Land Use/Cover Change, Fragmentation, and Driving Factors in Nepal in the Last 25 Years

Abstract

Because of the influence of climate change and human activities, an in-depth analysis of land use/cover change (LUCC) and its drivers in Nepal is important for local community forestry management and sustainable development. This paper analyzed the direction, magnitude, and rate of LUCCs and their spatial aggregation, as well as landscape fragmentation in Nepal, from 1995 to 2020 using the ESA/CCI (European Space Agency Climate Change Initiative) dataset. A total of 10 factors including population, socioeconomic development, climate factors, and forest management factors were selected to determine the dominant driving factors affecting LUCC in Nepal by Principal component analysis (PCA) and linear regression analysis. Our study showed that climate change, human activities, and forest management (e.g., community forestry) all influenced LUCC. In Nepal, land use/cover shifted among forest, shrub, grassland, and cropland from 1995 to 2020, mainly from forest to cropland. The most significant LUCC in recent decades has been caused by the expansion of cropland and urbanization. The area of coniferous and broadleaf forests decreased from 1995 to 2001 due to deforestation and forest degradation, and recovered gradually after 2001, which was attributed to the successful practice of community forestry in Nepal. Accelerated urbanization was also found in Nepal, and the significant expansion of construction land came mainly came from cropland. Land fragmentation in Nepal was severe and exhibited spatial aggregation characteristics. Human activities played a greater role in LUCC in Nepal than climate factors. The community forestry, GDP growth, and precipitation were positive driving factors for increases in forest area, while the development of the services sector and rising temperature were negative driving factors.

1. Introduction

Land use/cover change (LUCC) is one of the main drivers of global environmental change [1,2,3], a direct representation of external influences on the regional ecosystem, reflecting the basic laws of human activity and socio-economic development [3,4,5]. Spatial and temporal LUCC and its driving forces have become a hot research topic in areas related to global environmental change and sustainable land use [5,6,7]. Studies using Landsat images showed that climatic change, human activities, and government policies all influence LUCC. Natural factors play an important role in LUCC, while the frequency and extent of LUCC caused by human activities are escalating worldwide. Identifying regional LUCC and its drivers is crucial for land use planning and management [8,9]. In addition, LUCC has significant impacts on the ecological integrity of forests, biodiversity, and natural resources, such as urbanization, which can block the connectivity of natural vegetation and cause fragmentation [10,11]. It has been suggested that the area of global land use change in the last 60 years is four times larger than previously estimated [12]. Human land use activities may lead to greater deforestation and are important determinants of forest fragmentation [13]. The increase in land demand due to massive land acquisition [14], population growth, urbanization [15,16], and agricultural land expansion [17,18,19] can all lead to forest fragmentation. Forest fragmentation increased in Nepal between 1930 and 2014 [20], and the Central Himalayan region of northern India experienced intensive forest fragmentation between 1990 and 2014, shifting from a forest-dominated landscape to a fragmented forest mixed with cropland and urban settlements [21]. Progressive fragmentation may have significant ecological impacts on species dependent on forests [20]. Various methods have been used to analyze and detect the characteristics of land use change. For example, the source and direction of LUCC and its spatial variation can be identified by the land use transfer matrix [5,22,23]. Previous studies have shown that principal component analysis (PCA) is an effective method for detecting the main drivers of LUCC transfer from different dimensions [24].

Nepal is one of the least developed countries (LDCs) identified by the United Nations [25]. Nepal is also a pioneer country in adopting community forestry, with two-fifths of the national forest under the community forestry system involving more than half of the households in the country [26]. Community forestry has succeeded in improving forest conditions in Nepal, with direct and indirect benefits, and is the foundation of the socioeconomic sector and community livelihood development in Nepal. However, global climate change, rising urban population, unplanned land utilization, shortage of clean water supply and reliable forest resources, and land degradation have posed serious threats to Nepal’s sustainable development. In the past hundreds to thousands of years, the dominant land use activity was deforestation for farmland and pasture in South Asia [27]. Limited studies have been conducted on past and present LUCC in Nepal and they have found that Nepal’s forests and snow/glacier cover have declined, agricultural land has increased, and rapid urbanization has led to significant changes in Nepal’s land use [20,28,29,30,31,32]. Biophysical drivers (e.g., climate change) are relatively less important, and socioeconomic drivers (e.g., population growth and economic growth) are more important for agricultural land to forest change in South Asia countries, such as Nepal [33]. In Nepal, India, and elsewhere, land fragmentation is becoming a key obstacle to improving the productivity and sustainability of land resources [34]. However, integrated research on LUCC and land fragmentation across this region is generally lacking.

