Driving Forces of Agricultural Land Abandonment: A Lithuanian Case

Abstract

The abandonment of agricultural land is now considered one of the primary land use changes driven by complex interactions between social, economic, and environmental factors. To understand and manage this process, a holistic approach that integrates multidimensional methodologies and interactions is essential. This study examines the key driving factors behind agricultural land abandonment in Lithuania using two methodological approaches. First, seventeen highly qualified land management experts were surveyed, and their insights were analysed using in-depth qualitative interviews, focusing on agricultural land abandonment and its underlying factors. Second, the development of agricultural land abandonment in a representative Lithuanian municipality was modelled using Markov chain models, incorporating freely available geographic data as factors influencing land use transformation. Actual areas of abandoned agricultural land were mapped using orthophotos from 2012, 2018, and 2021, for both model development and validation. The importance of predictors in the model was then assessed in relation to their significance as drivers of agricultural land abandonment. The findings indicate that natural factors, such as the proximity of forests and topographical constraints, play a significant role in explaining land abandonment processes. Additionally, agricultural land abandonment is influenced by social, economic, and legal factors, including land ownership structures, migration, and infrastructure accessibility. The importance of soil quality, productivity, and the presence of nearby arable land was found to vary depending on data accuracy and local environmental conditions, highlighting the complexity of agricultural land use patterns. The chosen mixed-method approach, combining qualitative surveys with numerical spatial modelling, demonstrates potential for identifying critical land use areas and providing insights to improve land management policies and decision making.

1. Introduction

Land-use patterns and their changes have a significant impact on the sustainable development of the global environment [1,2]. There are many reasons why land cover and its use by humans is important. First, the properties of land cover and their changes over time have an impact on the environment and influence the functioning and value of ecosystems. On the other hand, land cover or, more appropriately in this context, land use, greatly defines the resources and benefits available for human use. Globally, the abandonment of agricultural land (ALA) is currently one of the primary land-use changes [3], driven by social, economic, and environmental factors [4,5]. Rapid population growth, extensive resource use, and growing environmental concerns have made the monitoring of ALA a key topic on the international research agenda [6,7,8]. Over the past 40 years, there has been a steady increase in the number of articles published regarding ALA [9]. In various regions of the world, the abandonment of agricultural land occurs due to biophysical factors and human activities [10,11]. Although the issues related to ALA are global and cause significant challenges in many countries, changes depend on local conditions influenced by numerous factors, such as policy, governance, economics, culture, human behaviour, and the environment [12,13,14,15,16,17]. Therefore, it is crucial to understand the processes shaping land abandonment at various scales—from regional to global. Such knowledge is essential for developing policies and management plans necessary to understand and improve land-use change trends [10,18,19] and to draft sustainable development strategies [20].

While some of the forces that influence the risk and rate of abandonment can be intuitively understood, research into the drivers of ALA has noticeably intensified since the 1980s. Several approaches exist for classifying the driving forces behind ALA [21]. Based on the extent of abandonment effects, these forces can be categorized as fundamental (e.g., the marginalization of cultivated land), direct (e.g., labour migration), and main (e.g., changes in socioeconomic conditions). When classified by their origin, driving forces can be external, including those acting as catalysts in ALA processes, or internal, such as agroecological, socioeconomic, or agricultural land characteristics that determine the dynamics, magnitude, scope, and location of ALA [5]. These factors often interact; for example, the lack of profit—an internal driver of agricultural land abandonment—is largely determined by various external factors that render land use unprofitable [22]. According to their attributes, driving forces may be grouped into physical factors, relating to geography, biology, physics, ecology, and environmental limitations on agricultural land use; socioeconomic factors, such as markets, population, technology, and mobility; and management factors, related to overall land management systems [23]. However, these classification approaches often overlap, leading to conceptual and interpretative ambiguities.

The process of ALA is usually the result of complex interactions between environmental and human factors. So, to understand and manage it effectively, a holistic approach that integrates social, environmental, technological, economic, and political dimensions is essential [24,25]. Investigating the drivers behind ALA requires methods rooted in the study of socioecological systems. Numerous studies have examined the factors contributing to ALA, varying in methodologies, scales, and problems addressed. A common approach is the use of modelling techniques that account for location and scale. Geographic information systems and remote sensing are frequently employed to map land-use changes and analyse spatial and temporal patterns of development [21,26,27,28]. For example, abandoned agricultural land is mapped using very high-resolution satellite imagery or laser scanning data [29,30,31]. Land-use process simulations are then applied to understand landscape transformations over time and to predict future trends, using mapped factors as model inputs [32,33,34,35,36]. The primary determinants of ALA often vary depending on the scale of analysis. A shared challenge across these factors is their inherent difficulty in being quantified and integrated into numeric models of ALA processes. To address these complexities, a comprehensive understanding of ALA requires a further combination of methods such as quantitative surveys, which systematically collect data from a large number of respondents to assess the prevalence and significance of various factors, and qualitative surveys, which are used to provide detailed, context-specific insights into the motivations, perceptions, and experiences of those involved [37]. Then, desktop research, which involves the analysis of secondary data from reports, academic articles, databases, and government publications, offers a broad and systematic understanding of ALA [38]. Finally, examination of legal frameworks, land-use regulations, and agricultural policies, such as the EU Common Agricultural Policy and national legislation, is crucial for understanding how governance and institutional structures influence land abandonment [7,39,40]. In summary, the study of ALA drivers involves synthesizing insights from desktop research and legal reviews, combined with stakeholder assessments and empirical investigations through spatial analysis and modelling. This integrated approach usually ensures a robust understanding of the diverse and interconnected factors driving agricultural land abandonment.

