Next Article in Journal
Pedogenesis of Fluvial Terrace Soils Related to Geomorphic Processes in Central Taiwan
Previous Article in Journal
Assessment of Water Yield and Water Purification Services in the Arid Zone of Northwest China: The Case of the Ebinur Lake Basin
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Land
Volume 12
Issue 3
10.3390/land12030534
land-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

An Alternative Method of Cultivated Land Identification and Its Actual Change from 2009 to 2019: A Case Study of Gaochun, China

by
Zhuoran Jiang
1,
Ming Jiang
1,
Yahua Wang
1,2,*,
Can Ma
1 and
Weifeng Qiao
1,2
1
School of Geography, Nanjing Normal University, Nanjing 210023, China
2
Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China
*
Author to whom correspondence should be addressed.
Land 2023, 12(3), 534; https://doi.org/10.3390/land12030534
Submission received: 20 December 2022 / Revised: 18 February 2023 / Accepted: 20 February 2023 / Published: 22 February 2023
(This article belongs to the Section Land – Observation and Monitoring)
Browse Figures
Versions Notes

Abstract

:
As the largest developing country, China has permanently attached great importance to cultivated land protection. However, due to the different rules of cultivated land identification in the second and third national land surveys, the cultivated land area in the two surveys has changed greatly. Some agricultural lands in the south, such as plantations, forests, grasslands, aquaculture ponds, etc., belonged to cultivated land during the second survey, but they were identified as non-cultivated land in the third national land survey. This change has led to a sharp reduction in the area of cultivated land in some places. In order to calculate the actual change in the area of cultivated land since the second survey and provide a reasonable basis for the standard of cultivated land protection, this paper takes Gaochun District, a developed area in China, as an example; interprets the images of the second national land survey period with the deep learning network HRNet; and compares the results with the second and third national land survey rules. The results show that the actual reduction of cultivated land in Gaochun District in the past ten years accounts for 35.1% of the reduction of cultivated land in the two land surveys, while the reduction of cultivated land caused by the change of cultivated land identification rules accounts for 64.9% of the reduction of cultivated land in the two land surveys, indicating that the significant reduction in local cultivated land was mainly caused by the changes in the rules, and these cultivated land reduction behaviors existed before the second survey.

1. Introduction

Cultivated land is the carrier of sustainable agricultural development, the fundamental natural resource of human survival and development, and the fundamental lifeblood of socially sustainable development [1,2,3]. China is the largest developing country in the world, which has an exceptional national condition of a large population but a small land and scarce cultivated land resources. The protection of cultivated land has always been highly valued [4,5]. In recent years, due to the restructuring of agricultural cultivation, the area of cultivated land in China has continued to decrease; more than half of the cultivated land in developed areas has even been left uncultivated, and half of the farmers do not cultivate their fields [6]. This kind of occupation of cultivated land for planting other cash crops is called “non-grain-oriented” behavior. Although the non-grain-oriented behavior can increase farmers’ economic income to a certain extent [7,8,9], the transformation from grain planting to non-grain-oriented production may induce the phenomena of soil barrenness, acidification, and degradation [10,11]. For instance, digging ponds to raise fish and crabs will destroy the plough horizon, which requires a large amount of foreign soil to rebuild before planting can be resumed [12]. In order to protect cultivated land, the state has issued many policies and implemented corresponding reward and punishment measures according to the realization of cultivated land protection objectives in various regions: the central government will reward the newly determined provinces whose cultivated land protection objectives are higher than the original ones. For provinces that have not completed the target of cultivated land protection, it is necessary to restore and supplement a part of cultivated land within the planning period with 5 years as a cycle and gradually fill the responsibility gap of the original target of cultivated land protection. For provinces that cannot fill the responsibility gap of the original target of protection within the first 5 years after signing the protection responsibility letter, it is necessary to pay compensation fees according to the standard. However, although these policies are reasonable at the level of cultivated land protection, there are some problems in the implementation process. According to the second national land survey definition, cultivated land is land for planting crops, including cultivated land for temporarily planting medicinal materials, turf, flowers, seedlings, etc., and other cultivated land for a temporary change of use. In the third national land survey, this cultivated land for temporarily planting other crops was classified as restoration of agricultural land, that is, it was no longer used as cultivated land. It can be seen that the definitions of cultivated land in the two national land surveys are not the same, and if the cultivated land areas of the two are directly compared, there will be severe problems. For example, the main agricultural land in a region is aquaculture ponds. These lands were cultivated land in the second survey, but they were restored agricultural land in the third survey. Therefore, when comparing the data of cultivated land in the second and third surveys, we can draw the conclusion that the cultivated land has decreased significantly, even though there has been no change in the agricultural land in this region in the past decade.
Therefore, in order to formulate a reasonable standard of cultivated land protection and explore the actual situation of cultivated land change, we need to reinterpret the images of the second national land survey period based on the third national land survey rules and the developed areas prone to non-grain-oriented behavior, to obtain the actual land type distribution in the second national land survey period under the third national land survey rules. In this way, the land type distribution data of the two periods can be unified to the same standard to obtain the real change of cultivated land in this decade.
In order to explore the change of cultivated land from the above perspective, we need to adopt a reasonable way to interpret the remote sensing images of the second survey period. Presently, domestic and foreign scholars have conducted a lot of research on image segmentation methods. Image segmentation methods can be divided into four categories, namely, the pixel-based method [13,14,15], the machine learning method [16,17,18], the object-oriented method [19,20,21], and the deep learning method [22]. The pixel-based method is an image segmentation method based on image pixels, mainly including the threshold method [23,24,25], spectral analysis method [26,27], edge detection method [28,29], etc. These methods have a good segmentation effect for features with more superficial characteristics but are sensitive to noise, errors, and other factors, with poor robustness and a high false detection rate, and the semantic association between pixels is not considered. Machine learning is an artificial intelligence method involving multiple interdisciplinary fields and is also widely used in image segmentation. Common machine learning methods include support vector machine [30,31], decision tree [32,33], random forest [34,35], extreme learning machine [36], artificial neural network [37], etc. However, the classification samples of machine learning depend heavily on the low-level features designed by hand, and the effect of the machine learning method is equally poor when faced with ground objects with different shapes and complex features. Object-oriented image segmentation methods based on ground objects mainly include level set [38], Markov random field [39], conditional random field [40], etc. Although these methods can extract complete figure information, they require higher image resolution and lower generalization ability. As a popular method in recent years, deep learning is the most effective and widely used method at present, which can learn features independently by convolution neural network [41], cyclic neural network [42], deep neural network [43], and so on, abstracting semantic information from the original remote sensing image layer by layer from low levels to high levels.
Therefore, we used the method of deep learning to interpret the high-resolution remote sensing images of Gaochun District, Nanjing, Jiangsu Province, in 2009 according to the third national land survey classification standard; explore the change of cultivated land in the second national land survey under the third national land survey rule; and analyze its change trend and reasons.
The structure of this paper is arranged as follows: First, in the Introduction section, we sorted out the research on the temporal and spatial evolution of cultivated land and analyzed the shortcomings and defects. Then, in the Methodology section, we will introduce the research area, technical route, and research methods used in this paper. Then, in the Results section, we will present the results of land type interpretation and cultivated land change and compare them with the survey data. Then, in the Discussion section, we will discuss the driving mechanism of cultivated land change, the reasons for this phenomenon, the methods used in this paper, the importance of consistency between survey and planning, and the shortcomings and prospects of the research. Finally, in the Conclusions section, we will summarize the conclusion of this paper.

