**1. Introduction**

The monitoring of land-use changes is a key way to understand and assess the dynamic processes in landscapes under different time and spatial scales. A fast-growing human population, the exhaustive use of resources, and increasing environmental concerns have made land-use change monitoring an important topic on the international research agenda [1,2]. The interaction between human activity and land-use changes is an increasing focus of researchers [3,4] due to their impacts on the climate [5], ecosystems [6], water resources [7], soil quality [8], and socioeconomic systems [9]. Land-use changes due to biophysical factors and human activities are accelerating in different regions of the world [10–12]. Even though the issues related to land-use changes are global and cause severe problems in many countries, the change patterns are dependent on local conditions due to numerous factors, such as policies, management, economics, culture, human behaviour, and the environment [13–17]. Thus, it is extremely important to understand the

103

**Citation:** Jukneliene, D.; ˙ Kazanaviˇciut¯ e, V.; Valˇ ˙ ciukiene, J.; ˙ Atkoceviˇciene, V.; Mozgeris, G. ˙ Spatiotemporal Patterns of Land-Use Changes in Lithuania. *Land* **2021**, *10*, 619. https://doi.org/10.3390/ land10060619

Academic Editors: Krystyna Kurowska and Cezary Kowalczyk

Received: 14 May 2021 Accepted: 7 June 2021 Published: 9 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

processes shaping land-use changes at different scales, ranging from regional to global. Such knowledge is of critical importance to build the policies and managemen<sup>t</sup> plans needed to understand and improve the land-use change trends [12,17–20].

Land-use change in Lithuania has always been dynamic. The radical political, economic, and social developments that took place in the country over the last half century undoubtedly had impacts on the land use. Official statistics indicate [21] that 45.6% of the country's area is covered by arable land, 33.5% by forest, 6.21% by meadows and natural pastures, 5.2% by wetland, 5.4% by built-up land, and 4.09% by other. The area proportions of all land-use types, except for agricultural land, have changed relatively steadily during the last five decades; however, the trends of producing land and grassland development changed their trajectories around 1990 and again in about 2005 [22]. The demand for up-to-date information on land cover and land-use changes is increasing due to rapid landscape development as a result of fast processes in the agricultural sector, the growth of urban areas, and the depopulation of some regions, followed by renaturalization [23]. To implement the European Landscape Convention (2020) [24], the Lithuanian authorities (the Environmental Protection Agency of the Ministry of Environment) conduct regular monitoring of landscape changes. Such monitoring delivers facts on landscape development peculiarities and the factors behind the trends, which are needed to predict potential future opportunities and risks [25]. Nevertheless, the data collected and methods of analyses differ from region to region. Scientific research in this area, to the best of our knowledge, has always been sparse. The changes of land cover structure were assessed on 100 test sites (totalling 2.5 km2) in 1976–1986, 2005–2006, and 2012–2013 by the Institute of Geology and Geography (2008; 2015). Often, CORINE information was mobilized to assess the historical development of land cover [26–29]. Information related to land use in Lithuania may also be available from several nationwide GIS databases, such as the Spatial dataset of georeference base cadastre (GRPK) or the Land Parcel Identification System (KZS) Database, which are maintained by state institutions and available for free from the Spatial Information Portal of Lithuania (geoportal.lt). Together with the information on declared land uses and agricultural parcels, this could make an excellent land-use dataset for scientific research; however, such data are only available from 2010 onward. Usually, only the most recent version of the data is freely available. Thus, the availability of suitable data could be another reason behind the limited research focus on land-use retrospection.

Land use and its changes are not only important for the development of the economy or the protection of the environment but are also recognized as having a significant impact on human-induced greenhouse gas (GHG) emissions [30,31]. Land use and its changes may result in GHG removal if certain active measures are applied, such as afforestation, reforestation, revegetation, etc. [32,33]. In order to estimate such emissions and removals, the land use, land-use change, and forestry (LULUCF) sector's GHG reporting was included under the requirements of UNFCCC reporting. Despite the sector's ability to capture GHG emissions from the atmosphere and sequestrate it in biomass or soil, the LULUCF sector was not included in the climate change mitigation target until 2021 [34]. Beginning in 2021, the LULUCF sector will play a role in the flexibility option to reach compliance with other sectors' GHG emission reduction target.