In Lithuania, the Law on Land Tax [41] states that “Abandoned agricultural land means areas of agricultural land located on a land plot or part thereof overgrown with woody plants (excluding plantations), determined by remote mapping methods”. That is, the legislation dictates a focus on woody vegetation within agricultural lands, as identified through specific mapping methods. The abandonment of agricultural land (ALA) in the country could be influenced by a complex interplay of socioeconomic, environmental, and policy-related factors. One of the primary drivers of ALA is socioeconomic change, particularly following the transition from a centrally planned economy to a market-driven type after the dissolution of the Soviet Union. This transition led to significant market disruptions, including price liberalization of agricultural inputs and outputs, which adversely affected the viability of farming in many regions [42,43]. The economic restructuring resulted in limited access to capital and increased competition, making it difficult for many farmers to sustain their operations [44]. Furthermore, rural depopulation, driven by migration to urban areas in search of better employment opportunities, has led to a reduced agricultural labour force, exacerbating the abandonment of farmland [29,44]. Environmental conditions also significantly influence ALA. In Lithuania, agricultural land that is less suitable for farming, often characterized by poor soil quality or unfavourable climatic conditions, is more prone to abandonment [45]. The increase in ALA is also promoted by factors unfavourable for agricultural activities, i.e., terrain, low productivity score, swampiness, and inoperability of land reclamation systems. The ecological consequences of ALA are significant, as abandoned lands can lead to changes in biodiversity, soil stability, and carbon sequestration [46]. Despite the growing body of research, several knowledge gaps still remain in regards to the understanding of ALA in the country. While there is a consensus concerning the socioeconomic and environmental drivers of abandonment, there is limited empirical data on the long-term impacts of ALA on local ecosystems and communities [29,43]. Furthermore, the integration of geographic information systems (GIS) and remote sensing technologies for monitoring ALA has been explored, yet there is a need for more comprehensive studies that combine GIS and remote sensing data with ground-level surveys to gain a holistic understanding of the phenomenon [46,47,48,49].

Thus, to understand and effectively manage the process of ALA, it is essential to integrate diverse information sources and research methodologies. However, studies on ALA are often challenged by several factors, including the ambiguous definition of ALA, differing interpretations of its impacts, a lack of transdisciplinary and interdisciplinary approaches, and the influence of socioeconomic and political contexts, as well as research traditions. These challenges become particularly pronounced when analyses focus on specific regions or countries. This paper aims to examine the key driving factors of agricultural land abandonment in Lithuania, while simultaneously addressing methodological challenges associated with the use of qualitative and quantitative approaches in land use research. Specifically, we integrate insights from in-depth qualitative interviews with highly qualified land management experts and employ spatial modelling to analyse ALA processes, using selected drivers as predictors. The discussion further extends to an exploration of policy implications for land use management.

2. Materials and Methods

2.1. Study Area

The study was conducted in Lithuania (Figure 1a). The total land area of Lithuania is 65,200 km². Geographically, even though Lithuania is situated in central Europe, with central coordinates of 55°10′ N, 23°39′ E, it has strong historical links with Eastern Europe. Land use development in Lithuania in recent decades strongly depended on the radical societal transformations after Lithuania broke away from the Soviet Union in 1990 and later joined the European Union in 2004 [50]. Lithuania lies on the Eastern European Plain, with characteristic lowlands and hills (the highest point in the country is only 293 m above sea level). More than 50% of its land area is used for agricultural purposes. The terrain features numerous lakes and wetlands, and a mixed forest zone covers over 33% of the country. Currently, Lithuania is dominated by rural landscapes, covering approximately 75% of its territory.

To enable a more in-depth analysis of the driving forces behind ALA, the Jonava Municipality, located in the central part of Lithuania, was selected as a case study (Figure 1b). Located in Kaunas County in the northern part of the Neris River region, the total land area of Jonava Municipality is 944 km². The southern and eastern areas are characterized by gently rolling moraine hills, while the central and northern regions are primarily flat plains, ideal for agriculture. Over 45% of the land is dedicated to agricultural use. The region is also home to several smaller lakes and streams, with mixed forests covering more than 41% of the district’s territory. Larger forested areas are found in the southern and northern parts of the district, comprising natural parks and recreational zones. Other land types, including tree and shrub plantations, wetlands, degraded land, and unused areas, account for more than 5% of the total area, while agriculture abandoned land occupies 0.56% of the district’s territory.

2.2. Input Data

Two methodological approaches were used in the study, i.e., firstly, a qualitative study using in-depth interviews with experts, and secondly, spatially based land use development modelling methods.

The qualitative part of our study involved in-depth interviews with ALA experts. Informants were selected based on a priori knowledge of the political significance of specific cases, initially focusing on experts who played key roles in particular sociopolitical contexts and were likely to provide the most substantial insights. Subsequently, the snowball sampling technique was employed to expand the participant pool, with interviewees recommending additional individuals whose perspectives were relevant to the research objectives and methodological approach. In total, 17 informants were interviewed, representing the full spectrum of stakeholders involved in ALA processes in Lithuania.

Table 1. Overview of the informants, involved in management of agricultural land abandonment processes in Lithuania.

Id	Type of Institution	Professional Experience	Functions	Education	Participation in Research
1	University	Over 20 years	Education and research in the field	Land management	Yes
2	University	Over 20 years	Education and research in the field	Environmental engineering	Yes
3	University	Over 20 years	Education and research in the field	Land management	Yes
4	Law firm	From 5 to 10 years	Legal services in the field	Land management	Yes
5	National land service (state authority)	From 15 to 20 years	Supervising implementation of land management policies	Land management	No
6	National land service (state authority)	Over 20 years	Leading the implementation of land management policies	Land management	Yes
7	National land service (state authority)	From 15 to 20 years	Supervising implementation of land management policies	Land management	No
8	State land fund (state company)	Less than 5 years	Implementation of land management activities	Land management	No
9	Ministry of Environment (state authority)	Over 20 years	Building land and landscape management policies	Forestry	No
10	Municipality administration	Over 20 years	Implementation of land management	Land management	No
11	State service of protected areas (state authority)	From 5 to 10 years	Supervising implementation of landscape protection policies	Landscape management	Yes
12	National land service (state authority)	Over 20 years	Supervising implementation of land management policies	Land management	No
13	University	Over 20 years	Education and research in the field	Environmental engineering	Yes
14	State land fund (state company)	Over 20 years	Implementation of land management activities	Land management	No
15	University	Over 20 years	Education and research in the field	Agronomy	Yes
16	Retired expert	Over 20 years	Former expert in landscape and forest management	Forestry	Yes
17	Ministry of Environment (state authority)	Over 20 years	Building land and landscape management policies	Land and landscape management	Yes

provides an overview of the informants who participated in this study.