2. Methodology

2.1. Case

This paper used Gaochun District of Nanjing as the research area. Gaochun District, Nanjing, is located in the marginal area of developed cities. The phenomenon of non-grain-oriented cultivation, dominated by aquaculture ponds, is the main feature of the spatial and temporal evolution of cultivated land in Gaochun District. There are large-scale plain and hilly areas in Gaochun District, which produce different types of non-grain-oriented cultivation, which provides rich research cases for the study of the spatial and temporal evolution of cultivated land.
Gaochun District is located at 31°13′~31°26′ N, 118°41′~119°21′ E, at the border of Jiangsu Province and Anhui Province. Liyang city is in the east, adjacent to Langxi County, Xuanzhou District, and Dangtu county, and Lishui District is in the north. The whole area is 790.23 square kilometers, of which the water area is 191 square kilometers. The plain covers an area of 291 square kilometers, accounting for 51% of the land area; hilly land covers an area of 275.5 square kilometers, accounting for 49%. There are the plains of Shijiu Lake, Gucheng Lake, and the valley floor of Xuxi River. The lake plain is between Gucheng Lake and Shijiu Lake in the region’s west. The plain of the lake area is crisscrossed by rivers and ditches, with dense water networks. The ground elevation is mostly between 4 and 7 m below the flood level. There are low hills around Gucheng Lake in the east, and there is a big difference in the terrain, which forms the terrain of upper hills and lower polders. The ground elevation is 7~50 m, which is characterized by outer flooding and inland waterlogging. The river valley plain is located in the southeast of the region, connected with the west Taowei area of Taihu Lake, and belongs to the Taihu Lake basin. The ground elevation is 5~15 m, and the terrain is flat, with abundant water sources. However, due to the hills around it, the water flow is not smooth, and heavy rain can easily cause disasters. The hills are in the east of the central part of the district, and they are roughly distributed in the southwest and northeast belts and connect at the bottom of the river crossing the whole village. It is bounded by Maodongjin sluice and belongs to two water systems: the Shuiyangjiang River–Qingyijiang River system and Taihu Lake. Gaochun District is located in the southwest of Jiangsu Province, which belongs to the monsoon climate zone in the south of the north subtropical zone. There are four distinct seasons with abundant rainfall, sunshine, and a long frost-free period.
Figure 1 shows an overview of the study area. Figure 1a shows the administrative division of Nanjing. Gaochun District, our research area, is located in the south of Nanjing. It is marked with a red box in the figure. Figure 1b,c show the distribution of cultivated land in Gaochun District of the second and third national land surveys, respectively, and it can be seen that there was a big difference in the distribution. According to the official statistics, the cultivated land area of Gaochun District in the second national land survey was 41,750 hectares, while that of the third national land survey was only 14,410 hectares. Compared with the cultivated land in the second national land survey, the cultivated land in Gaochun District in the third national land survey was obviously reduced by 27,340 hectares, especially in the western part of Gaochun District. Therefore, there may have been a large amount of non-grain-oriented phenomena in Gaochun District of Nanjing.

2.2. Data

The research data of this paper were the 0.3 m high-resolution remote sensing images of Gaochun District in Nanjing in 2009 and the data of the second national land survey and the third national land survey of Gaochun District.
Among them, the definition of cultivated land in the second and third national land surveys changed. In the second national land survey, the cultivated land included land on which crops were grown; land on which crops (including vegetables) were mainly grown, interspersed with scattered fruit trees, mulberry trees, or other trees; and reclaimed beaches and seashores that were guaranteed to be harvested for one season per year on average. Cultivated land included fixed ditches, canals, roads, and ridges with a width of <1.0 m in the south and <2.0 m in the north; cultivated land for temporary planting of medicinal materials, turf, flowers, and seedlings; and other cultivated land for a temporary change of use, covering a wide area. Any agricultural land that had not destroyed the plough horizon was classified as cultivated land.
However, in the third national land survey, the category of cultivated land changed considerably. All land types were investigated according to the actual planting situation, and the land types included in cultivated land were strictly defined as paddy fields, dry land, and irrigated land. The diagram of the change in cultivated land rules is shown in Table 1.
According to the existing LUCC data and socio-economic data, a large-scale adjustment of agricultural structure took place in Gaochun District around the time of the second national land survey. To clarify whether the main reason for the large-scale reduction in cultivated land in Gaochun District in the third national land survey was the different rules for the two national land surveys, it was necessary to go back to the second national land survey images and compare the cultivated land area in Gaochun District under the two national land surveys that established unified rules.

2.3. Methods

2.3.1. Technology Roadmap

The technical route of this paper is shown in Figure 2. First of all, we established a sample database of the cultivated land category changes between the second and third surveys (Figure 2a). We selected the land types that were considered cultivated land in the second survey and tea plantations, forests, land for construction, ponds, and aquaculture ponds in the third survey; analyzed the main changes of cultivated land between the second and third surveys; and established the sample database of cultivated land change in the second survey. Second, we used the sample database to interpret the images during the second survey. Using the “non-grain” sample database of cultivated land to interpret the images in the second survey period, the results of the land classification of the images in the second survey period under the cultivated land identification rules of the third survey (Figure 2b) were obtained, and the phenomenon of “non-grain” occurred in the spot tilling map during the second survey period. Third, we calculated the cultivated land interpreted by the images in the second survey period under the cultivated land identification rules of the third survey. The difference between the total area of cultivated land in the second survey and the total area of the interpreted land category was used to obtain the cultivated land results during the second survey based on the cultivated land identification rules of the third survey (Figure 2c). Then, we compared the interpreted cultivated land with that of the rules of the second survey (Figure 2d) and the rules of the third survey (Figure 2e) and obtained the reduced area of cultivated land caused by the change of the rules (Figure 2f) and the actual reduced area of cultivated land (Figure 2g). Finally, the cause analysis and countermeasures were proposed. We analyzed the actual degree and causes of “non-grain” in the study area and put forward corresponding suggestions.

2.3.2. Deep Learning

The deep learning method is one of the most advanced image interpretation methods at present. It can accurately identify cultivated land parcels, divide cultivated land boundaries, and classify crops through convolution neural networks and attention mechanisms. Compared with traditional algorithms, it has the advantages of high accuracy and high speed. In this paper, the deep learning network HRNet [43] was used as the interpretation network to interpret different land types. HRNet is one of the most accurate image segmentation networks at present. Different from previous segmentation networks, HRNet connects feature layers with different resolutions in parallel. On this basis, it fuses feature layers with different resolutions to obtain more comprehensive feature information. Figure 3 is the schematic diagram of the feature fusion of HRNet. The feature fusion directly copied the feature layer with the same resolution. For feature layers that required resolution enhancement, bilinear upsampling and 1 × 1 convolution were used to unify the fused layers into the same channel, while for feature layers that required resolution reduction, a stride 3 × 3 convolution operation was used. During the feature fusion, different transformed feature layers were fused in addition, and feature layers with multi-resolution information were obtained.

