Introduction

In recent years, with the continuous growth of China’s economy and population, limited natural resources and the environment have become insufficient to meet the increasing material demands of human society1. Excessive human activities have altered original land use patterns, disrupted ecological balance, and triggered a series of environmental problems2,3. As a result, ecological protection, the rational utilization of land resources, and long-term monitoring of land use types have become key research priorities4,5. Land use/land cover change is the most direct manifestation of human impact on the natural environment6, representing the outcome of human intervention in ecosystems7. Analyzing land use changes helps reveal the patterns of land use in different regions and provides scientific evidence and theoretical support for subsequent monitoring and studies of various land use types8,9.

Among these land use types, orchards occupy a crucial position10. Orchard cultivation not only plays a significant role in ecological protection, natural resource management, and human health assessment, but also serves as an important means for rural poverty alleviation and economic development11,12. Long-term monitoring of orchard area changes and providing scientifically grounded recommendations can promote the sustainable development of the orchard industry, improve orchard ecosystems, ensure fruit quality and safety, enhance overall orchard productivity, and drive regional economic growth13.

Research on orchard area changes is of great importance for maintaining and safeguarding regional ecological environments and has drawn wide attention from both domestic and international scholars14. Traditional methods of monitoring orchard area changes rely primarily on field surveys15. However, such methods are time-consuming, unsuitable for large-scale studies, and require substantial human, material, and financial resources, resulting in high costs and low efficiency16. At present, remote sensing (RS) technology has gradually become a key method for dynamically monitoring orchard planting areas due to its advantages of speed, accuracy, broad coverage, convenience, low cost, rapid data updates, strong timeliness, and high utilization value.

Some researchers have already applied remote sensing technology to orchard information extraction. For example, Yuan et al.17 synthesized multi-temporal remote sensing images and used object-oriented classification methods to extract and map orchard types in the study area. Zhang et al.18 conducted preprocessing on Landsat remote sensing images and established orchard information extraction models using NDVI and DEM data to analyze spatial and temporal patterns of orchard area changes. Recio et al.19 employed the K-means method, a form of unsupervised classification, to extract orchard areas and analyzed changes in orchard extent, type conversion, and spatial dynamics. Additionally, integrating RS with GIS allows for precise crop identification and facilitates the analysis of spatial distribution changes. For instance, Rahman et al.20 used a spatial analysis model combining RS and GIS to analyze cropland area and spatial dynamics in the Andes. Pan et al.21 used GIS and multiple regression models to analyze spatiotemporal changes in crop areas such as rice in Chongqing over the past 50 years. Huang et al.22 used remote sensing to monitor changes in winter wheat growth in their study area. Most of these studies focus on single temporal snapshots of remote sensing imagery, and few analyze orchard area changes specifically from a land use perspective. Nevertheless, they provide valuable references for conducting long-term continuous sequence analysis of orchard area changes.

Most long-term studies on cropland area changes are based on single-time-point remote sensing images23. However, monitoring based on such typical time slices can lead to the omission of key information and fail to capture the objective trends within continuous time series, resulting in an incomplete and inaccurate reflection of cropland area changes in the study region24. Therefore, this study aims to investigate orchard changes over a continuous 30-year time series, analyzing changes in orchard area, land use type conversions, and spatial dynamics.