This study focuses on land use/cover change in Nepal under the influence of climate change and human activities over a long time series. The direction, magnitude, and rate of LUCC and land fragmentation in Nepal from 1995 to 2020 were analyzed. PCA and linear regression analysis were used to determine the main driving forces behind the land use/cover changes in Nepal. The analysis is intended to provide a basis for community forestry management and planning, and the development of sustainable resource management in Nepal.

2. Materials and Methods

2.1. Study Area

Nepal is a landlocked mountainous country in South Asia (26°22′–30°27′ N, 80°04′–88°12′ E), located in the southern foothills of the Himalayas, bordering China to the north and India on the remaining three sides. Nepal has a total land area of 147,181 km² and a population of 28.9 million, with a huge topographic gap from north to south. Elevations ranging from the south to north as plain, hill, mountain, and high Himalaya [35] (Figure 1a). Nepal has a subtropical monsoon climate, and the year is basically divided into two seasons, with the rainy season from April to September and the dry season from October to December and January to March of the following year [36]. From 1995 to 2020, increase in both precipitation and temperature were trends in Nepal (Figure 2).

The land types of Nepal in 2020 were dominated by forest, cropland, and grassland, which accounted for 49.9%, 29.2%, and 15.6% of the total area of Nepal, respectively (Figure 1b,c). The forest resources of Nepal were abundant, and the forest cover had reached 44.7% in recent years due to the successful practice of community forestry [35]. Cropland in Nepal accounts for 28.7% of the total area and crops are mainly rice, wheat, and maize [37].

2.2. Forests in Nepal

Forest is the dominant land cover type in Nepal, with a majority of coniferous and broadleaf forests [38]. As can be seen from Figure 3, the area of the broadleaf forest was significantly more than that of the coniferous forest until 1999, but the forest area began to decrease significantly in 1999, especially the broadleaf forest decreased sharply. The area of both types of forest reached a minimum in 2001 and then recovered [22,28]. Fluctuations in broadleaf and coniferous forest areas were generally consistent after 2001, and the area of broadleaf forests was larger than that of coniferous forests after 2016.

2.3. Key Economic Indicators and Population of Nepal

In Nepal, the percentage of the primary sector (agriculture, forestry, and fishing) and the secondary sector (industry and construction) has declined significantly since 2001, while the third sector (services) has increased significantly, which corresponds to the recovery of forest area since 2001 (Figure 4). The change in GDP indicators (Figure 5a) showed an increasing trend in GDP and an overall increase in GDP growth. The population in Nepal increased fast and the total population was estimated at about 30.0 million people in 2020 (Figure 5b).

2.4. Data

LUCC data are from the latest version v2.1.1 of ESA/CCI (European Space Agency Climate Change Initiative) land use/cover product, which uses the UN Land Cover Classification System (LCCS) with a spatial resolution of 300 m. The overall accuracy of this data at the global scale is 74.4% [40], and the LUCC data period used in this study is 1995–2020. The statistical data were obtained from the World Bank and related literature, and the precipitation and temperature data were extracted from the ERA5-Land dataset, as shown in

Table 1. Data sources used in this study.

Data Type	Source	Download Address
LUCC	ESA/CCI	http://maps.elie.ucl.ac.be/CCI accessed on 13 January 2023
Elevation data	SRTM 90 m	https://www.resdc.cn/data.aspx?DATAID=284 accessed on 19 December 2022
Statistical data	World Bank	https://data.worldbank.org/country accessed on 9 January 2023
Meteorological data	ERA5-Land	https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land-monthly-means?tab=overview accessed on 14 February 2023

.

Compared to the national scale dataset provided by ICIMOD (International Centre for Integrated Mountain Development), the area extracted from ESA/CCI for each land type in Nepal is slightly different. However, the new ESA/CCI product provides the first, more detailed, long-time series of global land cover changes [28,41]. The uncertainty of ESA/CCI land cover maps is from the spectral reflectance classification algorithm to the land classes [42,43]. ESA/CCI defines more hybrid categories. Some studies have pointed out its high accuracy under the first level of classification, such as forest land, cropland, construction land, and bare land, and the existing global-scale land cover products have low shrub accuracy [44,45,46].

2.5. Methods

2.5.1. Data Pre-Processing

The ESA/CCI data were reclassified and spatially overlaid using ArcGIS 10.5 to finally extract 10 types [44]; the specific classification is shown in

Table 2. Reclassification description.

Land Cover Types	Value of ESA/CCI 1
Coniferous forest	70, 80
Broadleaf forest	50, 60
Mixed forest	90, 100, 110
Shrub	120
Grassland	130, 140
Cropland	10, 20, 30, 40
Wetland water	160, 170, 180
Construction land	190
Permanent snow and ice	220
Bare land	150, 200

¹ Legend of ESA/CCI land cover maps is based on LCCS.

. The area of each land/cover type from 1995 to 2020 in Nepal was summarized and PCA and linear regression analysis of the drivers were conducted using SPSS 24.0.