2.3.3. Classification of Interpretation

According to the different definitions of cultivated land in the second and third national land surveys, we analyzed statistics on the parts of cultivated land in Gaochun District that had changed their definitions. The statistical results are shown in Table 2. It was seen that the cultivated land specified in the second survey in Gaochun District was mainly turned into the plantations, forests, land for construction, ponds, and aquaculture ponds in the third survey, and the sum of dozens of other types of land was only 2%. As cultivated land was divided into paddy field, dry land, and irrigated land in the third survey, and as shown in Figure 4, not only were there huge differences in characteristics among the three land types, but also there were different manifestations within the same land type (Figure 4(a1–c3)), so direct training of cultivated land interpretation model may not achieve good results. Therefore, in this study, we ignored the 2% of other types of land, and trained the interpretation models on five land types, namely, tea plantations, forests, land for construction, ponds, and aquaculture ponds. First, we interpreted these five types of land types in the image of the second modulation period and then compared them with the map of the second survey and calculated the distribution of actual cultivated land of the second survey period in the form of “subtraction”. At the same time, according to this step, we could also obtain the specific destination of the cultivated land in the second survey, that is, what type of land the specific land types evolved into during the third survey.

3. Results

3.1. Training Model

In order to interpret the land types, we chose samples of five land types, tea plantations, forests, land for construction, ponds, and aquaculture ponds, and generated corresponding samples (Table 3). The total number of samples of each land type was 6000, with a size of 256 × 256 pixels. We divided these samples into a training set, a verification set, and a test set according to the ratio of 7:2:1, and the numbers were 4200, 1200, and 600, respectively. The training set was used for model training, and the verification set was used to test the model accuracy during model training. To prevent the over-fitting of the model on the verification set from affecting the accuracy evaluation, the final model accuracy was calculated on the test set.
According to these samples, we trained the model of HRNet. Based on the PyTorch deep learning framework, we set up and ran the HRNet deep learning network. All training data and segmentation codes were deployed on a Linux server, which had an Ubuntu16.04.4 system, a Core i7-8700k processor (INTEL CORE, Santa Clara, CA, USA), 32 G memory, and a GTX2080Ti-12G graphics card. In the training process of the model, the optimizer was set as SGD, the initial learning rate was 1 × 10−3, and the number of the epoch was 200, in which the learning rate was adaptive to the number of the epoch, that is, with the increase in iterations, the learning rate gradually decreased and finally returned to 0.
After the training of the model was completed, we evaluated its accuracy. We used the accuracy rate A, precision rate P, recall rate R, and F1 coefficient as the detection accuracy indicators of the preliminary evaluation model.
The accuracy rate is the ratio of the number of correctly segmented samples to all samples, and its calculation formula is as follows:
A = T P + T N T P + T N + F P + F N
In Equation (1), TP is the number of samples with the same detection value as the actual value and a positive detection value, TN is the number of samples with the same detection value as the actual value and a negative detection value, FP is the number of samples with different detection values and a positive detection value, and FN is the number of samples with different detection values and a negative detection value. The higher the value of the precision indicator, the more pixels were correctly segmented and the better the accuracy of the model, while the opposite indicated poor segmentation.
The accuracy rate is the ratio of the number of detected positive samples to the number of actual positive samples, reflecting the ability of the model to “find the right positive samples”, and is calculated as follows:
P = T P T P + F P
The recall rate is the ratio of the number of detected positive samples to the total number of positive samples, which reflects the ability of the model to “find all positive samples”. Its calculation formula is as follows:
R = T P T P + F N
The F1 coefficient is the harmonic average of precision and recall, which is an important index to evaluate the performance of the model. Its calculation formula is as follows:
F 1 = 2 T P 2 T P + F P + F N
Finally, we obtained the accuracy of each model, as shown in Table 4.

3.2. Comparison and Analysis of Data Based on the Second Survey

After obtaining the global interpretation results of the Gaochun area, we generated an intersection with the cultivated land range of the second national land survey and only retained the interpretation results within the cultivated land range of the second national land survey. According to the processed results (Table 5), 849 hectares of tea plantations (Figure 5a), 2154 hectares of forests (Figure 5b), 265 hectares of land for construction (Figure 5c), 1439 hectares of ponds (Figure 5d), and 13,024 hectares of aquaculture ponds (Figure 5e) were interpreted and then differenced with the second national land survey cultivated land plots to obtain the second survey inverse cultivated land plots (Figure 6), with a total area of 24,019 ha. According to the above results, the actual cultivated land area of the second national land survey in Gaochun District accounted for 57.5% of the cultivated land area of the second national land survey statistics. The remaining areas were mainly aquaculture pit ponds, accounting for 31.2% of the cultivated land area of the second national land survey statistics, which were mainly distributed in Zhuanqiang Street, Yangjiang Street, and Chunhua Street, indicating that the aquaculture product (crab) was one of the leading local cash crops.
Table 6 shows the statistics of the amount of cultivated land in Gaochun District by street (town). The statistical items in the table are divided into two parts: cultivated land area and change in cultivated land area. The cultivated land area includes second land, second inversion land, and third land, where the second land is the cultivated land area in the official statistics of the second survey, the second inversion land is the actual cultivated land area of Gaochun District during the second national land survey period calculated in this paper, and the third land is the cultivated land area in the official statistics of the third survey. The cultivated land area changes include the third minus the second land, the third minus the second inversion, and the reduction of the rules. Among them, the third minus the second land is the difference of subtraction between the official cultivated land statistics of the third and second surveys, the third minus the second inversion is the difference of subtraction between the official cultivated land statistics of the third survey and the actual cultivated land area of the second national land survey, and the reduction of rules is the difference of subtraction between the third minus the second land and the third minus the second inversion, which is the reduced cultivated land area due to the change of the second and third survey rules.
According to the results in Table 6, the actual cultivated land area in the second national land survey period was 24,019 hectares, while that in the third national land survey period was 14,410 hectares, so the actual reduced cultivated land area during the two national land surveys period was 9609 hectares, accounting for 40.0% of the actual cultivated land area in the second national land survey period. According to the official data of the second and third national land surveys, the cultivated land decreased by 27,340 hectares during the two national land surveys, which indicated a reduction of 17,731 hectares of cultivated land due to changes in cultivated land rules, accounting for 64.9% of the cultivated land decrease during the two national land surveys. In spatial distribution, the decrease in hilly areas in the east was slight, while the decrease in plains in the west was significant, and most of the cultivated land in the west was transformed into aquaculture pit ponds (crab ponds). In Gaochun District, the amount of cultivated land in Yangjiang town and Zhuanqiang town increased, and the other six streets showed a decreasing trend, among which Qiqiao Street had the highest decreasing ratio, accounting for 48.5%.