To meet its international climate change mitigation commitments and fulfil the obligation of reporting on GHG emissions and removals in the LULUCF sector, Lithuania introduced an original land-use monitoring system, which became an integral part of the National Forest Inventory (NFI), implemented by the State Forest Service [35,36]. The inventory uses a network of 16,349 systematically allocated sampling points. The land-use type and subtype were identified at each point following the Good Practice Guidance for Land Use, Land-Use Change and Forestry (IPCC 2003), also taking into consideration the requirements of the United Nations Framework Convention on Climate Change and the Kyoto protocol for each year starting in 1971. Past land uses at each point were identified using available historical maps, such as topographic maps, land managemen<sup>t</sup> maps, orthophotos, or satellite images [37]. The information collected in the sampling plots was

used to prepare a land use and land-use change database, in addition to conventional forestry statistics, traditionally attributed to forest inventories. This information has been used in Lithuania to conduct greenhouse gas (GHG) accounting and reporting in the Land Use, Land-Use Change, and Forestry (LULUCF) sector since 2010. Usually, conventional land-use-data-based exercises are based on aggregated statistical information at the country level. Considerable spatial patterns of land-use distribution may be seen in a relatively small country such as Lithuania.

The fast progress of geographic information systems (GIS) during the last few decades provided researchers with powerful tools with which to conduct spatial analyses and modelling [38]. In Lithuania, there were few attempts to use GIS as a tool in land-use-related studies. For example, Kucas et al. [39] applied a multiscale analysis of forest fragmentation in Lithuania to demonstrate the technique with CORINE data. Lazdinis et al. [40] suggested an alternative—the average shortest distance to the closest forest—to forest cover percentage, better describing the spatial distribution of forested habitats for birds in an afforestation study. Jukneliene and Mozgeris [41] compared two GIS databases, representing the forest cover at a nominal scale of 1:10,000 and referring to two dates—1950 and 2013. The data were aggregated for the analyses up to the municipality level. The Global Moran's *I* statistic and Anselin Local Moran's *I* were used to identify global and local patterns in the distribution of forest cover characteristics in Lithuanian municipalities. The authors provided the reader with updated statistics on forest cover in Lithuania just after WWII and discussed the trends of forest cover dynamics during the second half of the 20th century. Recently, Manton et al. [42] used a local hotspot analysis to study peatlands in the Nemunas River basin. However, all these studies used wall-to-wall land-cover and land-use maps, referring to specific dates. The lack of continuously supplied information over time introduces some uncertainties in land-use change trajectories and, simultaneously, makes generalizing about land-use changes more challenging. A distinctive feature of the current study is that we analyse land-use data collected through sampling annually and covering the period since 1971. Another advantage of GIS is the opportunity to integrate for joint analysis data collected using different techniques, formats, time periods, and sometimes applications, but all sharing the same geographic location [22]. The availability of free multisource and multipurpose GIS data in the country has notably increased during the last decade since the implementation of the Spatial Information Portal of Lithuania [43]. All this potentially offers enhanced opportunities for a better understanding of the processes behind land-use development and facilitating land managemen<sup>t</sup> policies.

The aim of current study is to map and explain the land-use changes in Lithuanian municipalities in the period since 1971. We map land use types that are considered the most significant in terms of carbon storage using land-use data originating from the Lithuanian NFI. Then, we evaluate and explain the land-use changes during different periods using factors that are extracted from freely available GIS databases. Finally, we discuss the spatial patterns observed in both land use and land-use change geography, associating them with land-use policy implications.

