−2.5043	0.2312	−10.8279	2.5385e-27	0.0817	<.0001
Area	0.6254	0.0998	6.26582	3.7085e-10	1.8691	<.0001
Slope (plain)	0.2011	0.2283	0.8808	3.7842e-01	1.2228	0.3784
Land quality	−0.5077	0.1078	−4.7093	2.4851e-06	0.6018	<0.0001

3.3. land-use change under scenarios

Constructed agricultural land-use change matrix and assessed agricultural land-use change under scenarios are presented in Annex 1 and Table 9, respectively. Land-use change under the ‘business as usual’ scenario was based on autologistic regression models. The probability of land use was assessed only for intensification and disintensification of agricultural land use. The assessed abandonment was not representative enough for building regression. The total amount of change was kept at the same level as the assessed land-use change between 2014 and 2018. Land-use change in ‘no subsidies’ and ‘payments for services’ was based on the change in the agricultural land-use category discussed at the expert panel. The ‘no subsidies’ scenario included the correction of the land-use change category for fields situated in the distance from farms (>10 km) and/or paved roads (3 km) and small size (<1 ha). Predicted amounts of agricultural land-use change under aforementioned scenarios are outlined Table 9 by using categories of the intensity of farmland management where under ‘intensive use,’ there are assembled the arable land and cultivated grasslands and under ‘low-intensity’ – permanent and natural grasslands.

Table 9. Farmland under land-use change in scenarios.

	2018	‘Business as usual’		‘No subsidies’		‘Payments for services’
	ha	ha	%	ha	%	ha	%
Intensive use (arable and cultivated grassland)	3145.12	+780.4	24%	+4460.61	41%	−1342.96	43%
Low-intensity (permanent and semi-natural grasslands)	12,328.3	+1615.6	13%	−5675.77	46%	+23,044.92	86%
Abandoned	8914.46	+50	0.6%	+14,251.5	60%	−8914.46	−100%

3.4. Distribution of potential of ESs under scenarios

We uploaded the agricultural land-use change matrix in Viva Grass tool to assess ES potential, trade-offs, and cold/hotspots. From the outcomes of the assessment, we produced maps and distribution charts.

Distribution of potential of ESs under the ‘business as usual’ scenario in the ‘habitats’ bundle (Figure 2) compared to a situation in 2018 illustrates a rise in the ‘moderate-value’ group and a decrease in the ‘high-value’ group. Under the ‘no subsidies’ scenario, the distribution of ES shows a strong increase in the ‘low’ value group, the disappearance of the ‘very high’ value group, and a slight increase in the ‘high’ value group. Under the ‘payments for services’ scenario, intense increase in the ‘very high’ value group and a decrease in proportion in all other value groups are apparent.

Figure 2. Distribution of potential of ecosystem services in the ‘habitats’ bundle.

Regarding the distribution of potential of ESs under the ‘business as usual’ scenario in the ‘production’ bundle (Figure 3) compared to 2018, there is a noticeable rise in the ‘moderate’ value group and a decrease in ‘low’ value. Conversely, the ‘no subsidies’ scenario demonstrates an increase in the ‘very low’ value group, a disappearance of the ‘very high’ value group, and a slight decline in the ‘high’ value group. On the other hand, the ‘payments for services’ scenario displays an increase in the ‘moderate’ value group, a decrease in the ‘low’ value group, and disappearance of the ‘very low’ value group.

Figure 3. Distribution of potential of ecosystem services in the ‘production’ bundle.

Analyzing the distribution of potential of ESs under the ‘business as usual’ scenario in the ‘soils’ bundle (Figure 4) compared to a 2018, there is an elevation in ‘low’ and ‘moderate’ value groups and a decrease in ‘high’ and ‘very high’ values. In contrast, under the ‘no subsidies’ scenario, the distribution of ES potential displays an increase in ‘low’ and ‘moderate’ value groups, the disappearance of the ‘very low’ value group, and a decrease in the ‘high’ value group. Additionaly, the ‘payments for services’ scenario reveals an increase in the ‘moderate’ value group, a reduction in the ‘high’ value group, and the elimination of the ‘very low’ and ‘very high’ value groups.