4. Discussion

4.1. Analysis on Driving Mechanism of Cultivated Land Change

According to the results of two land surveys, the area of cultivated land for the third survey in Gaochun District in 2019 (14,410 hectares) decreased by 27,340 hectares compared with the area of cultivated land for the second transfer in 2009 (41,750 hectares), accounting for 65.5% of the total area of cultivated land for the second transfer. We explored the reasons for the decrease in cultivated land in Gaochun District, Nanjing, from the perspective of the change of cultivated land identification rules and came to the conclusion that the main reason for the sharp decrease in cultivated land in the third national land survey in Gaochun District was that the cultivated land identification rules used in the second national land survey were different from those used in the third national land survey. By interpreting the images in the second survey period in 2009 according to the rules of the third survey, the actual cultivated land area during the second survey period was 24,019 hectares, and the actual reduction in cultivated land area in the third survey in 2019 was 9609 hectares, a reduction of 23.0% of the total cultivated land area in the second survey period and 40% of the actual cultivated land area in the second survey period. The reduction in cultivated land area due to rule changes was 17,731 hectares, accounting for 64.9% of the reduction. The cultivated land that decreased due to the change of rules changed into the plantations, forests, ponds, aquaculture ponds, and other land types in the second survey period. The second survey identified it as “cultivated land for temporary planting of medicinal materials, turf, flowers, seedlings, etc., and other cultivated land for temporary change of purpose”, while the third survey investigated it as non-cultivated land according to the actual situation, indicating that the reduction of cultivated land area in Gaochun District occurred more before the second national land survey results in 2009. Therefore, although a certain degree of the non-grain-oriented phenomenon occurred in Gaochun District during the second and third national land surveys, the decrease in cultivated land caused by the change of cultivated land identification rules accounted for most of the decrease in cultivated land during the second and third national land surveys. Therefore, the main reason for the sharp decrease in cultivated land in Gaochun District in the third survey was that cultivated land identification rules used in the second and third surveys were different.
According to the existing research on cultivated land change based on the actual land category extraction results in China, the total area of cultivated land in China gradually decreased during the period from the second survey to the third survey, but the decrease was relatively stable, not reaching 7.53 million hectares [44] (that is, the difference of cultivated land area in the second and third surveys). Some scholars have also studied the change of cultivated land area in various districts of Nanjing [45], and the results also show that the cultivated land area in Gaochun District has not decreased significantly during the period of the second and third surveys. Therefore, the results of this study are consistent with the above results.

4.2. Reason Analysis

Judging from the actual decrease in cultivated land in Gaochun District, the decrease in the eastern hilly area was small, while the decrease in the western plain was large, and most of the cultivated land in the western area has been transformed into aquaculture pit ponds (crab ponds). There are two reasons for this result, namely, economic factors and policy factors.
The first is economic factors. On one hand, the comparative profit of crab farming is higher than that of grain production. The comparative profit in the production process refers to comparing the harvest of different production operations on the same land. Producers’ decision-making is driven by comparative profits, and they often choose products with higher comparative profits for production. Under the influence of the mutual restraint of domestic and international food prices, the comparative profit of planting food crops in Gaochun is low, with a net income of about 400–500 yuan/mu, while the net income of crab farming is as high as 3000–5000 yuan/mu, which is about ten times that of planting food crops. The comparative benefit is very obvious. Therefore, crab farming can bring high economic benefits to farmers and, to a certain extent, improve farmers’ enthusiasm for production. On the other hand, the demand for non-crop products in the food structure is increasing, and the consumption structure should be adjusted. In recent years, the dietary structure of urban and rural residents in China has shown the characteristics of decreasing the proportion of staple food and diversifying the structure. With the improvement in living standards, people prefer foods with high protein and nutrition, and the market of non-crop products such as shrimp and crab is constantly expanding. Therefore, in order to improve the economic benefits, hundreds of thousands of people in Gaochun are directly or indirectly engaged in crab-related breeding, processing, and sales, and river crab breeding in Gaochun has reached 14,667 hectares, with an annual output value of more than one billion yuan. Table 7 shows the comparison of the data related to pit pond farming and grain production.
Then there are policy factors. In 2017, the report of the 19th National Congress of the Communist Party of China formally put forward the implementation of the rural revitalization strategy, aiming at solving the three rural issues related to the national economy and people’s livelihood. With the in-depth implementation of the new socialist countryside construction and rural revitalization strategy, collective economic organizations and farmers have actively carried out land circulation and agricultural structure adjustments under the call of policies, some cultivated land has been adjusted to land types such as aquaculture pit ponds, and a diversified agricultural economy has been implemented. In the past ten years, Gaochun District has adhered to the strategy of “Ecological District”, taking industrial revitalization as the cornerstone and taking crab industry development as an important starting point for rural revitalization. Ecological aquaculture, which focuses on crab farming, has become the pillar industry of agriculture in the whole district and a real industry for enriching the people and has embarked on a new way of increasing agricultural efficiency, increasing farmers’ income, and enhancing rural beauty. In order to achieve high-quality development, Gaochun carried out exploration and embarked on the road of large-scale breeding, standardized production, intelligent management, and brand marketing. In 2012, the area of crab cultivation in Gaochun District was 14,200 hectares, and the output of crabs was 15,500 tons, with an output value of 1.52 billion yuan. By 2021, the area of crab cultivation was stable at 14,100 hectares, with an output of 18,000 tons and an output value of 2.3 billion yuan. Through land transfer, farmers have benefited, not only receiving subsidies but also putting their energy into working and helping large-scale aquaculture farmers raise crabs on a large scale.

4.3. Methodology

At present, the research methods of cultivated land change are mainly divided into two categories. One is based on the actual land category extraction results, and the other is based on the official land category statistics. The former mainly uses machine learning methods such as maximum likelihood method [17], decision tree [46], and random forest [47] or deep learning methods such as convolution neural network [48], cyclic neural network [49], and deep neural network [50]. The data source is mainly satellite remote sensing data, including Landsat, MODIS, Sentinel, and other medium-ground-resolution data [51,52,53,54] or QuickBird, World-View, SPOT, and other high-spatial-resolution data. The latter is to analyze the existing statistical data directly, and use the transfer matrix [55], the center of gravity transfer model [56], and regional statistics to analyze the land use change. The data source is mainly official statistical data such as the second and third national land surveys. These studies have been very effective. Whether it is to monitor the changes of cultivated land in different regions with different technical means or to analyze the various causes of the reduction of cultivated land, a large number of empirical cases have provided reliable technologies and research methods for the existing research of cultivated land. However, the impact of the change of cultivated land identification rules in different periods on cultivated land area has been less considered in the existing studies. Although some domestic scholars have studied the changes in the classification criteria of land types in the second and third surveys [57], they are only limited to the rule changes themselves. Few studies will take into account the interpretation of remote sensing images in the field at the same time, so these studies cannot analyze the impact of changes in cultivated land identification rules on the real farmland change research.
Therefore, different from the conventional methods, we combined the interpretation of cultivated land in the actual image with the change of cultivated land identification rules and analyzed the reasons for the reduction of cultivated land from a multidimensional perspective. Compared with the single-dimensional research, our research considered the problem from a more comprehensive and practical perspective. From the perspective of data statistics and analysis, we not only conducted research on the change of cultivated land area based on the official land category statistical data. In addition to the statistics and analysis of the data themselves, we also conducted research on the change of the identification rules of cultivated land and thus found the inadequacy of direct analysis and calculation of the cultivated land area of the second and third adjustments. From the perspective of the land classification interpretation method, the method we used was the deep learning method. We used the deep learning network HRNet to interpret the plantations, forests, land for construction, ponds, and aquaculture ponds. The F1 scores of these models reached 0.881, 0.899, 0.876, 0.893, and 0.914, to ensure the accuracy of the interpretation results. Different from the conventional study of cultivated land interpretation, we did not interpret the cultivated land directly. Because in the third survey, cultivated land contained a variety of land types with different characteristics, in order to ensure the accuracy of the interpretation results, we interpreted several land types other than the cultivated land first and then compared them with the official second survey cultivated land result, and the resulting difference set was the final cultivated land interpretation result. Since the official second survey cultivated land results used in this paper are field survey data, and the total area of other land types except for cultivated land, plantations, forests, land for construction, ponds, and aquaculture ponds only accounts for 2% of the total area of the cultivated land in the second survey, it could be considered that the interpretation results of cultivated land in this paper were close to the actual distribution results of cultivated land, with an error of about 2% of the total area of the cultivated land in the second survey.