The Alar (also transliterated Alaer) reclamation area is a major production base in Xinjiang for fruits such as jujube, apple, fragrant pear, grape, peach, apricot, and pomegranate25. Alar Reclamation Area, as part of the Xinjiang Production and Construction Corps, plays a crucial role in agricultural production. Its oasis agriculture system is highly sensitive to water diversion from the Tarim River. In recent years, with the increasing population in the reclamation area, large-scale orchard development has been undertaken to meet living needs and generate economic benefits. According to data from the Xinjiang Production and Construction Corps Statistical Yearbook (1990–2019), annual fruit production in the area has increased from 24,000 tons in 1990 to 1.6029 million tons in 2019. A large amount of reserve land has been converted into orchards, which has alleviated some of the pressures brought by population growth. However, it has also led to a series of ecological and environmental problems. This is especially true in the arid northwest region, where Alar is located. The region faces severe soil salinization, and orchard cultivation requires large volumes of fresh water for leaching. As the area’s water supply primarily depends on meltwater from the Tianshan Mountains, freshwater resources are limited. The continuous expansion of orchard areas has intensified water-use conflicts and exacerbated the human–land contradiction, which is detrimental to long-term sustainable development. Therefore, it is necessary to use scientific methods to accurately monitor dynamic changes in orchard area over continuous long-term sequences, understand the conversion processes among land use types, analyze spatial dynamics of orchards, and develop sound protection strategies. Such efforts are of great significance for the sustainable development of the fruit industry and agriculture at large.

This study takes the Alar reclamation area—a typical artificial oasis in the arid northwest region—as a case study. A total of 30 Landsat images from 1990 to 2019 that meet mapping requirements were selected for preprocessing. Based on actual land use conditions, an optimal classification system and interpretation features were developed. Supervised classification was performed using the Support Vector Machine (SVM) method, and a confusion matrix was constructed to evaluate classification accuracy. Orchard change analysis was then carried out on the accurately classified images from the perspectives of area change, type conversion, and spatial dynamics. The results aim to provide scientific support for ecological governance, land resource planning, and macro-level regulation of orchard areas in the reclamation zone.

Overview of the research area

The Alar Reclamation Area is located in the southern region of Xinjiang Uygur Autonomous Region and is administered by the First Division of the Xinjiang Production and Construction Corps (XPCC). Its geographical coordinates range from 80°30′ to 81°58′ E and from 40°22′ to 40°57′ N (Fig. 1). It stretches from the southern foothills of the Tianshan Mountains in the north to the northern edge of the Taklimakan Desert in the south, bordered by Keping County to the west and Shaya County to the east, covering a total area of 4,105.92 km².

The reclamation area contains three major reservoirs—Shengli, Shangyou, and Duolang—and lies adjacent to the Tarim River. The terrain rises slightly on both sides of the river channel, sloping from northwest to southeast, and forms a belt-shaped oasis landscape. The ecosystem in the region is classified as desert–oasis type, and the climate belongs to a warm temperate, extremely arid continental desert climate zone. With abundant solar radiation and thermal resources, the area has great potential for agricultural development. Large-scale cultivation of crops such as jujube, apple, fragrant pear, and grape is commonly practiced in the region14,25,26.

Fig. 1
figure 1

Geographic position of study area. Maps created using ArcGIS 10.2 (https://www.esri.com/).

Full size image

Materials and methods

Acquisition of remote sensing image data

Based on the remote sensing imagery provided by the United States Geological Survey (https://earthexplorer.usgs.gov/), a total of 30 scenes of remote sensing images covering the Alar Reclamation Area were selected for the period from 1990 to 2019. The selected images meet the following criteria: path/row number of 146/32, spatial resolution of 30 m, imaging dates between July and November (the peak vegetation growing season), and cloud cover below 10%. Each image was individually subjected to a series of preprocessing steps, including radiometric calibration, atmospheric correction, image registration, image enhancement, and cropping. Although the temporal range spans June to November, the scenes selected for analysis were during the peak vegetation growing season (July–September) to minimize seasonal variation. The specific acquisition dates and the post-processing status of each image are shown in Fig. 2.

Fig. 2
figure 2

Satellite remote sensing images of Alar reclamation area from 1990 to 2019. Maps created using ArcGIS 10.2 (https://www.esri.com/).

Full size image

Establish a classification system

To accurately distinguish between different land cover types and provide reliable data support for the quantitative analysis of orchard area in the Alar Reclamation Area, this study established a classification system and interpretation criteria tailored to the region. The classification system was developed based on the current land use/land cover status of the Alar Reclamation Area, land use statistics of the First Division from the Statistical Yearbook of the Xinjiang Production and Construction Corps (1990–2019), and relevant literature13.