2.5.2. Land Use/Cover Transfer Matrix

The transfer matrix can effectively obtain information about each land type, such as transfer area and transfer direction so that the characteristics of LUCC in Nepal can be effectively monitored and analyzed. The transfer matrix was established using the spatial analysis module in ArcGIS 10.5 [47]. The transfer matrix equation is shown below.

$$P_{ij}=\left(\begin{array}{ccc}{p}_{11}& \cdots & p_{1j}\\ ⋮& \ddots & ⋮\\ p_{i1}& \cdots & p_{ij}\end{array}\right)$$

(1)

In Equation (1), P_ij is the area transfer matrix; p_ij is the area transformed from i land use type in period k to j land use type in period k + 1.

2.5.3. Land Use Type Dynamic Degree

The rate of LUCC is expressed in terms of the single land use type dynamic degree [48]. It is calculated as follows,

$$K=\frac{U_b-U_a}{U_a}\times \frac{1}{T}\times 100\%$$

(2)

In Equation (2), K is the intensity of land change, U_a and U_b are the quantities of a certain land use type at the beginning and end of the study, and T is the length of the study period. When T is set to years, the value of K is the annual rate of change of a certain land use type.

We chose 2001 as the middle year for analysis, mainly because the forest area in 2001 was the smallest, and the forest cover was the main type of land cover accounting for 44.7% of the land area in recent years.

2.5.4. Land Use/Cover Fragmentation Index

The landscape fragmentation index was used to describe and reflect the degree of fragmentation of land use/cover fragmentation, and to some extent, the disturbance and impact caused by human activities on land use/cover change [49]. The landscape fragmentation indexes mainly include the number of patches (NP), patch density (PD), maximum patch index (LPI), and landscape separation index (DIVISION) [50]. The specific value domains and ecological significance are shown in

Table 3. Range and significance of landscape fragmentation index.

Landscape Fragmentation Index	Abbreviation	Range	Ecological Significance
Number of patches	NP	NP ≥ 1	Number of all patches in a landscape type. The larger the value, the higher the degree of fragmentation of the landscape pattern.
Patch density	PD	PD > 0	Number of patches per unit area of a certain landscape type, reflecting the density of patches.
Maximum patch index	LPI	0 < LPI ≤ 100	The proportion of the total area of the largest patch of a certain landscape type. Reflects the degree of patch concentration and landscape dominance type.
Landscape separation index	DIVISION	0 < DIVISION ≤ 1	Reflects the degree of landscape separation. The more the value tends to 1, the higher the degree of landscape segmentation.

.

2.5.5. Spatial Aggregation Analysis

Randomly distributed events may also exhibit some degree of spatial clustering, and the goal of hotspot analysis is to identify regions with statistically significant clustering, as this indicates that these events are being influenced by some spatial process factor and there is a spatial correlation. To be a statistically significant hot spot, elements should have high values and be surrounded by other elements that also have high values. Similarly, to be a statistically significant cold spot, elements should have low values and be surrounded by other elements that also have low values.

In this paper, 10 classes were used to represent 10 land types, such as coniferous forests, and hotspots were analyzed using the spatial statistical analysis tool of ArcGIS, which classifies each pixel according to its type (the object of study in this paper is land type) to represent spatial aggregation or spatial dispersion [51].

2.5.6. Driving Factor Selection

In this paper, a total of 10 drivers were selected in four categories including population, socio-economic development, climate factors, and community forest development based on the principles of science and accessibility [5,39], as shown in

Table 4. Drive factor description.

Types	Driving Forces (Unit)	Symbol
Population	Population	X1
Socio-economic development	GDP (constant 2015 US$)	X2
	GDP growth (annual %)	X3
	GDP per capita (constant 2015 US$)	X4
	Agriculture, forestry, and fishing, value added (% of GDP)	X5
	Industry (including construction), value added (% of GDP)	X6
	Services, value added (% of GDP)	X7
Climate factors	Mean annual precipitation (mm)	X8
	Annual mean temperature (degree Celsius)	X9
Community forest development	Number of new community forests (PCS/year)	X10

.

2.5.7. Analysis of Driving Factors

SPSS 24.0 software was used to perform principal component and linear regression analyses.

Step 1: PCA is performed, using the idea of dimensionality reduction to transform multiple indicators into several composite indicators, keeping as much as possible what is reflected by the original multiple indicators [5,22,24,47]. The standardized (Z-score) independent variables (drivers X₁, X₂, …, X₁₀) are subjected to PCA, and the top k (k ≥ 2) principal components F₁, F₂, …, F_k are selected based on the selected cumulative contribution to reveal the driving mechanism of LUCC changes in Nepal.