Questionnaires with preliminary questions for the informants were developed. The questionnaires were designed to first determine the informants’ areas of work and interest, followed by a discussion of their professional relevance to ALA. Subsequently, the informants were asked about the decisions they make in their work, with an emphasis on the processes of ALA and the drivers behind them. At the end of the survey, informants were invited to suggest other experts whose perspectives might be valuable.

The survey was conducted by two researchers, with one engaged directly with the informant, while the other took notes and recorded the interview, provided that the informant consented to the audio recording. Interviews lasted between 30 and 140 min, with an average duration of approximately one hour. Following each interview, the audio recording was reviewed, and a transcript was produced.

The survey data, including the summarized responses, were organized in MS Excel tables, categorizing responses by informant and question. Each interview was analysed using qualitative methods, including identifying recurring themes and contradictions, refining key ideas relevant to the study, and summarizing the findings. Direct quotes were used to maintain authenticity and accurately represent the informants’ perspectives. To enhance clarity and conciseness, some quotes were shortened by omitting less relevant content (indicated by “…”) or by inserting clarifying phrases in parentheses, based on the broader context of the question.

Informants were anonymized and assigned ID numbers based on the date and time of their interview. However, their identities are known to the researchers, and additional details about them can be obtained from the authors of this paper. All citations in this paper were translated from Lithuanian into English.

The second part of the analysis of ALA drivers involved a spatial modelling-based exercise using openly available geographic information to represent specific drivers as input in ALA models at the case-study level.

First, a detailed mapping of abandoned agricultural land was conducted. Abandoned agricultural land was identified across Jonava Municipality using a 25 × 25 m grid, which aligns with the network of virtual observation points currently being developed by the Lithuanian National Forest Inventory (NFI) for greenhouse gas accounting in the LULUCF sector (Figure 1b). The ALA status was determined manually by analysing a digital raster orthophoto map of the Republic of Lithuania (ORT10LT) at a scale of 1:10,000 (https://www.geoportal.lt/arcgis/rest/services/NZT/ORT10LT_2012_2013/MapServer, accessed on 12 April 2025), following the ALA definition of Law on Land Tax [41], with areas of agricultural land overgrown with woody plants classified as abandoned. The analysis was based on aerial images acquired in 2012, 2018, and 2021. The methodological framework for the identification of ALA is summarized in Figure 2. The process began with the identification of virtual observation points overgrown with woody vegetation. Subsequently, areas outside the scope of the current study—such as forests, built-up areas, infrastructure, water bodies, orchards, and plantations—were excluded. The remaining areas, primarily agricultural lands, were classified into three categories based on the location of the virtual observation point relative to a contiguous polygon of at least 0.1 ha overgrown with woody vegetation, i.e., “abandoned”, “not abandoned”, and “other”. The “other” category typically included areas covered by tree and brush vegetation located outside, but adjacent to, forest land—such as along roads or in similar settings—and usually pertaining to the precision of forest edge delineation. Changes in ALA status were only identified for subsequent dates, specifically 2018 and 2021. Interpretation was carried out by an experienced land management expert following an aerial photo interpretation methodology adapted from forest inventories [51].

Free geographic data available from the Spatial Information Portal of Lithuania (www.geoportal.lt) were used to analyse the factors influencing land use changes. The data were preprocessed to meet the requirements of further modelling and to address the research question regarding ALA drivers in this study.

The following factors influencing land use transformation, or ALA drivers, were considered:

• Land ownership—categorized as private land, state-owned land, and unidentified land (most likely state-owned).
• Steepness of terrain—measured as slope in degrees, based on digital terrain models developed using elevation data from the GDR50LT dataset (https://www.geoportal.lt/metadata-catalog/catalog/search/resource/details.page?uuid=%7BC8F4086B-03E0-38A5-F5EF-F7A29A15923B%7D, accessed on 21 January 2025).
• Topographic wetness index, derived from digital terrain models using elevation data from the GDR50LT dataset.
• Soil granulometric composition—based on the DIRV_DB10LT dataset (https://www.geoportal.lt/metadata-catalog/catalog/search/resource/details.page?uuid=%7B449450A9-AD8C-6E9E-6FCB-06A0584BF88C%7D, accessed on 21 January 2025).
• Soil productivity score—based on the DIRV_DB10LT dataset.
• Density of the underground drainage network—according to the Mel_DR2LT dataset (https://www.geoportal.lt/metadata-catalog/catalog/search/resource/details.page?uuid=%7B2325543B-D0F7-4D2D-B486-6C10821DE7B5%7D, accessed on 21 January 2025).
• Distance to the nearest drainage network element—based on the Mel_DR2LT dataset.
• Distance to the nearest built-up block—from the Land Parcel Identification System (KZS_DR5LT) database (tps://www.geoportal.lt/metadata-catalog/catalog/search/resource/details.page?uuid=%7B05B2E0DB-31DC-7382-EBB7-7EC5531858A8%7D, accessed on 21 January 2025).
• Distance to the nearest road or railroad block—from the KZS_DR5LT database.
• Distance to the nearest linear water body—from the KZS_DR5LT database.
• Distance to the nearest areal water body—from the KZS_DR5LT database.
• Distance to the nearest agricultural block—from the KZS_DR5LT database; mainly used as a mask to define the search area for potential ALA.
• Distance to the nearest agricultural block with no history of agricultural crop declarations—from the KZS_DR5LT database; mainly used as a mask to identify areas without declarations for EU subsidies.
• Distance to the nearest unused land block—from the KZS_DR5LT database; mainly used as a mask to specify unused areas.
• Distance to the nearest forest land block—from the State Forest Cadastre database (https://www.geoportal.lt/metadata-catalog/catalog/search/resource/details.page?uuid=%7BBF626474-9B36-BFE6-2E43-EAEC396FCFE9%7D, accessed on 21 January 2025).
• Type of landscape use—classified as conservative, conservative-sustaining, sustaining-preserving, sustaining-intensive, intensive-conservative, or intensive.
• Degree of landscape culturalization—classified as forested, forested-agricultural, forested-lowly urbanized, agricultural, agricultural-lowly urbanized, agricultural-urbanized, or urbanized.