4.4. The Importance of Consistency between Survey and Planning

In China, the area of cultivated land is a very important indicator. The state will formulate the objectives of cultivated land protection based on the previous statistical data of cultivated land area and determine the degree of “non-agricultural” and “non-grain” in each region. At present, the state requires that the regions with the cultivated land protection goal of the overall planning of land space lower than the 2020 cultivated land retention goal should fill the gap of responsibility for the cultivated land retention. The target of cultivated land retention in 2020 is based on the identification rules of cultivated land of the second survey, that is, “cultivated land temporarily planted with medicinal materials, turf, flowers, seedlings, etc., and other cultivated land temporarily changed for use” are identified as cultivated land. The third survey adjusted the identification rules of cultivated land and divided these cultivated lands into plantations, forests, land for construction, ponds, aquaculture ponds, and other land types according to the field survey, resulting in a sudden change and large-scale reduction in the area of cultivated land in the third survey compared with the second survey. The cultivated land protection targets defined based on the results of the third survey and the 2020 land change survey will naturally be significantly reduced compared with the 2020 cultivated land retention target. The two cultivated land identification rules are obviously different. The requirements of the state to fill the gap in cultivated land protection responsibilities in such areas are obviously not based on the same criteria. Therefore, the 2020 cultivated land retention target corresponds to the identification criteria of the second survey of cultivated land, and the cultivated land protection target of the overall land space planning should also correspond to the identification criteria of the third survey of cultivated land, so that the planning target and the land survey criteria can match each other, that is, “the old target corresponds to the old rule, and the new task adopts the new criteria”. The 2020 cultivated land retention target is used to constrain the cultivated land protection target based on the third survey, which is obviously not under the same cultivated land identification rule as the areas identified as cultivated land in the second survey. If a certain area is classified as cultivated land in the second survey but is classified into other land types in the third survey due to the change of cultivated land identification rules, resulting in the reduction of cultivated land area in this area, then this area will face the embarrassment of failing to fulfill the responsibility of cultivated land protection.
In the actual scenario, the local government will pay compensation due to the reduction of cultivated land area. However, if the reduction of cultivated land area is not due to the destruction of the cultivation layer but because of the change of the definition of cultivated land in the two land surveys, it is not appropriate to simply assess and supplement according to the requirements of the previous round of cultivated land task; otherwise, it may lead to all kinds of chaos and social contradictions in the restoration of cultivated land. Therefore, the impact of policy factors should also be considered when exploring the reasons for the reduction of cultivated land; otherwise, it will interfere with the release of subsequent agriculture-related indicators.
Figure 7 is the statistical table of cultivated land area change and cultivated land protection targets in Gaochun District. The cultivated land protection target in 2020 was set according to the cultivated land area of the second national land survey in 2009, which required the cultivated land area to reach 39,072 hectares. As the cultivated land area of the second national land survey was 41,750 hectares, the cultivated land protection target set according to the second national land survey rule was numerically reasonable. However, according to the third national land survey rule, the actual cultivated land in 2009 was only 24,019 hectares, while the actual cultivated land in 2019 was 14,410 hectares, far from the cultivated land protection target in 2020. If the protection target was set according to the second national land survey of cultivated land area and then checked according to the third national land survey of cultivated land standard, there would inevitably be a substantial reduction of cultivated land, and it would be difficult to reach the standard. Therefore, it was necessary to unify the standard of cultivated land division.
In addition to Gaochun District, there are many similar phenomena in China. Therefore, the research ideas in this paper are applicable to other regions affected by changes in cultivated land identification rules and can help relevant departments to formulate more reasonable indicators.

4.5. Shortcomings and Prospects of the Research

In addition, although this paper puts forward a novel perspective and method of cultivated land change analysis, there are some shortcomings. For example, this study only carried out experiments in Gaochun District but did not cover other areas of China. Therefore, the follow-up research scope of this paper will be gradually extended to the whole China region, and according to the results of an extensive range of research, the changing trend of cultivated land in different regions and the degree to which it is affected by policies will be analyzed.

5. Conclusions

In this paper, the deep learning network HRNet was used to interpret the images of Gaochun District in the second national land survey period in 2009 according to the third national land survey rules, and the actual cultivated land area in Gaochun District in the second national land survey period was calculated. By comparing it with the third national land survey data, the actual cultivated land reduction and the cultivated land reduction caused by the change of rules were obtained. According to the experimental results, we reached the following three conclusions:
(1)
The research method in this paper combined the interpretation results of the actual land category and the official cultivated land statistics, taking into account the impact of the change of the identification rules of cultivated land on the study of cultivated land area change, and could excavate the real change of cultivated land in the study area. This method revealed the distribution of cultivated land in the second survey under the identification rules of cultivated land in third survey and could provide a reference for the establishment of cultivated land protection objectives in some areas where the cultivated land area has been significantly reduced due to the change of the identification rules of cultivated land. According to the existing research on cultivated land change in China, the method used in this study has certain reliability and reference value.
(2)
The main reason for the sharp decline of cultivated land in the third national land survey in Gaochun District was that the cultivated land identification rules used in the second and third national land surveys were different. The total area of cultivated land for the second national land survey was 41,750 hectares, and that for the third national land survey was 14,410 hectares, a decrease of 27,340 hectares. However, according to the interpretation of the images in the second national land survey period, according to the third national land survey rules, the actual cultivated land area was only 24,019 hectares. If we take this as the total amount of cultivated land in the second national land survey period, the reduction in cultivated land between the second and third surveys was only 9609 hectares, and the reduction in arable land caused by the different rules was 17,731 hectares, accounting for 64.9% of the reduction in cultivated land between the two national land surveys. Therefore, the actual reduction of cultivated land during the second and third adjustment periods accounted for 35.1% of the reduction of cultivated land in the two land surveys. This showed that although there was a reduction in the area of cultivated land to a certain extent, the amount of reduction was not as much as the data of the two national land surveys, and the main reason for this phenomenon was the change of the identification rules of the cultivated land.
(3)
The cultivated land decrease in Gaochun District due to the change of the identification rules of the cultivated land was mainly the plantations, forests, ponds, and aquaculture ponds, which were called “cultivated land for temporary planting of herbs, turf, flowers, seedlings, etc., and other cultivated land for temporary change of purpose” in the second survey. Therefore, according to the cultivated land identification rules based on the third survey, the reduction of these cultivated land existed since the second survey, and its occurrence process could be traced back to before the second survey, rather than between the second and third surveys. The reduction of cultivated land and the increase in crab ponds and other land types in Gaochun District were the result of the local adaptation to the superior natural and geographical conditions of the lake plain. It was also formed after years of development with the improvement of people’s living standards and the improvement of food structure, driven by the market law of comparative income, and under the guidance of national policies such as new rural construction, poverty alleviation, and rural revitalization. The control of “non-grain” cultivated land should take into account food security, food security, and livelihood security.