Step 2: The dependent variable is the area of coniferous or broadleaf forests in Nepal and the independent variables are the driving factors affecting LUCC, and a linear regression analysis model was applied to quantitatively reveal the driving mechanisms of forest area change [6]. After determining the k principal component factors used in the analysis, linear regression analysis was applied to calculate the functional relationship between the dependent variable and the extracted principal components, with the equation,

$$Y={\beta }_0+{\beta }_1{F}_1+{\beta }_2{F}_2+\dots +{\beta }_k{F}_k$$

(3)

In Equation (3), Y is the dependent variable, i.e., land cover type area, F₁, F₂, …, F_k (k ≥ 2) are the independent variables, i.e., the principal component score coefficients for each year, β₀ is the constant term, and β₁, …, β_k are the regression coefficients.

Step 3: Since each principal component F₁, F₂, …, F_k (k ≥ 2) is a linear combination of the independent variables (drivers X₁, X₂, …, X₁₀), the final linear regression model can be obtained after transformation, with the equation,

$$Y={\alpha }_0+{\alpha }_1{X}_1+{\alpha }_2{X}_2+\dots +{\alpha }_{10}{X}_{10}$$

(4)

In Equation (4), Y is the dependent variable, i.e., land use type area. X₁, X₂, …, X₁₀ are the independent variables, i.e., drivers. α₀ is the constant term, α₁, …, α_k are the regression coefficients.

3. Results

3.1. Characteristics of Land Use/Cover Transfer in Nepal

3.1.1. Area and Direction of Land Use/Cover Transfer

Land use and land cover in Nepal mainly shift between forest, shrub, and grassland. Cropland, and mixed forest, grassland, snow/ice, wetland water, and bare land are relatively stable. From 1995 to 2001, land in Nepal shifted mainly between forest and cropland (

Table 5. Land use/cover transfer matrix of Nepal from 1995 to 2001.

1995/km2	2001/km2
	CF	BF	MF	SH	GL	CL	WW	UB	IS	BL	SUM
CF 1	32,560	67	24	29	13	400	0	0	0	0	33,093
BF	76	32,332	60	40	14	2556	1	4	0	1	35,084
MF	20	7	6708	0	1	12	0	0	0	0	6748
SH	41	42	0	530	0	17	0	0	0	0	630
GL	9	4	1	0	23,097	20	0	5	1	2	23,139
CL	186	289	8	1	8	41,178	1	30	0	1	41,702
WW	0	0	0	0	0	2	317	0	0	0	319
UB	0	0	0	0	0	3	0	137	0	0	140
IS	0	0	0	0	1	0	0	0	4894	0	4895
BL	0	0	1	0	19	8	0	0	0	1869	1897
SUM	32,892	32,741	6802	600	23,153	44,196	319	176	4895	1873	147,647
Change from 1995 to 2001/km2	−201	−2343	54	−30	14	2494	0	36	0	−24	−201
Rate of change from 1995 to 2001/%	−0.61	−6.68	0.80	−4.76	0.06	5.98	0.00	25.71	0.00	−1.27	−0.61

¹ CF, coniferous forest; BF, broadleaf forest; MF, mixed forest; SH, shrub; GL, grassland; CL, cropland; WW, wetland water; UB, construction land; IS, permanent snow and ice; BL, bare land.

, Figure 6). The area of coniferous and broadleaf forests decreased, and the largest transfer was made to cropland, accounting for about 6.7% of cropland. Cropland was transferred to coniferous and broadleaf forests and construction land, with about 0.6% of coniferous forest area transferred from cropland and 0.9% of broadleaf forest area transferred from cropland. About 19.9% of construction land was transferred from cropland. From 2001 to 2020, the land transfer occurred between coniferous forests, broadleaf forest, shrub, cropland, and construction land (

Table 6. Land use/cover transfer matrix of Nepal from 2001 to 2020.

2001/km2	2020/km2
	CF	BF	MF	SH	GL	CL	WW	UB	IS	BL	SUM
CF 1	32,481	107	121	6	10	166	0	1	0	0	32,892
BF	100	31,845	49	44	3	696	1	4	0	0	32,742
MF	171	49	6546	1	15	20	0	0	0	0	6802
SH	102	144	4	321	0	28	0	0	0	0	599
GL	47	31	21	0	22,966	38	1	3	1	45	23,153
CL	490	1249	69	17	56	42,140	1	165	0	6	44,193
WW	0	1	0	0	0	2	316	0	0	0	319
UB	0	0	0	0	0	2	0	175	0	0	177
IS	0	0	0	0	1	0	0	0	4894	0	4895
BL	0	0	0	0	26	6	0	0	0	1840	1872
SUM	33,391	33,426	6810	389	23,077	43,098	319	348	4895	1891	147,644
Change from 2001 to 2020/km2	499	684	8	−210	−76	−1095	0	171	0	19	-
Rate of change from 2001 to 2020/%	1.52	2.09	0.12	−35.06	−0.33	−2.48	0.00	96.61	0.00	1.01	-

¹ CF, coniferous forest; BF, broadleaf forest; MF, mixed forest; SH, shrub; GL, grassland; CL, cropland; WW, wetland water; UB, construction land; IS, permanent snow and ice; BL, bare land.