All geographic data were converted into raster format with 25 × 25 m cells, corresponding to the NFI grid.

Then, ALA processes in the municipality were analysed using Markov chain models, developed based on observed changes in abandoned areas over the past decade. The modelling was conducted using the TerrSet (v18.21) Land Change Modeler [52]. The factors or drivers listed above were incorporated to assess potential transitions to and from ALA categories, and their performance was evaluated within the ALA models. A multilayer perceptron (MLP) neural network algorithm was applied to model these transitions. Notably, all driver variables underwent the Cramér’s V test to ensure their suitability as explanatory variables. All driver variables were considered static. Two types of models were developed: (1) the transition of abandoned land to non-abandoned land and (2) the transition of non-abandoned land to abandoned land, covering two periods (2012–2018 and 2018–2021). To assess the importance of driver variables in ALA processes, we evaluated the model’s sensitivity by holding independent variables as constant. This paper reports the following performance parameters of the driver variables: influence order, driver ranking (with 1 indicating the most influential and 17 the least influential), model accuracy when a single independent variable is held constant, and model accuracy when all independent variables except one are held constant.

The abandoned agricultural land in 2021 was predicted using transition potential modelled for the 2012–2018 period using ALA maps from the corresponding years, the driver variables listed above, and the change matrix for 2018–2021. The overall model performance was assessed by comparing actual and modelled abandoned agricultural land in 2021. The modelled area exceeded the validation figure by 0.2%.

3. Results

First, we identify the driving forces behind ALA, based on insights from interviewed experts. These findings are expected to provide context for further quantitative investigations. Simultaneously, additional perspectives are explored to elucidate the mechanisms underlying ALA management. ALA-related processes are closely linked to land management, legal administration, monitoring, and decision making concerning land-use changes. However, the interviewed land management stakeholders primarily regard ALA as one of several types of land-use change, i.e., “... we are also responsible for accounting for state land” (4); “... we participate in management decision-making at all levels …” (5); we are “... directly involved in land use monitoring …” (10).

ALA is considered an integral component of broader land-use policy, intended to support decision making and provide a scientific basis for policy development. The scale and scope of information required depend on the specific task: “We use information ranging from the parcel to the national level” (5); “There are multiple levels—one pertains to public sector planning, while another focuses on the planning of specific areas” (6); “Detailed information is only needed for specific applications, such as estimating land tax” (10); “ALA statistics, abandoned land characteristics, and other data are utilized down to the most detailed parcel level” (14).

The conceptual understanding of ALA remains somewhat ambiguous, necessitating a clearer definition of abandoned land: “The definition itself should be improved. Currently, the assessment is overly abstract; the indicators of abandonment should be more clearly defined” (12);“The concept of ALA should perhaps be explored more comprehensively from its foundations” (15). Various factors shape the interpretation of ALA, including legal status, potential land use, vegetation cover, and designated land purpose: “Abandoned land refers to a plot or area that cannot be immediately used for its intended purpose. This may include agricultural land becoming overgrown or other land parcels that are obstructed and therefore unusable” (4); “Abandoned land is overgrown with vegetation but not trees. If trees are present, it is no longer considered abandoned” (10); “Abandonment should be classified based on the designated land use. Shrublands should not be categorized as forest land” (8). It should be noted that the insights provided by informants may differ from the legal definitions officially used in Lithuania of abandoned land or forest land.

Different types of ALA are associated with distinct driving factors and management approaches: “Abandonment varies depending on land use type. The abandonment of forest land, agricultural land, residential areas, and industrial sites each follows different patterns. A more detailed classification into specific subsystems is necessary, as the criteria for each differ significantly” (7); “A separate classification is required: (i) woody vegetation, (ii) wetland areas, and (iii) remnants of buildings or ruins, both in rural and urban contexts” (3).

The processes of ALA are multifaceted and influenced by various natural, legal, economic, and social factors. One of the primary determinants is the quality of land parcels in terms of their suitability for agriculture, which is assessed using indicators such as soil fertility, the condition of drainage systems, topographical characteristics, and parcel size. Informants emphasize the importance of these factors: “In all cases, the key factor for agricultural land is its productivity and holding size” (7) and “The correlation between arable and abandoned land is linked to land productivity. Less fertile land is more likely to become abandoned” (17).

The location of a land parcel is also significant, particularly regarding legal and regulatory conditions that either permit or restrict specific activities (7; 16). Over time, abandoned land frequently becomes overgrown with trees and may eventually transition into forest, highlighting a clear relationship between abandoned and forested land: “The link exists simply because abandoned land becomes overgrown with forest” (15); “This can be interpreted in two ways: when land becomes overgrown with self-seeded trees, it can, over time, be classified as forest land. If we establish a priority for a given area, we can say that it will develop into a forest within a certain period” (11); “Often, overgrown land is reclassified as forest land, and there is no turning back” (4).

However, no such correlation is perceived between abandoned and arable land. Informants stress that arable land is rarely abandoned, and its maintenance is directly dependent on the landowner: “If the land is arable, it is maintained and does not have the opportunity to become overgrown” (17); “Cases of abandoned arable land are relatively rare” (4).