Figure 4. Distribution of potential of ecosystem services in the ‘soils’ bundle.

The distribution of trade-offs among ESs remains largely unchanged under the ‘business as usual’ scenario when compared to the situation in 2018 (Figure 5). However, the ‘no subsidies’ scenario exhibits a substantial increase in trade-offs in the benefit of the ‘production’ bundle, accompanied by a significant decrease in trade-offs in the benefit of the ‘habitats’. In stark contrast, the ‘payments for services’ scenario reveals a completely opposite pattern of trade-offs.

Figure 5. Distribution of trade-offs among ecosystem services.

The distribution of cold/hotspots of ES potential exhibits minimal changes under the ‘business as usual’ scenario when compared to the situation in 2018 (Figure 6). However, the ‘No subsidies’ scenario demonstrates the elimination of intense cold and intense hotspots, accompanied by an increase in the prevalence of moderate hotspots. Conversely, the ‘payments for services’ scenario reveals the disappearance of intense coldspots and a substantial increase in the proportion of intense hotspots of ES potential.

Figure 6. Distribution of cold/hotspots among ecosystem services.

4. Discussion

Farmland abandonment, which over the past few decades was the dominating land-use change in marginal landscapes of post-soviet rural areas (Bakker et al., 2011; Lieskovský et al., 2015; Nikodemus et al., 2005; Plieninger et al., 2016; Prishchepov et al., 2013; Ruskule et al., 2012; Zariņa, 2013), now occurs rarely and in insignificant areas. The field survey and comparison of land-use and IACS-LPIS maps clarified that modifications in this section of the land-use change matrix could be referred to the more precise delineation of field boundaries (in the case of ‘to abandoned land’) or returning to more intensive agricultural management of the farmland in favourable agroecological conditions. This tendency has been documented elsewhere in similar situations (Meyfroidt et al., 2016; Smaliychuk et al., 2016). Agroecological conditions are the drivers behind ongoing relatively small-scale land-use changes in both (intensification and disintensification) directions acknowledged in other studies (Kolecka et al., 2017; Pazúr et al., 2020).

The results of the assessed agricultural land-use change (used for modelling the ‘business as usual’ scenario) show that area under intensification of land use, i.e. elimination of permanent grasslands by turning them into arable or cultivated grasslands, is twice as less as the area under disintensification. That could lead to the assumption that the overall landscape change pattern is heading towards the direction of less intense anthropogenic pressure. That change is reported to positively impact the supply of various regulatory ESs (Palm et al., 2014). Our modelling results for the distribution of ES values within ES bundles reveal a distinct trend. This trend indicates a subtle yet noticeable intensification, signifying increased human pressure on the studied landscape. From the results of the autologistic binary model for intensification, it could be assumed that intensification can be more explicit in close proximity to farms, especially intensive dairy farms, where cows are kept in confinement and fed a highly formulated diet. However, this needs further research as even subtle changes in farming methods can have an impact on ecosystem and landscape functions (Houet et al., 2010).

There is quite a clear consensus that subsidies can be a useful tool for supporting low-intensity farming and helping farmers to overcome challenges and remain competitive. However, the need for subsidies in low-intensity farming can vary depending on the specific context and goals of the farming system (Cullen et al., 2018). It is important to note that low-intensity farmland can provide a range of ESs that have value beyond their direct contribution to farm income (Dardonville et al., 2022) and can benefit both – agricultural production and biodiversity (Erisman et al., 2016).