### **2. Materials and Methods**

### *2.1. Study Area*

The study focuses on land use and land-use changes in Lithuania. Geographically, even though Lithuania is situated in central Europe with central coordinates of 55◦10 N, 23◦39 E (Figure 1), it has strong historical links with Eastern Europe. The total land area of Lithuania is 65,200 km2. 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). The terrain features numerous lakes and wetlands, and a mixed forest zone covers over 33% of the country. Lithuanian climate conditions and natural soil productivity are generally favourable for crop production. Consequently, more than 50% of its land area is used for agricultural purposes. Currently, Lithuania is dominated by rural landscapes, covering approximately 75% of its territory. The proportion of natural landscapes does not exceed

15% of the country's area and they are concentrated in the eastern and southeastern regions, the hilly western parts of the country, and the ancient delta on the shoreline [44]. The rest of the country is covered by rapidly expanding urban or urbanized landscapes. The administrative units of the Republic of Lithuania are 10 counties and 60 municipalities.

**Figure 1.** Specification of the study area: **left**—location of the study area in Europe, **centre**—elevation in Lithuania, **right**— average soil productivity grade. Sources of the data used: **left**—thematicmapping.org/downloads/world\_borders.php (accessed on 13 May 2021), **centre**—GDB200 database from www.gis-centras.lt/ (accessed on 13 May 2021), **right**—derived using Dirv\_DR10LT from www.geoportal.lt (accessed on 13 May 2021).

### *2.2. Input Data*

Two types of input data were used in the study—(i) data describing the land uses in Lithuania and (ii) data describing the factors influencing the land-use changes. Land-use information was available from the Lithuanian National Forest Inventory, which involves permanent observation of land-use types on a network of 16,349 systematically distributed sampling plots [36,45]. NFI sampling plots are distributed in all land-use types across the country in clusters of four sampling plots on a 4 × 4 km grid. One-fifth of the sampling plots are visited each year by field measurement specialists; therefore, the whole country is covered in a five-year inventory cycle. Land-use types and subtypes are identified annually at the centre of each plot from 1971 using the nomenclature of GHG inventories [46], and land-use changes, if occurring, are detected and reported according to the measurement year. The land cover is further grouped according to the GHGC Level 1 coding of land cover: forest, producing land, grassland/pasture, wetlands, built-up areas, and other land. It should be noted that the identification and monitoring of land-use types became the responsibility of the Lithuanian NFI in 2011. To reconstruct the land-use types for each of the nonforest sampling plots for the period 1990–2011, a special study was conducted based on the use of all available historical materials, e.g., remotely sensed data, including orthophotos and satellite image archives, and land managemen<sup>t</sup> and real estate maps [37].

Land-use statistics were aggregated to the level of Lithuanian municipalities. The borders of municipalities (USE\_3 level) were acquired from EuroBoundaryMap (v3.0), which is a European reference database of administrative units and boundaries established within the framework of EuroGeographics (Available online: eurogeographics.org/maps-foreurope/ebm/, accessed on 13 May 2021). We excluded from the study nine predominantly urban municipalities (Figure 1); thus, the study was done on 51 municipalities with a mean area of 1260 km<sup>2</sup> (standard deviation = 452). The municipality for each observation point was identified using the Spatial Join tool of ArcGIS (v10.7) by specialists of the State Forest Service responsible for GHG inventories in the LULUCF sector. Summarized data on all the land-use types and subtypes from 1971 to 2015 were joined to the borders of

each municipality. Usually, the proportions of observation points belonging to particular land-use types were calculated for each municipality and used in further analyses.