, Figure 6). The area of coniferous and broadleaf forests increased, mostly transferred from cropland and shrub. Moreover, 1.5% of the area of coniferous forests was transferred from cropland, 3.7% of the area of broadleaf forests was transferred from cropland, and 2.0% of the cropland was transferred from forests. About 47.4% of the construction land was transferred from cropland.

Spatially, the increase of cropland was mostly distributed in the hill, which was consistent with the distribution of population, mainly transferred from the forest (Figure 7). Forest was mainly and widely transferred from cropland, grassland, and shrub, and the transfer of cropland to forest land mainly occurred in the high mountains in the central east. The increase in construction land was concentrated around cities, such as Kathmandu and Kaski. In addition, some bare land was developed into grassland and cropland.

3.1.2. Transfer Magnitude and Speed

According to the changes in the area of coniferous and broadleaf forests, the period 1995–2020 was divided into two time periods, 1995–2001 and 2001–2020, to analyze the amount and rate of change in the area of land use/cover types in Nepal.

Figure 8a showed the amount of area change of land use/cover types. The area of coniferous and broadleaf forests decreased from 1995 to 2001 and increased significantly from 2001 to 2020. The increase in forest area in Nepal was attributed to local community forestry management [52,53]. The increase in construction land from 2001 to 2020 was much greater than that from 1995 to 2001, which was consistent with the rapid urbanization in Nepal. The area of cropland increased from 1995 to 2001, with some studies pointing to an expansion of cropland during this period [29], and a significant decrease between 2001 and 2020 associated with rapid urbanization [30].

The largest relative rate of change was in construction land (Figure 8b), which was related to the relatively small area itself. Due to their large original area base, coniferous forest, broadleaf forest, and cropland had a smaller rate of change in the area but a larger absolute change. The absolute amount and relative rate of change in the area of grassland, wetland waters, snow and ice, and bare ground were insignificant.

The degree of construction land was the largest in all periods, the degree of shrub was the smallest, and all other land types were less than 1%. Among them, the dynamic degree of construction land was positive in both periods and was greater after 2001 than before, indicating that urbanization in Nepal has been accelerating in recent years. The dynamic degree of coniferous and broadleaf forests was > 0 from 2001 to 2020, indicating that the area of coniferous and broadleaf forests increased faster in recent years compared to 1995–2001 (Figure 8c).

3.2. Analysis of Spatial Agglomeration of Land Use/Cover in Nepal

3.2.1. Land Fragmentation

From 1995 to 2001 (

Table 7. Land fragmentation index of Nepal from 1995 to 2020.

Land Cover Types	NP 1			PD			LPI			DIVISION
	1995	2001	2020	1995	2001	2020	1995	2001	2020	1995	2001	2020
CF	7973	8138	8621	0.50	0.51	0.54	1.49	1.49	1.87	1.00	1.00	1.00
BF	9472	10,262	11,129	0.60	0.65	0.70	8.51	6.35	4.06	0.99	1.00	1.00
MF	6011	6067	6262	0.38	0.38	0.40	0.00	0.00	0.01	1.00	1.00	1.00
SH	1254	1490	1386	0.08	0.09	0.09	0.01	0.01	0.00	1.00	1.00	1.00
GL	4358	4323	4553	0.28	0.27	0.29	12.75	12.78	12.68	0.98	0.98	0.98
CL	10,154	8866	9774	0.64	0.56	0.62	8.44	8.63	8.48	0.99	0.99	0.99
WW	387	387	391	0.02	0.02	0.02	0.06	0.06	0.06	1.00	1.00	1.00
UB	214	252	411	0.01	0.02	0.03	0.05	0.08	0.13	1.00	1.00	1.00
IS	1015	1015	1015	0.06	0.06	0.06	0.35	0.35	0.35	1.00	1.00	1.00
BL	2117	2108	2094	0.13	0.13	0.13	0.11	0.12	0.12	1.00	1.00	1.00

¹ NP: the number of patches; PD: patch density; LPI: maximum patch index; DIVISION: landscape separation index.

), the number of patches (NP) of forest, shrub, and construction land in Nepal increased, with broadleaf forest and shrub increasing the most and cropland decreasing the most. The patch density (PD) of forest and shrub also increased, and fragmentation was obvious, and the cropland decreased, while the other land types did not change much. The maximum patch index (LPI) of broadleaf forest decreased significantly, indicating that the proportion of the maximum patch to the total area was decreasing and the concentration of the patch was significantly worse. The other land cover types did not change significantly, indicating that the maximum patch in other land cover types in Nepal maintained its original integrity. The landscape separation index (DIVISION) of all land types tends to be 1, indicating that Nepal’s overall land use/cover fragmentation is high.