‘No subsidies’ – a scenario describing a rather dystopian situation – discloses quite a pessimistic situation in the distribution of ES. The distribution of ES in the ‘habitats’ bundle – losing the ‘very high’ value group and heavily increasing the ‘very low’ value group – shows an increase in the ‘high’ value group. This is accomplished by intensifying agricultural land use in favourable agroecological conditions and abandoning farmland in unfavourable agroecological conditions. The most alarming change under this scenario is the elimination of ‘intense hotspots’ – places that had high potential in both ‘habitats’ and ‘production’ bundles – described as possible areas for conservation or restoration with high service diversity and capability of supplying ES (Schröter & Remme, 2016; Liu et al., 2017; Gilby et al. 2020). Both ‘intense hotspots’ and ‘strong trade-offs in benefit of habitats’ were provided by semi-natural grasslands, which under the ‘No subsidies’ scenario will disappear as non-productive agricultural land use (Dengler et al., 2020).

The scenario ‘payments for services’ displays an overall increase in the potential of regulatory services – both in ‘habitats’ and ‘soils’ bundles, thus demonstrating a pathway to reach goals related to halting biodiversity loss (Wunder & Wertz-Kanounnikoff, 2009; Wunder, 2013; Chen et al. 2020) and mitigating soil degradation (Mander et al., 1999). In the results of an assessment of this specific scenario, we want to highlight the distribution of values of ES in the ‘production’ bundle (Figure 3). Complete elimination of ES provided at the ‘very low’ value is reached through the withdrawal of intensive agriculture practices from sites in unfavourable agroecological condition. These sites include low-quality lands and steep slopes. Instead of abandoning them as in the ‘no subsidies’ scenario, these sites are managed as low-intensity agroecosystems. Usual trade-offs of such agroecosystems, i.e. inverse relationships between productivity and stability and productivity and sustainability (Viglizzo et al., 1998), are overcome by payments for multiple benefits in the form of a wide range of ES these systems potentially provide (Villoslada Peciña et al., 2019). The overall growth of production values is based on the return to agricultural use of abandoned farmland. This approach not only supports the viability of production by extending lands designated for agricultural practices (Corbelle-Rico et al. 2022) but also upholds the mosaic of the landscape (Ruskule et al., 2013). By doing so, it increases the potential provision of landscape goods and services relevant to these types of landscapes (Smaliychuk et al., 2016). Nevertheless, even such a forward-looking scenario can possibly create certain trade-offs (Boke Olén et al., 2021) that should be addressed through informed agricultural policy and spatially and temporally explicit agri-environment schemes (Mewes et al., 2015).

5. Conclusions

The study reveals a variety of multi-functional values of low-intensity agroecosystems by studying the possible futures of ES supply potential under different scenarios. Overall, it emphasizes the benefits of low-input land use in marginal mosaic-type landscapes, i.e. ES in ‘habitats’ bundle – pollination, maintenance of habitats, climate control, herbs for medicine; ES in ‘production’ bundle – reared animals, fodder, biomass for energy; and ES in ‘soils’ bundle – weathering, accumulation, bioremediation. Comparing outcomes of two antagonistic and one ‘business as usual’ scenarios creates cogent advice for elaborating landscape-specific agri-environmental schemes and agricultural land-use planning. The ‘business as usual’ scenario is elaborated by studying agricultural land-use change and statistical analysis of drivers of identified undisclosed intensification, which becomes evident only by studying the distribution of ES and not the proportions of land-use types. The spatial allocation of possible intensification calls for further detailed study as it could lead to changes or even loss of vital ecosystem functions and services. The ‘no subsidies’ scenario points out the importance of subsidies for farmers, as, without such contributions, the loss of the economically less favourable low-intensity agroecosystems, i.e. semi-natural grasslands (ecosystems providing hotspots of ES supply potential), could be inevitable. The ‘payments for ES’ scenario calls attention to an arrangement of distribution of ES values – designating that low-intensity agroecosystems bear the potential to not only provide wide range of regulatory ES and biodiversity but are also capable of productive outcomes.

Ultimately, the analysis of explanatory scenarios reveals that for agricultural land-use planning and policy recommendations in marginal mosaic-type landscapes, choosing the most sustainable or beneficial scenario is not necessarily the case. The range of multi-functional values that disclose under these diverse scenarios in the context of social-ecological sustainability suggest developing a management plan that incorporates elements from multiple scenarios.