Using the year 2019 as an example, the land cover in the study area was categorized into six types: orchard, cultivated land, forest and grassland, water bodies, construction land, and unused land. The surface characteristics, image features, and field survey results for each land cover type are summarized in Table 1.

Table 1 Classification system and interpretation signs of Alar reclamation area.
Full size table

Based on the established classification system and interpretation criteria, training samples corresponding to each land cover type were created. The number and spatial distribution of sample points for each category are shown in Fig. 3. The sample points were selected following the principles of uniformity and comprehensive spatial coverage. The final number of pixels for each category is as follows: orchard – 192,516; cultivated land – 128,321; forest and grassland – 119,278; water bodies – 118,261; construction land – 37,668; and unused land – 109,303.

Since the primary focus of this study is the analysis of orchard area change, the orchard category was assigned the largest number of samples to ensure higher classification accuracy. Following the same approach, training samples for each land cover type were constructed individually for the remaining years to perform land cover classification.

Fig. 3
figure 3

Number and distribution of sample points for each category. Maps created using ArcGIS 10.2 (https://www.esri.com/). Note: The number of pixels of each type of sample is in parentheses.

Full size image

Remote sensing image classification methods

Remote sensing image classification is typically conducted by analyzing features such as the tone, shape, spatial distribution, and differences in vegetation growth of various land cover types in the imagery, combined with field survey knowledge of the study area27. Based on the actual surface conditions of the study area, statistical data on land cover types from the statistical yearbooks, and classification methods adopted in relevant case studies, a comprehensive analysis was conducted28. As a result, the Support Vector Machine (SVM) method29, a supervised classification technique, was ultimately selected for image classification30. SVM was chosen for its robustness in high-dimensional feature spaces, particularly when sample sizes are small. Recent studies have demonstrated its effectiveness in similar applications31.

User Accuracy (UA) refers to the probability that a pixel classified as a certain land cover type actually represents that type on the ground. This metric provides insights into how reliable the classification is from the perspective of the user, i.e., how often the classified pixel truly matches the land cover type in reality. Producer Accuracy (PA), on the other hand, reflects the likelihood that a pixel on the ground, which actually represents a specific land cover type, is correctly classified. It measures the classification’s reliability from the producer’s perspective, i.e., how often the correct land cover type is classified by the algorithm. Both UA and PA are crucial for evaluating classification accuracy and ensuring the results’ reliability.

To validate the classification results, a confusion matrix was constructed for the year 2019, which compares the classified land types with ground truth data. The overall accuracy of the classification was calculated, and the confusion matrix was used to compute both User Accuracy (UA) and Producer Accuracy (PA) for each land cover type.

One limitation of the SVM classification method is that it is highly sensitive to the quality and quantity of training data. SVM can overfit when the number of samples is too small or poorly distributed, leading to inaccurate classification results, especially for heterogeneous landscapes. Additionally, the performance of the SVM classifier depends on the proper selection of kernel functions and tuning parameters. In areas with high spectral variability, SVM may struggle to distinguish between similar land cover types, leading to lower classification accuracy.

Orchard area change analysis methods

The characteristics of orchard change from 1990 to 2019 were analyzed from three main aspects: area change, land type conversion, and spatial dynamic change (The former refers to land cover transformations, while the latter describes the spatial redistribution of orchards over time.). Area change refers to the variation in the total area of orchard land from the initial year to the final year, as well as the average annual change rate during different time periods. Land type conversion and spatial dynamic change refer to the amount of area converted between orchards and other land cover types, as well as the spatial redistribution patterns over time.

The spatial statistical tools in ENVI and GIS software were employed to generate land use transition matrices and spatial dynamic change maps. These tools provided a visual representation of orchard area inflows and outflows, effectively illustrating the trends in orchard expansion and reduction across the study period.