Land ownership patterns also play a crucial role in ALA processes. State-owned land, if left uncultivated, is at risk of abandonment: “… if the land belongs to the state, has not been cultivated, has not been leased, and no one is interested in taking it, it becomes abandoned. How many state-owned plots remain that were never returned? If no one has worked the land for 15 years and there is no owner, it turns into abandoned land” (3). Conversely, the presence of large, active farmers in the vicinity may prevent land abandonment: “… if there are any, they often take over and cultivate these plots” (3). Geographic location also plays a role, as evidenced by municipal oversight efforts: “… municipal oversight has played a significant role. The Vilnius region, for example, has visibly improved” (4).

Lithuania has a significant number of small, state-owned agricultural land plots that receive insufficient government attention, making them particularly susceptible to abandonment: “There are small state-owned land plots located in unattractive areas. If cadastral surveys, property valuation, and auction organization were conducted, the costs would exceed the value of these small plots, and there would be no demand to purchase them” (6); “These small fragmented plots are a major challenge” (5).

A strong correlation is also observed between land abandonment and the condition of drainage systems. The mechanism is straightforward: land with malfunctioning drainage systems is more likely to be abandoned, while abandonment, in turn, accelerates the degradation of these systems. Without regular maintenance, drainage systems become clogged, and plant root systems cause further damage: “There is a direct correlation—if the system fails, the farmer will get stuck in that swamp” (3); “Most abandoned lands were observed in areas where drainage systems had broken down, leading to waterlogging” (10); “As a result of land abandonment, parts of the drainage system deteriorate. If the land is not maintained, drainage systems gradually clog over time. Where people do not take care of their plots, drainage systems also fail” (12); “If the drainage system is not functioning, wetlands emerge very quickly. The land soon becomes abandoned. If the land is left uncultivated, the roots of plants and trees damage the drainage system” (17).

Insufficient funding for the renewal of drainage systems further exacerbates the problem. However, informants indicate that restoring these systems is not always a viable solution: “Projects were initiated to restore drainage systems and make them operational, but many environmental concerns arose. Some areas need to be drained, while others should be left for natural development” (11).

Social factors, particularly migration, emigration, aging landowners, generational transitions, and a lack of continuity in land management, also play a crucial role in ALA. Informants describe how demographic shifts contribute to land abandonment: “Older generations hoped to pass their land on to their children for farming. However, younger generations think differently and are not interested in agriculture. This can be seen as a lack of continuity” (11); “The majority of abandoned lands, based on inspections, are linked to the departure of landowners. Owners who do not live in Lithuania rarely ensure that their land is maintained. The reasons are typically old age or emigration” (12); “A person abandons their lifestyle and former activities, and as a result, the land becomes abandoned. The landowner plays a key role in this process” (15).

Demographic factors, such as the number of farmers in a given area and its geographical location, are also significant determinants of ALA: “The number of large-scale farmers in a region matters—where there are many, abandoned land should not exist” (3); “Topography plays a significant role, as does proximity to central areas. The number of farmers also matters, the more farmers in an area, the less abandoned land there is” (10).

The second part of our study involved identifying and modelling abandoned agricultural land in Jonava Municipality, which is representative of average Lithuanian land conditions. The identified abandoned agricultural land areas were as follows: 1421.6 ha in 2012, 1562.3 ha in 2018, and 2070.8 ha in 2021 (Figure 3). Nearly all abandoned lands identified in 2021 were located within areas classified as grassland by the Lithuanian National Forest Inventory, with only a few instances involving producing land or permanent orchards. It is important to note that virtually none of the grasslands classified as abandoned were situated in areas declared as permanent pastures or meadows for the purpose of obtaining EU agricultural subsidies. The rate of land abandonment was relatively higher during the 2018–2021 period compared to that during 2012–2018, with 246.8 ha transitioning into abandoned land annually, versus 88 ha per year in the earlier period. The rate of land reclamation showed similar trends, with 75.8 ha and 64.5 ha being reclaimed annually during the 2018–2021 and 2012–2018 periods, respectively. The transformations occurred in the southern part of the study area during both periods (Figure 4).

By assessing the potential for transformations between abandoned and not abandoned land, we also determined the influence of various factors on land abandonment-related processes. In all cases, the accuracy of the developed models was sufficiently high, i.e., above 80% (

Table 2. Influence of driver variables on the performance of agricultural land abandonment models.

Driver	2012–2018						2018–2021
	Abandoned to Not Abandoned			Not Abandoned to Abandoned			Abandoned to Not Abandoned			Not Abandoned to Abandoned
	IO *	One **	All ***	IO	One **	All ***	IO	One **	All ***	IO	One **	All ***
1. Land ownership	5	76.1	61.6	6	80.5	50.0	11	84.1	50.5	5	84.6	50.1
2. Steepness of terrain	13	84.2	51.8	14	83.0	50.0	15	87.4	50.5	14	86.3	50.1
3. Topographic wetness index	17	84.9	50.0	16	83.4	50.0	17	88.5	50.5	15	86.4	50.1
4. Soil granulometric composition	6	76.6	51.8	2	74.8	50.0	3	73.5	50.5	7	85.1	50.1
5. Soil productivity score	14	84.4	50.0	17	83.7	50.0	12	85.6	50.5	13	86.2	50.1
6. Density of underground drainage network	9	80.0	52.2	5	80.5	57.4	4	77.4	50.5	4	83.8	63.3
7. Distance to drainage network element	2	70.2	50.0	8	80.8	50.0	6	77.7	50.5	9	85.6	50.1
8. Distance to built-up block	8	80.0	59.3	9	82.2	50.0	7	79.5	50.5	8	85.3	50.1
9. Distance to road or railroad	16	84.8	51.0	10	82.8	50.0	13	86.7	50.5	11	86.0	50.1
10. Distance to water stream	10	80.8	58.6	15	83.2	49.0	5	77.6	50.5	17	86.4	50.1
11. Distance to areal water body	7	76.7	56.3	11	82.9	49.8	9	81.7	50.5	12	86.1	50.1
12. Distance to agricultural block	4	74.6	44.7	3	75.6	50.9	1	57.1	60.0	2	77.4	50.9
13. Distance to agricultural block with no crops declared	3	72.9	41.5	4	79.6	50.0	2	69.3	50.4	6	85.1	54.0
14. Distance to unused land block	1	62.7	41.2	7	80.7	50.0	8	81.3	50.5	3	83.4	52.1
15. Distance to forest land	15	84.8	50.0	1	67.5	57.1	16	87.5	50.5	1	70.7	76.9
16. Type of landscape use	12	84.0	50.0	12	82.9	50.0	10	83.0	50.5	10	85.8	50.1
17. Degree of landscape culturalization	11	82.1	50.0	13	83.0	50.0	14	86.7	50.5	16	86.4	50.1
With all variables	-	85.0		-	83.8		-	88.5		-	86.5