From 2001 to 2020 (Table 7), the NP of coniferous forest, broadleaf forest, grassland, and construction land patches all increased, the NP of cultivated land turned to an increasing trend compared to the previous period, and shrub and bare land NP values slowly decreased. The PD of cropland and construction land increased and became more fragmented, while the other types of land were relatively stable. The LPI of the broadleaf forest still decreased significantly. The DIVISION of each type of land in this period tended to be 1 compared to 1995–2001, indicating that the fragmentation of land use/cover in Nepal was getting worse during the period from 2001 to 2020.

3.2.2. Spatial Aggregation Analysis

The changes of cold and hot spots in Nepal during the period from 1995 to 2020 were dominated by coniferous forest, broadleaf forest, cropland, and construction land, with significant spatial aggregation of forest and cropland changes occurring mainly in the hills, and spatial dispersion occurring mostly in the plains and mountains. The area of coniferous forest aggregation first decreased and then increased (Figure 9, CF). The changes mainly occurred near Kalikot and Rasuwa, indicating that the coniferous forest here was protected and changed from dispersion to aggregation after 2001. The spatially significant aggregation of broadleaf forest in the southwestern plains decreased and then recovered (Figure 9, BF). The broadleaf forests in the area had been protected since 2001 with a significant concentration. The spatial dispersion area was concentrated in Surkhet city and the south-central plain, and became larger than the area in 1995, indicating that the spatial dispersion of broadleaf forest in the region was serious. The hot spot change area of cropland was the largest (Figure 9, CL); the spatial aggregation area of cropland in the southwestern plain decreased first and then increased. The spatial aggregation area of the eastern plain increased significantly, indicating that cropland became aggregated after 2001 and crops were planted intensively. The spatial aggregation area of construction land increased and then decreased (Figure 9, UB). The aggregation area around Kathmandu continued to increase between 1995 and 2001 and became larger but turned spatially insignificant from 2001 to 2020, indicating urban expansion here but the loss of spatial aggregation characteristics by 2020. The agglomeration area near Kaski continued to increase, and a significant increase in the number of small urban agglomerations, indicating the rapid urbanization of Nepal.

3.3. Analysis of Land Use/Cover Drivers in Nepal

3.3.1. Principal Component Analysis

According to the principles and requirements of PCA, the raw data of the 10 selected drivers (Table 4) were standardized (Z-score) to remove the correlation between the factors. From the analysis results, it could be seen that KMO = 0.72 > 0.70, and Bartlett’s sphericity test significance was 0, indicating that the PCA was great. Based on the eigenvalues, contribution rates, and cumulative contribution rates of each factor, it could be seen that the top three factors made the percentage of variance > 85%, which could reflect the driving force of LUCC in Nepal comprehensively.

The rotated three factors represented the combined information of 10 driving factors, as shown in

Table 8. Principal component matrix after rotation.

Index	F1	F2	F3
X1	0.98	−0.08	0.02
X2	0.98	0.05	0.08
X3	−0.05	−0.03	0.98
X4	0.95	0.09	0.09
X5	−0.97	−0.03	−0.00
X6	−0.97	0.09	0.12
X7	0.94	−0.16	−0.13
X8	0.07	0.95	0.01
X9	0.42	−0.33	0.13
X10	−0.92	0.02	0.20

. The first principal component represented the population, value added of each industry, and community forest indicators, which could be summarized as the role of population, economic, and community forest factors, accounting for 66.36% of the original variance. The population and GDP drove LUCC positively the most, and agriculture, forestry, fisheries, and industrial value added drove LUCC negatively the most. The second principal component was the annual precipitation and temperature, representing the influence of meteorological factors on land use change, accounting for 10.85%, where precipitation exhibited a positive driving effect on LUCC and temperature exhibited a negative effect. The third principal component was the GDP growth rate, which reflected the role of economic growth and accounts for 10.25%, positively driving LUCC. This result indicated that human activities played a greater role in LUCC than natural factors and the factors were not isolated but were mutually constraining and reinforcing, jointly driving the land use/cover change process [52,54].

3.3.2. Regression Analysis of Forest Drivers

The area of coniferous and broadleaf forests in Nepal in each year after standardization was used as the dependent variable and the extracted score coefficients of the principal components in each year were used as the independent variables for linear regression. The results are shown in

Table 9. Principal component regression coefficients.

Land Use/Cover Types	Principal Component	Nonstandardized Coefficient
CF	F1 1	0.19
	F2	0.23
	F3	0.20
BF	F1	−0.71
	F2	0.25
	F3	0.09

¹ NP: F1 represented the influence of population, value added of each industry, and community forest indicators, F2 represented the influence of meteorological factors, and F3 represented the influence of GDP growth rate.