Results

Classification results and accuracy assessment

After establishing classification samples for each of the 30 remote sensing images from 1990 to 2019, the Support Vector Machine (SVM) method was employed to perform land use/land cover classification for each year. The spatial distribution of different land use types for each year is shown in Fig. 4. As illustrated in the figure, there has been a noticeable increase in vegetation-covered areas across the Alar Reclamation Area over the past three decades, while the area of unused land has exhibited a continuous decreasing trend.

Fig. 4
figure 4

Land use/cover in Alar reclamation area from 1990 to 2019. Maps created using ArcGIS 10.2 (https://www.esri.com/).

Full size image

To verify the reliability of the land use/land cover classification results, a confusion matrix was constructed for the year 2019 to evaluate the classification accuracy of each land use type. As shown in Table 2, the UA and PA for orchards were the highest among all land categories, reaching 95.28% and 97.69%, respectively, ensuring the high classification accuracy of orchards. Meanwhile, the UA and PA values for all other land types were above 70%. Overall, the Overall Accuracy (OA) for the Alar Reclamation Area reached 87.97%. According to relevant studies32, when the UA, PA, and OA values of all land use types exceed 70%, the classification results can be considered reliable and applicable for further research. Therefore, the classification results in this study are deemed credible. User Accuracy (UA) refers to the probability that a pixel classified as a certain land cover type actually represents that type on the ground. It measures the reliability of the classification from the user’s perspective. Producer Accuracy (PA), on the other hand, indicates the likelihood that a pixel of a specific land cover type on the ground is correctly classified. It measures the reliability of the classification from the producer’s perspective. Both accuracy metrics are crucial for evaluating the overall performance and reliability of remote sensing classification results, providing insights into both the correctness of classifications and the errors associated with them.

Table 2 Evaluation on the precision of land types in Alar reclamation area.
Full size table

Characteristics of orchard area change

The area changes of different land use types in the Alar Reclamation Area over the past 30 years are illustrated in Fig. 5. During this period, orchard area exhibited a continuous increasing trend, expanding from 417.57 km² (10.17%) in 1990 to 1,091.76 km² (26.59%) in 2019. This represents a net increase of 674.19 km², accounting for 16.42% of the total area of the reclamation region. In addition, cultivated land also showed a steady upward trend, with a net increase of 1,147.19 km² over the 30 years. The area of forest and grassland increased slightly, with a net change of 33.67 km². In contrast, the areas of water bodies and construction land remained relatively stable but showed slight growth, with net increases of 107.58 km² and 36.54 km², respectively. Additionally, the increase in water bodies is primarily due to the development of irrigation to support the growing orchards and cultivated land. The most significant change occurred in unused land, which experienced a sharp decline of 1,931.83 km². This suggests that the expansion of orchard area primarily came from the reclamation and conversion of unused land. Additionally, the evolution of cultivated land is worth noting. Initially, in 1990, the area of cultivated land was smaller than that of orchards. Over time, both the areas of cultivated land and orchards have continuously increased, but the area of cultivated land surpassed that of orchards in 2017.

Fig. 5
figure 5

The area change of land types in Alar Reclamation area from 1990 to 2019.

Full size image

The annual average growth rate of orchard areas varies across different time periods. The annual average growth rate of each land use type from 1990 to 2019 is shown in Fig. 6. The periods are divided into six intervals with a 5-year gap:

(1) From 1990 to 1995, the annual average growth rate of orchards was the highest, at 6.04%. The area increased from 417.57 km² to 563.33 km². This was due to the smaller initial orchard area and the large development potential, resulting in the highest annual growth rate. Additionally, cultivated land also experienced significant changes during this period, as evident from the graph where the change in cultivated land was the highest from 1990 to 1995.

(2) From 1995 to 2000, following the large-scale orchard reclamation in the previous period, the development intensity of orchards decreased, and the annual average growth rate dropped to 2.58%. Unused land showed a negative growth rate during this period, indicating that humans had already started reclaiming wasteland thirty years ago.