* IO—influence order; **—accuracy (%) forcing a single independent variable to be constant; ***—accuracy (%) forcing all independent variables except one to be constant.

). Generally, the transformation of abandoned land into not abandoned land was modelled more accurately, as the number of such cases was lower. The influence of various factors on the modelled phenomenon is further analysed to highlight the significance of specific ALA drivers, complementing the previous discussion based on expert insights.

Three groups of drivers could be distinguished based on the magnitude and type of their influence on the accuracy of ALA models. Some drivers had either (i) strong to moderate or (ii) weak to no effects on both types of ALA transitions, i.e., ALA abandonment and land reclamation, and (iii) the influence of some drivers depended on the type of transformation. The proximity of a forest (Driver 15 in Table 2) appears to be the most important factor influencing land abandonment processes. It is the primary driver associated with land abandonment. At the same time, it is mathematically the least significant factor in explaining the reverse process—land reclamation—since these processes occur the most slowly near forests. The position relative to cultivated land (Driver 12) is often a significant factor in modelling both processes; however, its importance is likely linked to the presence or absence of a specific cultivated land block. The location of an area in relation to drainage systems (Driver 7) and soil granulometric composition (Driver 4) also emerged as relatively strong factors. Meanwhile, the soil productivity score (Driver 5) and topographic wetness index (Driver 3) were weak factors in all cases. However, it is important to note that these factors are derived from other indicators. Additionally, the mapping of soil productivity may lack the accuracy needed for modelling land abandonment processes. The form of land ownership (Driver 1) was a moderately strong factor in almost all cases. The distance to topographic elements—such as the hydrographic network (Drivers 10 and 11) and roads (Driver 9)—was generally a weak factor. Similarly, factors describing the degree of landscape use (Drivers 16 and 17) were weak, likely due to the relatively coarse nature of their identification.

4. Discussion

Therefore, two primary aspects—the definition of abandoned land and the factors influencing its existence—were identified when examining the underlying causes of ALA. The definition of abandoned land is a crucial component of land-use policy. While the importance of information on abandoned land is widely acknowledged and there are definitions provided in relevant legal acts—areas of agricultural land overgrown with woody plants [41]—there is currently still no clear and precise understanding of the term. Nearly all surveyed experts agree that the legal framework requires improvement, particularly through clearer regulations that define the characteristics of abandoned land and establish a classification system to facilitate more effective management strategies. The existing definition of ALA is often considered overly abstract, as the concept encompasses multiple dimensions, including legal status, potential uses, vegetation coverage, and intended land-use purposes. To enhance clarity, abandoned land could be categorized according to its specific characteristics, such as areas with woody vegetation, waterlogged territories, or sites containing remnants of buildings and ruins. Additionally, legal ownership—whether private or state-owned—plays a significant role in determining responsibility for land maintenance.

The lack of a universally accepted and acceptable definition of ALA is a well-recognized issue [3,53]. Abandoned land is interpreted differently depending on disciplinary perspectives, geographic context, land-use history, scale, and temporal aspects, as well as functional versus legal considerations. This ambiguity makes abandoned land an important scientific challenge, affecting data collection, process monitoring and modelling, and policy formulation for land management. Without a clear definition, land management decisions may become inconsistent, leading to misguided policies, economic inefficiencies, and environmental risks. Future research should focus on developing a classification system that integrates ecological, economic, and social dimensions of land abandonment while remaining flexible enough to accommodate specific contexts.

The presence of abandoned land is influenced by various natural, economic, social, and legal factors. These factors or drivers are further examined in the discussion, where we also reference other studies that substantiate or provide context for our findings. Some factors were identified as ALA drivers by interviewed experts. Among the natural determinants, land productivity and geographical location are particularly significant. Land productivity is often recognized as an ALA driver by other researchers. Land productivity is often intricately linked to agricultural land abandonment (ALA), as declining agricultural viability leads landowners to cease cultivation due to insufficient economic returns [14,22,54,55,56,57]. Low-yielding land is more susceptible to abandonment, especially if it is situated in remote or inconvenient locations, often near forests [58,59]. Afforestation and agricultural land abandonment to the Global North and expansion of agricultural land in the Global South are major causes of LULC changes [58].