Free data available from the Spatial information portal of Lithuania (Available online: geoportal.lt, accessed on 13 May 2021) were used to describe the factors influencing the land-use changes. The datasets used to ge<sup>t</sup> the explanatory variables were the Georeference spatial dataset (GDR10LT), a soil spatial dataset at a scale of 1:10,000 (Dirv\_DR10LT), a land reclamation and wetness dataset at a scale of 1:10,000 (Mel\_DR10LT), a dataset of special land-use conditions at a scale of 1:10,000 (SŽNS\_DR10LT), a dataset of abandoned agricultural land (AŽ\_DRLT), CORINE land covers for 1995, 2000, 2006 and 2014, a land parcel block database referring to 2004, 2008 and 2014 (KŽS), population census data for 1970, 1989, and 2011, including geospatial data for 2011, data on agricultural crops declared to the National Paying Agency for 2010–2015, and a digital raster elevation model (cell size: 100 m) built based on information available in the GDB200 GIS database. Each vector dataset was overlain with the municipality polygons and summary statistics, such as total area or length, and the area/count proportion was extracted for a specific geographic object or phenomenon. If the explanatory variable was available in the raster, we used ArcGIS function Zonal Statistics to estimate the statistics of a certain attribute within each municipality. In the case, the geographical data required additional processing, so the standard functionality of ArcGIS Desktop was used. In such a way, e.g., the slope was estimated using the digital elevation model as the input. To estimate the population within a 15-min driving distance of the centre of each municipality, we used a road database, referring to the year 2007. The road network was constructed using input vector data corresponding to current data of the Georeference background cadastre (GRPK), with all field and forest roads included. Accessibility was calculated using standard ArcGIS Network Analyst New Service Area functionality within the framework of the FP7 RURALJOBS project [47]. Additionally, we used agricultural census data, available from the Official Statistics Portal of Lithuania [48]. All the attributes characterising the municipalities are summarised in Table A1.

### *2.3. Mapping and Evaluating the Land-Use Spatial Pattern*

The proportions of forest, producing land, meadow/pasture, wetlands, built-up land, and other land in municipalities were plotted on the map. The Global Moran's *I* statistic and Anselin Local Moran's *I* were used to identify global and local patterns in the distribution of land-use characteristics in Lithuanian municipalities, respectively. To estimate the spatial distribution patterns, we used the spatial statistics tools available in ArcGIS Desktop. The land uses in municipalities were visualized and analysed at the following points: 1971, 1990, 2005, and 2015. The first and last years refer to the starting and ending points of land-use data available for the study, and the years 1990 and 2005 were chosen to correspond to the restoration of Lithuanian independence and joining the European Union, respectively. These dates also fit the overall development trajectories of producing land and meadow/pasture for the whole country [22]. To quantify the presence of a monotonic increasing or decreasing trend in the changes of land-use proportions during a specific period, we performed a nonparametric Mann–Kendall test and then estimated the slope of the linear trend with the nonparametric Sen's method using MAKESENS tools [49]. The spatial distribution of the slope was visualized and analysed using the same approaches as used with the land-use proportions and described above. The trends were analysed for the following periods: 1971–2015, 1971–1990, 1990–2005, and 2005–2015.

To understand the factors behind the land-use changes in Lithuanian municipalities, we applied an ordinary least squares (OLS) regression. The focus was on the changes in proportions of forest, producing land, and meadow/pasture during all the periods mentioned above. As the dependent variable, the slope of the linear trend in land-use proportion changes was used. All the variables extracted from the freely available GIS databases were considered as candidates for explanatory or independent variables. We checked all possible combinations of input candidate explanatory variables using the Exploratory Regression

tool of ArcGIS Desktop. The number of independent variables ranged from two to five. The following conditions for the fit of the regression models were set: only explanatory variables with statistically significant coefficients (95% confidence level) and with a variance inflation factor under 7.5 were exploited to avoid multicollinearity; the minimum Jarque–Bera *p*-value was 0.1 to consider the model residuals to be normally distributed; and model residuals were tested for spatial clustering using Global Moran's *I* (maximum value allowed: 0.1) for the cases that met all the above search criteria. We evaluated the extent to which each candidate independent variable met the above conditions. Only the best regression models (in terms of adjusted *R*<sup>2</sup> and corrected Akaike information criterion, under the condition that all other statistical tests—Jarque–Bera statistic, Koenker (BP) statistic, variance inflation factor, and spatial autocorrelation of the regression residuals—were passed) are presented in the current paper.

The methodological framework of our study is summarized in Figure 2.

**Figure 2.** Flowchart summarizing the overall structure of the study.