, it can be seen that the area of coniferous forest was positively correlated with the three principal components, and the area of broadleaf forest was significantly negatively correlated with the first principal component and positively correlated with the second and third principal components. This result indicates that population size, economic factors, and community forestry had a greater influence on the change in the forest area.

Since linear regression using the results of the PCA can only explain the relationship between the extracted principal components and the influence of land use change, it has some limitations [5]. The regression equations expressed by the standardized independent variables (drivers) were constructed from a matrix consisting of the three principal components’ driving factor loading coefficients and the principal components’ linear regression coefficients, using the area of coniferous and broadleaf forests as the dependent variable. The regression equation of coniferous forest area change expressed by the driving factors is shown in Equation (5),

$$Y=0.04\times {\mathrm{X}}_1+0.09\times {\mathrm{X}}_2+0.19\times {\mathrm{X}}_3+0.10\times {\mathrm{X}}_4-0.07\times {\mathrm{X}}_5-0.01\times {\mathrm{X}}_6-0.001\times {\mathrm{X}}_7+0.23\times {\mathrm{X}}_8-0.02\times {\mathrm{X}}_9-0.01\times {\mathrm{X}}_{10}$$

(5)

Similarly, the regression equation for broadleaf forest area change is shown in Equation (6),

$$Y=-0.11\times {\mathrm{X}}_1-0.19\times {\mathrm{X}}_2+0.06\times {\mathrm{X}}_3-0.07\times {\mathrm{X}}_4+0.09\times {\mathrm{X}}_5+0.12\times {\mathrm{X}}_6-0.13\times {\mathrm{X}}_7+0.2\times {\mathrm{X}}_8-0.10\times {\mathrm{X}}_9+0.11\times {\mathrm{X}}_{10}$$

(6)

From Equation (5), it can be seen that population indicators (X₁), GDP indicators (X₂, X₃, X₄), and average annual precipitation (X₈) had positive effects on coniferous forest area, while other factors showed negative effects. From Equation (6), except for GDP growth rate (X₃), value added in the agriculture, forestry, fishery, and industry (X₅, X₆), annual precipitation (X₈), and annual (X₁₀) had a positive effect on the change of broadleaf forest area; the other factors showed a negative effect.

In general, human activities had more impacts on coniferous and broadleaf forest area change than climate factors. GDP growth and precipitation had positive effects on forest areas, and service industry development and the rising temperature had negative effects. In addition, community forestry development showed positive effects on broadleaf forest area change.

4. Discussion

4.1. Driving Factors Influencing the Forests

Generally, demands for natural resources increase in tandem with local and regional socioeconomic development. Demand pressures are further exacerbated by population growth. Together, these two factors are the main (ultimate) driving forces leading to over-exploitation in Nepal. As can be seen from Figure 4, the third sector (services) increased significantly, which corresponded to the recovery of forest areas since 2001. The value added of agriculture, forestry, and fishing reflected the final results of agricultural production and business activities and their contribution to society, and the fact that forest conservation implied a reduction in the economic value generated by local people by obtaining income activities such as timber.

The change in GDP indicators (Figure 5a) showed an increasing trend in GDP and an overall increase in GDP growth. An increase in GDP could lead to an increase in inputs for community forestry. Meanwhile, population growth (Figure 5b) could result in increased use of forest products and increased pressure on forest resources [55]. In Nepal, both precipitation and temperature have increased in the past decades. For coniferous and broadleaf forests, increased precipitation was a positive effect, while the increased temperature was not conducive to forest area change. Sigdel pointed out that the warming and drying trend in winter and spring may reduce forest growth [56].

Reducing emissions from deforestation and forest degradation is one of the main mechanisms for addressing climate change, biodiversity loss, and improving livelihoods [57]. Deforestation in Nepal not only threatens the functioning of complex socioeconomic systems but also affects biodiversity and other ecosystem services, with two-thirds of Nepal’s population dependent on forests for their livelihoods [58]. There is potential for a win-win situation between forest conservation and rural development, but attention needs to be paid to the challenges of future population growth and increased demand for cropland on forest conservation [59].

Cropland in Nepal was in a state of growth due to agricultural land reclamation and deforestation in the early years. Later on, the area of cropland was reduced due to the expansion of construction land and forest land protection policies on the one hand, and on the other hand, the reduction of the labor force, geological hazards, and drought may also be the main factors for the reduction of cropland [52]. Especially with the epidemic in recent years, there is an urgent need to improve people’s livelihoods. The revitalization of fallow land and low-slope potential agricultural areas have been proposed to address livelihood issues [60].