(3) From 2000 to 2005, the annual average growth rate of orchards began to rise again, reaching 5.50%, with the area increasing from 636.01 km² to 810.92 km². Moreover, forest and grassland areas, as well as water bodies, first showed negative growth rates during this period, indicating that the increase in orchards and cultivated land came at the cost of forest, grassland, and water bodies.

(4) From 2005 to 2010, the annual average growth rate of orchards decreased again to 2.41%.

(5) From 2010 to 2015, the annual growth rate of orchards slightly increased to 3.05%. During this period, the negative growth of forest and grassland areas, as well as water bodies, turned positive, signaling that the government had started to recognize the worsening ecological environment and began to protect forest, grassland, and water bodies.

(6) From 2015 to 2019, the annual average growth rate of orchards decreased to the lowest, at 1.06%, although the area increased from 1047.42 km² to 1091.76 km². During this period, the increase in construction land was significant, indicating that the expansion of human activities was intensifying.

Fig. 6
figure 6

Average annual change rate of land use types in Alar reclamation area from 1990 to 2019.

Full size image

Characteristics of orchard type conversion

The land use transfer between different categories in the Alar reclamation area from 1990 to 2019 is shown in Table 3. During this period, the total area conversion from orchards reached 860.38 km², while the area converted to orchards amounted to 1,534.58 km², resulting in a net increase of 674.2 km² in orchard area.

Moreover, significant transitions occurred between orchard and non-orchard land types. The areas converted from orchards to other land use types, in descending order, are: cultivated land (832.59 km²) > water bodies (15.84 km²) > construction land (11.59 km²) > forest/grassland (0.36 km²) > unused land (0.00 km²). Conversely, the areas converted to orchards from other land types are ranked as follows: unused land (1,160.44 km²) > cultivated land (349.64 km²) > water bodies (15.69 km²) > forest/grassland (8.81 km²) > construction land (0.00 km²).

These results indicate that over the past 30 years, most newly developed orchards were converted from unused land, while the primary type converted out of orchards was cultivated land.

Table 3 The transformation matrix of land use types in Alar reclamation area from 1990 to 2019/km².
Full size table

Spatial dynamics of orchard changes

From 1990 to 2019, the spatial dynamic changes of orchards in the Alar reclamation area were analyzed across six distinct time periods (Fig. 7):

(1) From 1990 to 1995, the number of orchard area transfers in and out was relatively balanced, with the main spatial distribution along the Tarim River region.

(2) From 1995 to 2000, the transfer of orchard areas into the region increased, especially in the southeastern part of the Tarim River basin, where large-scale orchard reclamation occurred.

(3) From 2000 to 2005, the transfer of orchard areas out of the region also showed an increasing trend, with the spatial distribution mainly along the Tarim River region.

(4) From 2005 to 2010, both the transfer in and transfer out of orchard areas increased simultaneously. The transfer in occurred mainly in the northwestern part of the study area, while the transfer out was concentrated near the Tarim River basin.

(5) From 2010 to 2015, the transfer out of orchard areas, represented by the red areas, significantly increased, primarily in the unused land reclamation areas in the northwest of the reclamation zone. Meanwhile, the transfer in of orchard areas, represented by the green areas, was relatively small and mainly distributed along the Tarim River.

(6) From 2015 to 2019, the transfer in of orchard areas exceeded the transfer out. The transfer in primarily involved the reclamation of unused land in remote areas, while the transfer out was mainly concentrated along the Tarim River region, where the land was transformed into arable land.

Previous studies on land use changes in arid regions, such as those by Ana et al.33 and Solomon et al.34, have reported significant increases in orchard and cultivated land areas. Yaxuan et al.35 observed a moderate increase in orchard areas over the past few decades, with the rate of expansion slowing due to water constraints. In contrast, our study demonstrates a much higher rate of orchard expansion in Alar, likely due to substantial improvements in irrigation systems, which have facilitated the growth of both orchards and cultivated land.