Author Contributions

Methodology, Z.J., M.J. and Y.W.; validation, Y.W.; formal analysis, Z.J., Y.W. and C.M.; investigation, M.J.; writing—original draft preparation, Z.J. and M.J.; writing—review and editing, Y.W.; visualization, W.Q.; supervision, Y.W.; project administration, Y.W.; funding acquisition, W.Q. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by National Natural Science Foundation of China under Grant 42001196 and 42271264; the National Social Science Foundation of China later funded the projects (No. 21FSHB014).

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Zhang, F.; Xue, Y.; Ju, Z.; Zeng, L. Quantitative and Qualitative Prediction of Farmland Changes in China. Resour. Sci. 1998, 20, 35–42. [Google Scholar]
  2. Deng, X.; Huang, J.; Rozelle, S.; Zhang, J.; Li, Z. Impact of urbanization on cultivated land changes in China. Land Use Policy 2015, 45, 1–7. [Google Scholar] [CrossRef]
  3. Huang, X. Ten years of sharpening a sword, facing the problem of cultivated land protection in China—Review of the book “Farmer Behavior and Social Responsibility of Cultivated Land Protection in China”. J. Agric. For. Econ. Manag. 2018, 17, 119–120. [Google Scholar]
  4. Cai, Y. The Mechanisms of Cropland Conservation in Chinese Rural Transformation. Sci. Geogr. Sin. 2001, 21, 1–6. [Google Scholar]
  5. Luo, X.; Zeng, J.; Zhu, Y.; Zhang, L. Who will feed China: The role and explanation of China’s farmland pressure in food security. Geogr. Res. 2016, 35, 2216–2226. [Google Scholar]
  6. Long, H.L. Land Use Transition and Rural Transformation Development. Prog. Geogr. 2012, 31, 131–138. [Google Scholar]
  7. Liu, Y.; Zhou, Y. Challenges and Countermeasures for Beautiful Countryside Construction in China. J. Agric. Resour. Environ. 2015, 32, 97–105. [Google Scholar]
  8. Liu, Y.; Lu, D. The Basic Trend and Regional Effect of Agricultural Structure Adjustment in China. Acta Geogr. Sin. 2003, 58, 381–389. [Google Scholar]
  9. Su, Y.; Li, C.; Wang, K.; Deng, J.; Shahtahmassebi, A.R.; Zhang, L.; Ao, W.; Guan, T.; Pan, Y.; Gan, M. Quantifying the spatiotemporal dynamics and multi-aspect performance of non-grain production during 2000–2015 at a fine scale. Ecol. Indic. 2019, 101, 410–419. [Google Scholar] [CrossRef]
  10. Kong, X. Problems, Causes and Countermeasures of Cultivated Land “Non grain”. China Land 2020, 11, 17–19. [Google Scholar]
  11. Chen, M. Analysis of the phenomenon of “non grain” cultivated land and countermeasures. China Land 2021, 5, 9–10. [Google Scholar]
  12. Li, C.; Cheng, F. Thinking about the determination of the destruction of the cultivated layer by “non grain”. China Land 2021, 7, 12–14. [Google Scholar]
  13. Ji, C.Y. Delineating agricultural field boundaries from TM imagery using dyadic wavelet transforms. ISPRS J. Photogramm. Remote Sens. 1996, 51, 268–283. [Google Scholar] [CrossRef]
  14. Biggs, T.W.; Thenkabail, P.S.; Gumma, M.K.; Scott, C.A.; Parthasaradhi, G.R.; Turral, H.N. Irrigated area mapping in heterogeneous landscapes with MODIS time series, ground truth and census data, Krishna Basin, India. Int. J. Remote Sens. 2006, 27, 4245–4266. [Google Scholar] [CrossRef]
  15. Rydberg, A.; Borgefors, G. Integrated method for boundary delineation of agricultural fields in multispectral satellite images. IEEE Trans. Geosci. Remote Sens. 2001, 39, 2514–2520. [Google Scholar] [CrossRef]
  16. Li, Y.; Gong, J.; Wang, D.; An, L.; Li, R. Sloping farmland identification using hierarchical classification in the Xi He region of China. Int. J. Remote Sens. 2013, 34, 545–562. [Google Scholar] [CrossRef]
  17. Ashourloo, D.; Shahrabi, H.S.; Azadbakht, M.; Rad, A.M.; Aghighi, H.; Radiom, S. A novel method for automatic potato mapping using time series of Sentinel-2 images. Comput. Electron. Agric. 2020, 175, 105583. [Google Scholar] [CrossRef]
  18. Miao, R.H.; Tang, J.L.; Chen, X.Q. Classification of farmland images based on color features. J. Vis. Commun. Image Represent. 2015, 29, 138–146. [Google Scholar] [CrossRef]
  19. Diaz-Varela, R.A.; Zarco-Tejada, P.J.; Angileri, V.; Loudjani, P. Automatic identification of agricultural terraces through object-oriented analysis of very high resolution DSMs and multispectral imagery obtained from an unmanned aerial vehicle. J. Environ. Manag. 2014, 134, 117–126. [Google Scholar] [CrossRef]
  20. Mohammed, I.; Marshall, M.; de Bie, K.; Estes, L.; Nelson, A. A blended census and multiscale remote sensing approach to probabilistic cropland mapping in complex landscapes. ISPRS J. Photogramm. Remote Sens. 2020, 161, 233–245. [Google Scholar] [CrossRef]
  21. Watkins, B.; Van Niekerk, A. A comparison of object-based image analysis approaches for field boundary delineation using multi-temporal Sentinel-2 imagery. Comput. Electron. Agric. 2019, 158, 294–302. [Google Scholar] [CrossRef]
  22. Shelhamer, E.; Long, J.; Darrell, T. Fully convolutional networks for semantic segmentation. In Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, 7–12 June 2015; pp. 3431–3440. [Google Scholar] [CrossRef] [Green Version]
  23. Que, H.; Huang, H.; Xu, J. Road edge detection based on dual-threshold SSDA template matching. Remote Sens. Nat. Resour. 2014, 26, 29–33. [Google Scholar]
  24. Shanmugam, L.; Kaliaperumal, V. Junction-aware water flow approach for urban road network extraction. IET Image Process. 2016, 10, 227–234. [Google Scholar] [CrossRef]
  25. Mu, H.; Zhang, Y.; Li, H.; Guo, Y.; Zhuang, Y. Road extraction base on Zernike algorithm on SAR image. In Proceedings of the IGARSS 2016—2016 IEEE International Geoscience and Remote Sensing Symposium, Beijing, China, 10–15 July 2016. [Google Scholar]
  26. Luo, Q.; Yin, Q.; Kuang, D. Research on Extracting Road Based on lts Spectral Feature and Shape Feature. Remote Sens. Technol. Appl. 2007, 22, 339–344. [Google Scholar]
  27. Zuo, J.; Li, Y. Road Extraction from High Resolution Remote Sensing Image Based on Texture and Shape Feature. Surv. Mapp. 2013, 3, 111–113. [Google Scholar]
  28. Gaetano, R.; Zerubia, J.; Scarpa, G.; Poggi, G. Morphological road segmentation in urban areas from high resolution satellite images. In Proceedings of the International Conference on Digital Signal Processing, Corfu, Greece, 6–8 July 2011. [Google Scholar]
  29. Teng, X.; Song, S.; Zhan, Y. An Improved Road Extraction Algorithm from High-resolution Remote Sensing Images Based on Structure Tensor. Bull. Sci. Technol. 2013, 2, 175–177. [Google Scholar]