Disclosure statement

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

Additional information

Funding

This study was supported by Funding for Science of the University of Latvia within the ‘Climate change and sustainable use of natural resources’ (AAp2016/B041//Zd2016/AZ03) program.

Annex 1. Agricultural land-use change matrix under exploratory scenarios

Table

Agricultural land-use category	‘Business as usual’ scenario	Land-use category under ‘no subsidies’ scenario	‘Corrections for no subsidies’	Land-use category under ‘payments for services’
1. Cultivated grassland on plain relief, low land quality	Land-use change was based on a probability derived from autologistic binary regression	41	41	11
2. Cultivated grassland on plain relief, medium soil fertility		32	42	12
3. Cultivated grassland on plain relief, high land quality		33	43	3
4. Cultivated grassland on plain relief, organic soils		44	44	14
5. Cultivated grassland on gentle slope in low land quality		45	45	15
6. Cultivated grassland on gentle slope, medium soil fertility		36	46	16
7. Cultivated grassland on gentle slope, high land quality		37	47	7
8. Cultivated grassland on gentle slope, organic soil		48	48	18
9. Cultivated grassland on steep slope low land quality		49	49	19
10. Cultivated grassland on steep slope, medium soil fertility		50	50	20
11. Permanent grassland on plain relief, low land quality		41	41	21
12. Permanent grassland on plain relief, medium soil fertility		12	42	22
13. Permanent grassland on plain relief, high land quality		33	43	23
14. Permanent grassland on plain relief, organic soils		44	44	24
15. Permanent grassland on gentle slope, low land quality		45	45	25
16. Permanent grassland on gentle slope, medium soil fertility		16	46	26
17. Permanent grassland on gentle slope, high land quality		37	47	27
18. Permanent grassland on gentle slope, organic soils		48	48	28
19. Permanent grassland on steep slope, low land quality		49	49	29
20. Permanent grassland on steep slope, medium soil fertility		50	50	30
21. Semi-natural grassland on plain relief, low land quality		41	41	21
22. Semi-natural grassland on plain relief, medium soil fertility		32	42	22
23. Semi-natural grassland on plain relief, high land quality		33	43	23
24. Semi-natural grassland on plain relief, organic soils		44	44	24
25. Semi-natural grassland on gentle slope, low land quality		45	45	25
26. Semi-natural grassland on gentle slope, medium soil fertility		36	46	26
27. Semi-natural grassland on gentle slope, high soil fertility		37	47	27
28. Semi-natural grassland on gentle slope, organic soils		48	48	28
29. Semi-natural grassland on steep slope, low land quality		49	49	29
30. Semi-natural grassland on steep slope, medium land quality		50	50	30
31. Arable land on plain relief, low land quality		41	41	11
32. Arable land on plain relief, medium soil fertility		32	42	32
33. Arable land on plain relief, high land quality		33	43	33
34. Arable land on plain relief, organic soils		44	44	14
35. Arable land on gentle slope in low land quality		45	45	15
36. Arable land on gentle slope, medium soil fertility		36	46	6
37. Arable land on gentle slope, high land quality		37	47	7
38. Arable land on gentle slope, organic soil		48	48	18
39. Arable land on steep slope low land quality		49	49	19
40. Arable land on steep slope, medium soil fertility		50	50	20
41. Abandoned land (with shrub) on plain relief, low land quality		41	41	11
42. Abandoned land (with shrub) on plain relief, medium soil fertility		42	42	12
43. Abandoned land (with shrub) on plain relief, high land quality		43	43	13
44. Abandoned land (with shrub) on plain relief, organic soils		44	44	14
45. Abandoned land (with shrub) on gentle slope in low land quality		45	45	15
46. Abandoned land (with shrub) on gentle slope, medium soil fertility		46	46	16
47. Abandoned land (with shrub) on gentle slope, high land quality		47	47	17
48. Abandoned land (with shrub) on gentle slope, organic soil		48	48	18
49. Abandoned land (with shrub) on steep slope low land quality		49	49	19
50. Abandoned land (with shrub) on steep slope, medium soil fertility		50	50	20