The state of expansion of construction land in Nepal is dominated by the sacrifice of cropland, and high-density human activities are rapidly changing urban land use/cover patterns. Sustainable urban land use management is urgently needed to balance the conflict between ecological conservation and urbanization [30,61].

4.2. Community Forestry of Nepal

Community forestry in Nepal has shown promise for reducing rural poverty, improving reforestation, and potentially offsetting carbon emissions [62]. Previous research showed that the recovery of forest cover in Nepal has been attributed to community forestry policies, which have successfully restored predominantly degraded forest land from hilly areas through forest restoration initiatives [63,64]. For example, in the mid-mountain regions of Nepal and India, the most densely populated areas of the Himalaya, the area and quality of forests increased significantly [62]. More specifically, about 23,000 hectares of degraded pasture and shrubland have been converted to evergreen plantations, and there was also an increase in on-farm tree planting and natural regeneration on abandoned land from the early 1980s to 2000 [65].

Our results are similar to previous research findings and found that community forestry has contributed positively to reducing deforestation and increasing tree cover. Since the promulgation of the Master Plan for Forestry Sector (MPFS) in Nepal in 1989, forest cover has continued to decline and then recovered [66]. From the number of newly added community forestry from 1995 to 2013 (Figure 3), we found that the amount of community forestry continuously increased since 1995, which has led to the restoration of forest areas gradually. However, community forestry also faces issues such as management practices and fairness in benefit sharing [64,67]. The population of Nepal is concentrated in the plains and hilly areas, and most rural farmers rely on forests for fuelwood and fodder for their survival, and the population demand as well as the expansion of agriculture have caused a significant reduction in broadleaf forest area [63,68]. Community forestry is oriented towards achieving sustainable forest management. At the same time, it needs to improve the planning and management of forest product diversification and marketing. Furthermore, there needs to be a coordinated effort among stakeholders to involve local communities in the policy process. Ensuring that community forestry reaches its full potential to achieve a win-win for both forest protection and livelihood support [69,70].

4.3. Possible Factors Affecting Land Fragmentation

A previous study indicated that the potential mechanisms leading to forest fragmentation are different at various altitudes. At lower elevations, human activities often caused fragmentation; while at higher altitudes, fragmentation could be attributed to natural causes [21]. Population growth is one of the main factors in land fragmentation, and the policies and legal provisions adopted to control land fragmentation are also important [71]. Singh Air indicated that in Nepal, factors that drive parcel fragmentation were cultural, social, legal, economic, frequent disasters, geographic variations, unmanaged migrations, and haphazard land use planning, amongst other things [72].

Generally, uncontrolled and unmanaged parcel fragmentation in Nepal is the major challenge for land use planning and its implementation. India and Nepal have made some efforts to promote land consolidation, such as land consolidation schemes, and cooperative agriculture projects, in addition to enacting laws to control land fragmentation [71]. This study showed that land fragmentation in Nepal was severe and exhibited spatial aggregation characteristics. The analysis of the landscape fragmentation index and spatial aggregation in this study provided some insight into the fragmentation pattern of land cover in Nepal, and its influencing factors were also worth exploring. There are some limitations to the data used in the paper. Future analysis will be further validated with field investigation data.

5. Conclusions

This paper used ESA/CCI data to extract land use/cover information for Nepal from 1995 to 2020, and analyzed the direction, magnitude, and rate of land use/cover change, as well as fragmentation patterns and spatial autocorrelation characteristics using the land transfer matrix, land use type dynamic degree, and landscape fragmentation index, and counted 10 driving factors to explore the causes of land use/cover change in Nepal from 1995 to 2020.

In Nepal, the land use/cover shifted mainly between forest, shrub, grassland, and cropland during the period from 1995 to 2020, with forest shifting mainly with cropland, and grassland, wetland waters, snow/ice, and bare land being more stable. The land fragmentation in Nepal was serious, and it was more serious in 2001–2020 than in 1995–2001. The hotspot changes in Nepal during the period from 1995 to 2020 were dominated by coniferous forest, broadleaf forest, cropland, and construction land, with spatial aggregation of each type in most areas and spatial dispersion in plains and mountains. We found that LUCC in Nepal, as elsewhere, was caused by a combination of natural drivers such as climate change and human activity such as land conversion. Specifically, the GDP growth, precipitation change, and community forestry development had positive effects on forest areas, while service sector development and temperature change had negative effects. In addition, community forestry development showed positive effects on broadleaf forest area change. We conclude that it is critical to balance livelihood improvement and forest conservation in the context of increasing population trends. Future land use activities are likely to further expand and intensify to meet the growing demand for food, materials, and energy. Therefore, more rational and comprehensive sustainable land management is needed in Nepal. The results of this study provided a reference for the government to implement appropriate land use policies, community forestry management, land consolidation in agricultural areas, and reducing parcel fragmentation in urban and urban–rural interface areas.