  30. Ho, T.K. The Random Subspace Method for Constructing Decision Forests. IEEE Trans. Pattern Anal. Mach. Intell. 1998, 20, 832–844. [Google Scholar]
  31. Han, J.; Kambr, M. Data Mining: Concepts and Techniques; Morgan Kaufmann Publishers: Burlington, MA, USA, 2001; pp. 279–333. [Google Scholar]
  32. Breiman, L.; Friedman, J.H.; Olshen, R.A.; Stone, J. Classification and Regression Trees; Chapman & Hall (Wadsworth, Inc.): New York, NY, USA, 1984. [Google Scholar]
  33. Shotton, J.; Fitzgibbon, A.; Cook, M.; Sharp, T.; Finocchio, M.; Moore, R.; Kipman, A.; Blake, A. Real-time human pose recognition in parts from single depth images. Commun. ACM 2013, 56, 116–124. [Google Scholar] [CrossRef] [Green Version]
  34. Lindner, C.; Bromiley, P.A.; Ionita, M.C.; Cootes, T.F. Robust and accurate shape model matching using random forest regression-voting. IEEE Trans. Pattern Anal. Mach. Intell. 2015, 37, 1862–1874. [Google Scholar] [CrossRef]
  35. Huang, G.B.; Zhu, Q.Y.; Siew, C.K. Extreme Learning Machine: A New Learning Scheme of Feedforward Neural Networks. In Proceedings of the 2004 IEEE International Joint Conference on Neural Networks, Budapest, Hungary, 25–29 July 2004; pp. 985–990. [Google Scholar]
  36. Chang, N.B.; Han, M.; Yao, W.; Chen, L.C.; Xu, S. Change detection of land use and land cover in an urban region with SPOT-5 images and partial lanczos extreme learning machine. J. Appl. Remote Sens. 2010, 4, 043551. [Google Scholar]
  37. Faugeras, O.; Keriven, R. Variational principles, surface evolution, PDE’s level set methods, and the stereo problem. IEEE Trans Image Process. 1998, 7, 336–344. [Google Scholar] [CrossRef]
  38. Simmons, J.; Przybyla, C.; Bricker, S.; Kim, D.W.; Comer, M. Physics of MRF regularization for segmentation of materials micro structure images. In Proceedings of the 2014 IEEE International Conference on Image Processing (ICIP), Paris, France, 27–30 October 2014; pp. 4882–4886. [Google Scholar]
  39. Lafferty, J.; McCallum, A.; Pereira, F.C. Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data. In Proceedings of the 18th International Conference on Machine Learning, Williamstown, MA, USA, 28 June–1 July 2001. [Google Scholar]
  40. LeCun, Y.; Bottou, L.; Bengio, Y.; Haffner, P. Gradient-based learning applied to document recognition. Proc. IEEE 1998, 86, 2278–2324. [Google Scholar] [CrossRef] [Green Version]
  41. Palangi, H.; Deng, L.; Shen, Y.; Gao, J.; He, X.; Chen, J.; Song, X.; Ward, R. Deep sentence embedding using long short-term memory networks: Analysis and application to information retrieval. IEEE/ACM Trans. Audio Speech Lang. Process. 2016, 24, 694–707. [Google Scholar] [CrossRef] [Green Version]
  42. Wang, J.; Sun, K.; Cheng, T.; Jiang, B.; Deng, C.; Zhao, Y.; Liu, D.; Mu, Y.; Tan, M.; Wang, X.; et al. Deep high-resolution representation learning for visual recognition. IEEE Trans. Pattern Anal. Mach. Intell. 2020, 43, 3349–3364. [Google Scholar] [CrossRef] [Green Version]
  43. Hou, Y.L.; Ma, J. Analysis of influencing factors and policy suggestions for the change of cultivated land area. J. Anhui Agric. Sci. 2019, 47, 60–64. [Google Scholar]
  44. Ren, Y.T.; Shi, G.Q. Regression analysis of influence factors of cultivated land area in Nanjing. Food Sci. Technol. Econ. 2016, 41, 23–26. [Google Scholar]
  45. Wu, Z.; Thenkabail, P.S.; Mueller, R.; Zakzeski, A.; Melton, F.; Johnson, L.; Rosevelt, C.; Dwyer, J.; Jones, J.; Verdin, J.P. Seasonal cultivated and fallow cropland mapping using MODIS-based automated cropland classification algorithm. J. Appl. Remote Sens. 2014, 8, 083685. [Google Scholar] [CrossRef] [Green Version]
  46. Phalke, A.R.; Özdoğan, M.; Thenkabail, P.S.; Erickson, T.; Gorelick, N.; Yadav, K.; Congalton, R.G. Mapping croplands of Europe, Middle East, Russia, and Central Asia using Landsat, Random Forest, and Google Earth Engine. J. Photogramm. Remote Sens. 2020, 167, 104–122. [Google Scholar] [CrossRef]
  47. Zhang, D.J.; Pan, Y.Z.; Zhang, J.S.; Hu, T.G.; Zhao, J.H.; Li, N.; Chen, Q. A generalized approach based on convolutional neural networks for large area cropland mapping at very high resolution. Remote Sens. Environ. 2020, 247, 111912. [Google Scholar] [CrossRef]
  48. Lu, H.; Fu, X.; Liu, C.; Li, L.G.; He, Y.X.; Li, N.W. Cultivated land information extraction in UAV imagery based on deep convolutional neural network and transfer learning. J. Mt. Sci. 2017, 14, 731–741. [Google Scholar] [CrossRef]
  49. Hansen, M.C.; Loveland, T.R. A review of large area monitoring of land cover change using Landsat data. Remote Sens. Environ. 2012, 122, 6674. [Google Scholar] [CrossRef]
  50. Zhu, Z. Change detection using landsat time series: A review of frequencies, preprocessing, algorithms, and applications. ISPRS J. Photogramm. Remote Sens. 2017, 130, 370–384. [Google Scholar] [CrossRef]
  51. Das, P. 20 years MODIS-NDVI monitoring suggests that vegetation has increased significantly around Tehri Dam reservoir, Uttarakhand, India. Remote Sens. Appl. Soc. Environ. 2021, 24, 100610. [Google Scholar] [CrossRef]
  52. Bontemps, S.; Arias, M.; Cara, C.; Dedieu, G.; Guzzonato, E.; Hagolle, O.; Inglada, J.; Morin, D.; Rabaute, T.; Savinaud, M.; et al. “Sentinel-2 for Agriculture”: Supporting Global Agriculture Monitoring. In Proceedings of the 2015 IEEE International Geoscience and Re-Mote Sensing Symposium (IGARSS), Milan, Italy, 26–31 July 2015; pp. 4185–4188. [Google Scholar] [CrossRef]
  53. Liu, R.; Zhu, D.L. Methods for Detecting Land Use Changes Based on the Land Use Transition Matrix. Resour. Sci. 2010, 32, 1544–1550. [Google Scholar]
  54. Wang, H.; Zhang, B.; Liu, Y.; Liu, Y.; Xu, S.; Zhao, Y.; Chen, Y.; Hong, S. Multi-dimensional analysis of urban expansion patterns and their driving forces based on the center of gravity-GTWR model: A case study of the Beijing-Tianjin-Hebei urban agglomeration. Acta Geogr. Sin. 2018, 73, 1076–1092. [Google Scholar]
  55. Sun, K.; Xiao, X.; Liu, D.; Wang, J. Deep high-resolution representation learning for human pose estimation. In Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, 15–20 June 2019. [Google Scholar]
  56. Wang, Y.J.; Di, Z.Z. On Land Classification Conversion between the Second and the Third National Land Surveying. Stand. Surv. Mapp. 2019, 35, 33–38. [Google Scholar]