The condition of drainage systems is another critical factor. Malfunctioning drainage infrastructure not only contributes to land abandonment but is further degraded when land is left uncultivated. Effective reclamation systems address the soil fertility and hydrological problems associated with abandoned lands and improve agricultural productivity by restoring soil biological status and nutrient composition [60,61,62]. In some cases, rather than restoring drainage systems, certain areas may be more appropriately designated for natural rewilding. Rewilding is a new strategy that promotes ecological restoration by allowing natural processes to reclaim agricultural landscapes [39]. It has great potential to address the challenges of ALA and poorly functioning drainage systems, e.g., by increasing biodiversity and improving ecosystem functions on abandoned farmland [38,63]. In Lithuania, land reclamation is considered an important factor that shaped Lithuanian landscapes [64,65,66]. It should be emphasized that the facilities available for land reclamation in Lithuania influence land use—e.g., afforestation of agricultural lands is dependent on the presence or absence of land with a functioning land reclamation system [67]. Legal and administrative factors also exert a considerable influence, particularly concerning state versus private ownership. Some abandoned land parcels are state-owned, yet insufficient government resources are allocated for their management. Small, fragmented plots located in remote or economically unattractive areas are often neither maintained nor sold due to their low market value [68]. Additionally, abandoned land frequently transitions into forest, as no effective policy measures exist to reintegrate it into agricultural use [69,70,71]. Social factors, including migration and demographic shifts, further contribute to ALA [72,73,74,75]. Many landowners have emigrated from Lithuania or have ceased to manage their properties, leading to progressive neglect. Generational transitions and a lack of continuity in land stewardship exacerbate the issue. Furthermore, economic conditions and market trends play a crucial role. In regions dominated by large-scale farming operations, abandoned land is less prevalent [76,77].

Although the ALA drivers identified by experts were also considered important by other researchers, their usability in local-level spatial planning and their contribution to a deeper understanding of land abandonment processes remain limited without numerical evidence. Therefore, our study was complemented by a modelling exercise designed to test certain mapped landscape characteristics as ALA predictors. There are various methodological approaches to model land use development in general and the ALA in particular, including economic models, system dynamics approaches, cellular automata, and agent-based models [33,36]. These diverse methods can yield differing interpretations of the same processes depending on the chosen approach. Spatial analysis and statistical methods are commonly employed to investigate ALA processes, often by examining relevant attributes within specific areal units [78]. In the current study, we applied spatially explicit modelling to map predictions, supporting more informed management decisions. Our modelling exercise focused on a relatively local scale, which significantly influenced the selection of the drivers tested. As a result, some potential drivers identified by experts could not be assessed or were irrelevant at the municipal level. Most ALA factors tested best fit the category of natural drivers, while only a few could be classified as economic, social, or legal.

Our findings, along with those of numerous other studies, highlight the proximity of forests as a significant factor influencing the natural overgrowth of land by woody vegetation. Despite this, forest proximity has not been widely recognized as a driver of ALA dynamics in national-level assessments. In the absence of human-driven deforestation, forests would likely expand considerably, although they would never entirely cover the Earth’s surface [21]. Landscapes near forest edges tend to exhibit higher natural regeneration rates and distinct ecological dynamics compared to more isolated agricultural lands [79]. Several natural factors facilitate the encroachment of woody vegetation in areas adjacent to existing forests. These include the dispersal of wind-borne seeds from nearby forest stands; elevated soil moisture levels; improved soil structure and nutrient cycling that support more rapid colonization by woody species; and edge effects that create more favourable microclimatic conditions for seedling establishment. Such edge effects may involve more stable temperature and precipitation regimes, protection from wind, modified snow cover, and reduced exposure to disruptive disturbance factors [80,81,82,83,84,85,86]. In addition to ecological factors, socioeconomic drivers such as rural depopulation, migration patterns [70,87], and historical land use legacies [88] also contribute to land abandonment near forests. Research further suggests that areas adjacent to forested regions experience slower rates of land reclamation due to ecological competition and the gradual pace of natural succession in these environments [89]. Nevertheless, the impact of current forest land on ALA was considered by the interviewed experts through natural afforestation as a contributing factor but not as a location-based driver.

Studies support our finding that proximity to cultivated land influences both abandonment and reclamation patterns in agricultural landscapes. For instance, hidden abandonment—a condition in which land is formally maintained as cultivated but not actively farmed—often correlates with the presence or absence of adjacent cultivated land blocks [90]. The intensity of agricultural practices significantly influences the likelihood of land transitioning from abandonment back to active use. Agricultural fields located near productive zones are more likely to be revitalized due to supportive institutional or economic frameworks, increasing the probability that abandoned lands in these areas will be reclassified as actively used [91,92]. A significant factor influencing land reclamation may be the legal classification of agricultural land overgrown with tree vegetation. In cases where the average age of such vegetation exceeds 20 years, the land is designated as forest and must be registered in the State Forest Cadastre [93]. Legally, forest land cannot be reverted to agricultural use. As a result, farmers often remove naturally regenerating tree vegetation before the relatively infrequent inventories can reclassify the land as forest, even if they do not anticipate substantial profits from renewed agricultural use.

The impact of terrain varies depending on geographic location and the specific attributes analysed. While derivatives of digital terrain models are relatively easy to generate, they are highly sensitive to the methods used in their production [94]. Nonetheless, topographical influences are widely regarded as crucial for understanding agricultural dynamics [95], as abandonment rates tend to be higher in areas where rugged terrain complicates farming activities [9]. Findings also suggest that factors such as soil productivity scores and the topographic wetness index are weaker determinants of land abandonment [9]. This slightly contradicts the experts’ opinions on the importance of soil productivity as an ALA factor. The accuracy of the soil maps available for the study could also explain the relatively low importance of soil-related model inputs.

Local hydrology plays a significant role in agricultural viability, with infrastructures such as roads and drainage systems affecting both farming practices and land tenure security. The absence of such infrastructure can contribute to abandonment [96,97]. Soil drainage conditions can either facilitate or hinder the establishment of woody vegetation [98]. In Lithuania, subsurface drainage systems were widely implemented throughout the 20th century to convert wetlands and poorly drained soils into arable land. These underground networks significantly altered local hydrology by lowering the water table and modifying soil moisture regimes, thereby making previously unsuitable areas temporarily viable for agricultural use. However, the degradation or failure of such drainage systems can result in soil re-wetting, which in turn affects the ecological dynamics and promotes conditions favourable for the natural encroachment of woody species [99]. Research further indicates that soils with favourable physical properties, such as good drainage and fertility, tend to sustain agricultural activity longer than those with poorer characteristics [100]. Poor soil quality is a key driver of land abandonment, directly linking soil properties to agricultural viability. Additionally, land ownership structures can either exacerbate or mitigate abandonment trends, particularly in contexts where land circulation mechanisms are inefficient [22].

The integration of quantitative and qualitative research methods to explore the drivers of agricultural land abandonment has been applied in previous studies [25,90,101,102,103]. Ideally, researchers would prefer to quantify all findings; however, measuring and assessing both the scale of ALA and its underlying drivers is a complex task [78]. This challenge is largely due to the availability, suitability, and performance of input data, which are often collected for purposes other than analysing ALA. A major limitation of numerical modelling, including the approach used in our study, lies in the preparation of modelling inputs, as well as their comprehensiveness and suitability for the analysis [104,105]. Additionally, uncertainty is introduced at various stages throughout the data processing chain. In this context, qualitative insights can complement quantitative findings by capturing subjective experiences and stakeholder perceptions, which are often difficult—or even impossible—to express using numerical techniques, particularly when these aspects are not well understood [90]. In-depth interviews provide valuable qualitative data by assessing indicators that are otherwise challenging to quantify, such as stakeholder expectations, aspirations, and intentions [37]. On the other hand, expert insights may not always be trustworthy or empirically supported, which can lead to uncertainties regarding management decisions. Contradictory outputs from qualitative surveys and modelling exercises—such as the differing interpretations of the role of soil productivity, in our case—may also highlight challenges in both conceptualizing the issue and modelling it effectively. Choosing effective informants has long been a key challenge in qualitative research. The qualitative component of our study is based on a relatively small sample of land management policy stakeholders in Lithuania. This necessarily limits the statistical generalizability of our findings to broader populations, such as farmers, who are directly involved in land abandonment or reclamation. Some scholars recommend including no more than 5–6 informants in case study research of this nature [106]. Therefore, our approach was to engage a limited number of experts who possess in-depth knowledge of agricultural processes and who are also informed about, and have the authority to influence, land management policies, including those related to ALA. Thus, integrating quantitative and qualitative methods enables a more comprehensive analysis of ALA, acknowledging the significance of human experiences while supporting findings with statistical trends—something that cannot be fully achieved by relying solely on either approach [38]. For this reason, we adopted a mixed-method approach, which offers a broad understanding of ALA processes based on the expertise of qualified professionals, while also substantiating certain aspects with model-based evidence.

Although this study primarily focused on the methodological aspects of understanding ALA processes, several policy implications for land use management can be observed. The management of ALA processes in Lithuania could be enhanced by refining legal documents and most importantly, by revising the definition of abandoned land. To improve clarity, a new ALA classification system should be developed, integrating the ecological, economic, and social aspects of land abandonment.

Additionally, further exploration of issues related to both the natural and artificial afforestation of non-forest land is essential. A comprehensive set of actions is needed to optimize the existing legal framework—encompassing strategic plans, laws, regulations, and provisions—that directly or indirectly influence afforestation policies [67,107,108,109]. A holistic approach is necessary, not only to support the expansion of forest coverage but also to contribute significantly to the ecological optimization of agricultural areas and the more efficient use of abandoned lands. Areas with naturally occurring woody vegetation have the potential to contribute significantly to the optimization of green infrastructure, particularly in the context of the EU Nature Restoration Regulation [110]. This can be achieved through measures such as the permanent removal of drainage systems on previously drained soils now overgrown with woody vegetation, or by introducing high-diversity landscape features—such as scattered trees, small forest patches, and similar elements—into agricultural areas.

On abandoned lands with non-functional and/or damaged drainage systems, conditions should be established to facilitate afforestation by eliminating the link between land productivity scores and afforestation opportunities. This would help preserve naturally regenerated forests on abandoned lands while ensuring that such land remains untaxed. However, in areas where afforestation on non-forest and abandoned land is not feasible, yet naturally regenerated tree stands exist, landowners should be required to use the land for its designated agricultural purpose and to pay land taxes [41,67,107].

Furthermore, strengthening inter-institutional collaboration between policymaking bodies (such as the Ministry of Agriculture and the Ministry of Environment) and policy-implementing institutions (such as the National Land Service, the State Enterprise Centre of Registers, and the State Forest Service) is essential. Such collaboration is crucial for achieving strategic national objectives, particularly in contributing to the European Union’s New Green Deal [111], which sets targets for climate change mitigation, biodiversity conservation, land restoration, and the promotion of sustainable farming and rural development in Europe [112,113,114,115].

5. Conclusions

First, the definition of agricultural land abandonment should be more precisely specified. Current definitions vary across contexts, leading to ambiguity. For studies like ours, a clear definition based on measurable criteria is essential to ensure accurate mapping and monitoring. This is also important for developing reliable spatial models for predicting and analysing ALA. Most experts agree that ALA involves land not used for its intended purpose, but this definition should be further refined for use in quantitative research.

Second, identifying the drivers of ALA requires a comprehensive approach that combines spatial analysis, expert input, and empirical modelling. Our results highlight the role of natural factors—such as forest proximity and terrain—as well as social, economic, and legal influences like land ownership, migration, and infrastructure access. Soil quality and nearby arable land also play context-dependent roles in abandonment and reclamation processes.

Combining quantitative and qualitative methods allows for a deeper understanding of ALA. Our mixed-method approach helped identify critical areas and supported better land management decisions. It also proved useful for modelling ALA processes in Lithuania by highlighting key influencing factors and drawing on expert knowledge.

Although the developed models demonstrate potential for predicting agricultural land abandonment, they can still be improved. We used only open-source data, but experts suggest including additional variables as potential predictors. Some factors identified by experts did not show strong effects in our results, suggesting that the selection and definition of driving variables should be refined. Future research should explore new variables related to land productivity, location, drainage, ownership, and social context. It is also important to consider how these factors are represented in models—both in terms of meaning and structure.