Kayitesi et al.36 noted that in other regions of Xinjiang, the expansion of cultivated land has slowed in recent years due to water shortages. However, our results indicate that in Alar, despite the challenges posed by water scarcity, cultivated land has expanded more rapidly, surpassing orchard areas by 2017. This difference may be explained by the development of new irrigation infrastructure and the availability of groundwater in the region, factors that have not been as prominent in Manikandababu et al.‘s study15.

These findings underline the role of water management practices in driving land use changes in Xinjiang, and suggest that the pace of agricultural expansion is closely tied to local irrigation capabilities.

Fig. 7
figure 7

Orchard spatial dynamic change map from 1990 to 2019. Maps created using ArcGIS 10.2 (https://www.esri.com/).

Full size image

Discussion

In this study, we observed significant changes in orchard area over the 30-year period, with an increasing trend in orchard expansion from 1990 to 2019. These results are consistent with previous studies, such as those by Ana et al.33 and Solomon et al.34, who reported similar trends of orchard area increase in arid regions due to advancements in irrigation and agricultural development. However, our study demonstrates a much higher rate of orchard expansion in the Alar region compared to other areas, likely due to substantial improvements in irrigation systems that have allowed for the rapid development of orchards and cultivated land.

Previous research has shown that orchard growth in arid regions is often limited by water constraints, as discussed by Yaxuan et al.35 in their study of orchard expansion in Xinjiang. In contrast, the Alar region’s orchard expansion appears to be less constrained by water shortages, possibly due to better water management practices and the availability of groundwater resources. This difference is noteworthy, as it suggests that improved irrigation infrastructure and water resource management can significantly impact land use changes in water-scarce regions.

Additionally, Kayitesi et al.36 found that land use expansion in other regions of Xinjiang has slowed due to water scarcity. In Alar, however, the expansion of cultivated land has outpaced orchard growth, which may be attributed to the region’s recent investments in irrigation infrastructure and new water resources that were not as readily available in other parts of Xinjiang. The spatial distribution of orchard area changes also aligns with findings from previous studies that highlight the role of water resources in shaping land use dynamics37.

Conclusion

This study selected a total of 30 scenes of Landsat remote sensing imagery from 1990 to 2019 for the Alar reclamation area. Based on the current land use/land cover status, relevant statistical data from the Statistical Yearbook of the Xinjiang Production and Construction Corps (1990–2019), a classification system and interpretation criteria were developed. The Support Vector Machine (SVM) method was employed for annual land use classification, and the classification results were evaluated for accuracy. Orchard change characteristics were then analyzed from three main aspects: area change, land type conversion, and spatial dynamics. The main conclusions are as follows:

(1) The UA and PA for orchards reached 95.28 and 97.69 respectively, while the UA and PA values for all other land use types were above 70. Overall, the classification achieved an Overall Accuracy (OA) of 87.97, indicating that the classification results meet the requirements for subsequent mapping tasks.

(2) From 1990 to 2019, the orchard area in the Alar reclamation area showed a continuous increasing trend—from 417.57 km² (10.17%) in 1990 to 1,091.76 km² (26.59%) in 2019. The net increase accounted for 16.42% of the total area of the reclamation zone. The average annual rate of orchard expansion varied across different periods: it was highest (6.04%) from 1990 to 1995, and lowest (1.06%) from 2015 to 2019. The expansion of orchards has exacerbated soil salinization and increased pressure on freshwater resources, which poses significant risks to the sustainability of the oasis.

(3) Over the 30-year period, a total of 860.38 km² of orchard area was converted to other land types, while 1,534.58 km² was converted into orchard land, resulting in a net gain of 674.2 km². Most of the orchard losses were due to conversion into cultivated land (832.59 km²), primarily along the Tarim River. Conversely, most of the orchard expansion came from the reclamation of previously unused land (1,160.44 km²), mainly located in the northwestern part of the reclamation area. The expansion of orchards must be balanced with the limitations of water resources and the ecological carrying capacity. Policies should promote sustainable land management practices to reconcile economic growth and environmental conservation.