  57. Cao, Y.; Li, G.Y.; Wang, J.Y.; Fang, X.Q.; Sun, K.Y. Systematic review and research framework of “Non-grain” Utilization of cultivated land: From a perspective of food security to muti-dimensional security. China Land Sci. 2022, 36, 1–12. [Google Scholar]
Land 12 00534 g001 550
Figure 1. Overview of the research area.
Figure 1. Overview of the research area.
Land 12 00534 g001
Land 12 00534 g002 550
Figure 2. Technical roadmap.
Figure 2. Technical roadmap.
Land 12 00534 g002
Land 12 00534 g003a 550Land 12 00534 g003b 550
Figure 3. HRNet network structure.
Figure 3. HRNet network structure.
Land 12 00534 g003aLand 12 00534 g003b
Land 12 00534 g004 550
Figure 4. Cultivated land in the third land surveys. (a1–a3) Paddy fields samples; (b1–b3) Dry land samples; (c1–c3) Irrigated land samples.
Figure 4. Cultivated land in the third land surveys. (a1–a3) Paddy fields samples; (b1–b3) Dry land samples; (c1–c3) Irrigated land samples.
Land 12 00534 g004
Land 12 00534 g005 550
Figure 5. Interpretation results.
Figure 5. Interpretation results.
Land 12 00534 g005
Land 12 00534 g006 550
Figure 6. Inversion of cultivated land by the second national land survey.
Figure 6. Inversion of cultivated land by the second national land survey.
Land 12 00534 g006
Land 12 00534 g007 550
Figure 7. Changes of cultivated land area and original protection targets of cultivated land in Gaochun District.
Figure 7. Changes of cultivated land area and original protection targets of cultivated land in Gaochun District.
Land 12 00534 g007
Table
Table 1. Changes of cultivated land identification rules.
Table 1. Changes of cultivated land identification rules.
Actual Land Types in the Second National Land Survey PeriodSecond National Land Survey ClassificationThird National Land
Survey Classification
Land on which crops are grown
Land on which crops (including vegetables) are mainly grown, interspersed with scattered fruit trees, mulberry trees, or other trees
Reclaimed beaches and seashores that are guaranteed to be harvested for one season per year on average
Fixed ditches, canals, roads, and ridges with a width of <1.0 m in the south and <2.0 m in the north
Cultivated land for temporary planting of medicinal materials, turf, flowers, and seedlings
Other cultivated land for a temporary change of use		Cultivated landCultivated land (paddy fields, dry land, and irrigated land)
Plantations, forests, ponds, and other agricultural lands
Table
Table 2. The situation of land types corresponding to the cultivated land patches in the second and third national land surveys.
Table 2. The situation of land types corresponding to the cultivated land patches in the second and third national land surveys.
Second
National Land Survey Land Type		Third National Land
Survey Land Type		Floor Space (Hectares)Proportion
Cultivated landCultivated land12,55530%
Tea plantation18925%
Forest433910%
Land for construction19825%
Ponds8592%
Aquaculture ponds19,29146%
Other8322%
Total41,750100%
Table
Table 3. Model samples.
Table 3. Model samples.
CategoryTea PlantationForestLand for ConstructionPondsAquaculture Ponds
ImageLand 12 00534 i001Land 12 00534 i002Land 12 00534 i003Land 12 00534 i004Land 12 00534 i005
LabelLand 12 00534 i006Land 12 00534 i007Land 12 00534 i008Land 12 00534 i009Land 12 00534 i010
Table
Table 4. Model accuracy.
Table 4. Model accuracy.
Model CategoryAccuracyPrecisionRecallF1-Score
Tea plantation0.9510.8450.9170.881
Forest0.9650.8750.9230.899
Land for construction0.9460.8370.9150.876
Ponds0.9540.8660.9200.893
Aquaculture ponds0.9760.8810.9460.914
Table
Table 5. Interpretation of cultivated land in Gaochun District. Unit: hectare.
Table 5. Interpretation of cultivated land in Gaochun District. Unit: hectare.
Land Use TypeTotalChunxi StreetDongba StreetGubai StreetGucheng StreetQiqiao StreetYaxi StreetYang Jian TownZhuanqiang Town
Statistical cultivated land41,7503361817325445054313410,21360183251
Inversion of cultivated land24,01987665681451342922558401729310
Inversion of tea plantations849039311075230510
Inversion of forests21544368862391305645200
Inversion of land for construction26526523244257394
Inversion of ponds143984269541981715607626
Inversion of aquaculture ponds13,024233220394488533622951832911
Table
Table 6. Statistical table of cultivated land quantity by street (town) in Gaochun District. Unit: hectare.
Table 6. Statistical table of cultivated land quantity by street (town) in Gaochun District. Unit: hectare.
RegionAgricultural AcreageCultivated Land Area Change
Second LandSecond Inversion LandThird LandThird Minus SecondThird Minus SecondReduction of Rules
Gaochun District41,75024,01914,410−27,340−960917,731
Chunxi Street3361876597−2764−2792485
Dongba Street817365684157−4016−24111605
Gubai Street25441451838−1706−6131093
Gucheng Street505434291920−3134−15091625
Qiqiao Street313422551161−1973−1094879
Yaxi Street10,21384014861−5352−35401812
Yangjian town6018729601−5417−1285289
Zhuanqiang town3251310276−2975−342941
Table
Table 7. Comparison of data related to pit pond farming and grain production.
Table 7. Comparison of data related to pit pond farming and grain production.
Pit Pond FarmingGrain Production
CostGenerally 5000–7000 yuan/acres;
8000–9000 yuan/acres along pond		90–1800/acres
ProfitAbout 5000 yuan
Gradually decrease after 2016		About 400 yuan
The income has been low and basically stable.
Purchase priceThe current purchase price for crabs is male crab 80 yuan/kg.2–4 yuan/kg
Planting timeAlmost twenty yearsLong term
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Jiang, Z.; Jiang, M.; Wang, Y.; Ma, C.; Qiao, W. An Alternative Method of Cultivated Land Identification and Its Actual Change from 2009 to 2019: A Case Study of Gaochun, China. Land 2023, 12, 534. https://doi.org/10.3390/land12030534

AMA Style

Jiang Z, Jiang M, Wang Y, Ma C, Qiao W. An Alternative Method of Cultivated Land Identification and Its Actual Change from 2009 to 2019: A Case Study of Gaochun, China. Land. 2023; 12(3):534. https://doi.org/10.3390/land12030534

Chicago/Turabian Style

Jiang, Zhuoran, Ming Jiang, Yahua Wang, Can Ma, and Weifeng Qiao. 2023. "An Alternative Method of Cultivated Land Identification and Its Actual Change from 2009 to 2019: A Case Study of Gaochun, China" Land 12, no. 3: 534. https://doi.org/10.3390/land12030534

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop