*2.3. Methods*

2.3.1. Crop Classification

This study used ENVI and ArcGIS for crop classification. First, we geometrically corrected the remote-sensing images based on DEM. After preprocessing remote-sensing images, we created a mask to crop images using the cropland vector data. Next, we performed a Landsat TM/ETM+ 453 RGB band composite and Landsat OLI 564 RGB band composite. According to crop spectral characteristics, we established the interpretation keys of soybean, maize, and rice by visual identification. Finally, we corrected the supervised classification results by visual interpretation to obtain crop information.

We assessed the accuracy of classification results through the confusion matrix and field survey. Taking the crop confusion matrix in 2020 as an example, the overall accuracy of crop classification was 91.5%, the Kappa coefficient was 0.89, and the user accuracies of soybean, maize, rice, and other crops were 86.36%, 96.36%, 100%, and 85.71%, respectively. Taking the crop field survey in 2020 as an example, the overall survey accuracy was 93%, and the field survey accuracies of soybean, maize, rice, and other crops were 92%, 88%, 98%, and 94%, respectively. The classification results met the requirements of the subsequent analysis.

#### 2.3.2. Temporal Dynamics of Crops

This section revealed the area dynamics of soybean, maize, and rice through the dynamic degree model and transition matrix.

We introduced the land use dynamic degree model to analyze the crop area change characteristics, which can quantify the change rate of the crop area. The expression is as follows [48]:

$$K = \frac{\mathcal{U}\_b - \mathcal{U}\_a}{\mathcal{U}\_a} \times \frac{1}{T} \times 100\% \tag{1}$$

where *K* is the dynamic degree of crop area during the study period; *Ua* and *Ub* represent the crop area at the beginning and end of the study, respectively; and *T* is the time interval of the study.

We introduced the land-use transition matrix to analyze the crop area conversion characteristics. The transition matrix reflects the transferred-out area at the initial period and the transferred-in area at the end period. The form of the transition matrix is as follows [49,50]:

$$S\_{ij} = \begin{bmatrix} S\_{11} & S\_{12} & \dots & S\_{1n} \\ S\_{21} & S\_{22} & \dots & S\_{2n} \\ \dots & \dots & \dots & \dots \\ S\_{n1} & S\_{n2} & \dots & S\_{nn} \end{bmatrix} \tag{2}$$

where *n* represents the number of crop types before and after the transfer; *i*, *j* (*i*, *j* 1, 2 ... , *n*) represent the crop type before and after the transfer, respectively; *Sij* represents the area of type-*i* crop before conversion to type-*j* crop after conversion.

#### 2.3.3. Spatial Dynamics of Crops

We introduced the crop kernel density to analyze crops' spatial aggregation and spatial dynamics [51]. Kernel density estimation reflects the spatial distribution density and changing trend of point groups. Before the kernel density estimation, all soybean, maize, and rice pixels were converted to points that were then algorithmically output to represent crop density. Suppose that (*x*1, ... ... , *xn*) is a series of *n* identically and independently distributed observations; the kernel estimator of the *x* density is given by [52,53]:

$$f\_{\mathbb{H}}(\mathbf{x}) = \frac{1}{nh} \sum\_{i=1}^{n} K(\frac{\mathbf{x} - \mathbf{x}\_{i}}{h}) \tag{3}$$

where *K* is the kernel function and *h* is a smoothing parameter called bandwidth. The bandwidth *h* is calculated as follows [54]:

$$h = 0.9 \ast \min\left(SD\_\prime \sqrt{\frac{1}{\ln(2)}} \ast D\_m\right) \ast n^{-0.2} \tag{4}$$

where *SD* is the standard distance, *Dm* is the distance from the mean center for all points, and *n* is the number of points.

In addition, we defined the kernel density levels as follows: kernel density values between 0 and 400 frequency/km2 are "low level", kernel density values between 401 and 800 frequency/km<sup>2</sup> are "medium level", and kernel density values between 801 and 1200 frequency/km2 are "high level".

#### 2.3.4. Determination and Spatial Characterization of CPS

The CPS type is determined based on one crop area as a percentage of the total area. Given previous studies in Northeast China [5,55], there are usually no more than three crops in a CPS type. The CPS type is determined as follows: if only a specific crop area

accounts for more than 30% of the total crop area in the county, the CPS is the single type of this crop; if there are two or three types of crop areas accounting for more than 30% of the total crop area, the CPS is a combination of these crops; if all crop areas account for less than 30% of the total crop area, the CPS is a combination of the top three crops. Considering the spatial variation of CPS, we further calculated the variation in CPS types on the 900 × 900 m unit.

#### **3. Results**

*3.1. Temporal Dynamics of Crops*

#### 3.1.1. Area Changes in Crops

Soybean and maize were the main crops with over 70% of the total area from 2000 to 2020. They alternated as the crops with the most area. Rice and other crops accounted for less than 30% of the area (Table 2).


**Table 2.** Area of major crops in Hailun County.

From 2000 to 2020, rice showed a significant relative increase in area (over 80%). The second-largest relative increase in area was for soybean (71.8%). In contrast, the area of maize and other crops decreased by 31.7% and 76.4%, respectively. In addition, the relative area change of major crops varies over the period. The area changes in soybean and maize were the most pronounced across periods (Table 3).

**Table 3.** Area changes of major crops in Hailun County.


(+) represents the increase in rate; (−) represents the decrease in rate.

#### 3.1.2. Area Conversion among Crops

Specifically, the area interconversion featured differently across stages. From 2000 to 2005 (Figure 2a), soybean had the most significant area increase, while maize was the crop with the most considerable area decrease. The gain area of soybean was 13.83 × <sup>10</sup><sup>4</sup> hm2, and the lost area of maize was 10.35 × <sup>10</sup><sup>4</sup> hm2. For soybean, 52% of the gain area came from the maize, and 40% was translated from other crops. For maize, 70% of the lost area translated to soybean and 17% to other crops. From 2005 to 2010 (Figure 2b), maize was the crop with the most area reduction. The area loss of maize was 6.68 × 104 hm2. A total of 55% of the lost area was converted to soybean and 28% to other crops. From 2010 to 2015 (Figure 2c), maize was the crop with the most significant area increase, and soybean was the crop with the most significant area decrease. The gain area of maize was 14.82 × 104 hm2, and the lost area of soybean was 14.72 × <sup>10</sup><sup>4</sup> hm2. Specifically, 79% of the gained area of maize came from soybean conversion, and 79% of the lost soybean area was converted to maize. From 2015 to 2020 (Figure 2d), the soybean area had the most significant increase, and the maize area had the largest decrease. The soybean gain area was 12.01 × 104 hm2, and the maize loss area was 13.41 × 104 hm2. For soybean, 81% of the gain area came from the maize conversion. For maize, 72% of the lost area was converted to soybean.

**Figure 2.** Transition matrix of crops 2000–2005, 2005–2010, 2010–2015 and 2015–2020 (**a**–**d**) (S for soybean; M for maize; R for rice; O for other crops; N for non-cultivated land).

Overall, the area interconversion was concentrated between soybean and maize. The interconverted area between paddy and dryland was almost negligible.

#### *3.2. Spatial Dynamics of Crops*

Specifically, crops' distribution ranges and aggregation characteristics differed significantly. According to Figure 3a–e, soybean was the most widely distributed and featured different aggregations. In 2005, 2010, and 2020, soybean was widely distributed and showed strong spatial aggregation. The frequency percentages of soybean kernel density at "high level" were about 30%. In 2000 and 2015, the spatial aggregation of soybean was weaker, with the frequency percentages of kernel density at a "high level" of about 4%. According to Figure 3f–j, maize was mainly distributed in the south-central part of Hailun County. In 2000 and 2015, the maize distribution was more widespread and aggregated. The frequency percentages of the maize kernel density at "high level" were 7.14% and 21.69%, respectively. In 2005, 2010, and 2020, the distribution and aggregation of maize were small, with some maize distributed along the Tongken River. According to Figure 3k–o, rice was the least widespread of the major crops. It was characterized by aggregation mainly near water sources, with the most pronounced spatial aggregation in the Tongken River and Zhayin River. In 2015 and 2020, rice showed more robust water aggregation than in other years.

**Figure 3.** Kernel density of soybean (**a**–**e**); maize (**f**–**j**); rice (**k**–**o**) in 2000, 2005, 2010, 2015, and 2020.

Overall, soybean and maize were spatially complementary in distribution and aggregation. In 2005, 2010, and 2020, soybean's distribution breadth and aggregation intensity far exceeded maize's. Conversely, maize's distribution breadth and aggregation intensity far exceeded soybean's in 2000 and 2015. Moreover, the aggregation centers of soybean and maize did not overlap and complemented each other.

#### *3.3. Determination and Spatial Characterization of CPS*

We obtained the CPS types for the past 20 years in Hailun County based on the CPS determination criteria. The results were "single type of maize" (2000), "single type of soybean" (2005), "single type of soybean" (2010), "single type of maize" (2015), and "single type of soybean" (2020), respectively. In addition, the CPS evolution pattern was the interconversion between a single type of soybean and maize.

The CPS spatial distribution based on 900 m × 900 m units is determined by the crop spatial distribution based on 30 m × 30 m units. Therefore, they shared similar spatial distribution characteristics. According to Figure 4, a "single type of soybean" and "single type of maize" were the main CPS types, accounting for more than 40% of the total area. A "single type of soybean" was distributed throughout the territory, and a "single type of maize" was mainly distributed in the south-central part of the study area. The area proportion of a "single type of rice" was about 10%, which was mainly distributed near water sources, especially near the Tongken River and Zhayin River. "Mixed type of soybean–maize" was the largest area of mixed types, but the area was much smaller than that of a single type. "Mixed type of soybean–maize" was mainly distributed in the central and northern parts of Hailun County. The other "mixed types of two crops" were the non-dominant CPS types, and irregularities characterized their spatial distribution. The area of "mixed types of three crops" was almost negligible.

**Figure 4.** Spatial distribution of CPS based on 900 m × 900 m units in 2000, 2005, 2010, 2015, and 2020 (**a**–**e**) (1: single type of soybean; 2: single type of maize; 3: single type of rice; 4: single type of other crops; 12: mixed type of soybean–maize; 13: mixed type of soybean–rice; 14: mixed type of soybean– other crops; 23: mixed type of maize–rice; 24: mixed type of maize–other crops; 34: mixed type of rice–other crops; 123: mixed type of soybean–maize–rice; 124: mixed type of soybean–maize–other crops; 134: mixed type of soybean–rice–other crops; 234: mixed type of maize–rice–other crops).

#### **4. Discussion**

#### *4.1. Explanation for Geospatial Distribution of Crops in Hailun County*

Soybean, maize, and rice showed significant differences in spatial distribution characteristics from 2000 to 2020 in Hailun County. The spatial distribution of regional crops is mainly influenced by natural conditions, socio-economic, and agricultural policies [56]. For soybean, Hailun County is blessed with widespread black soil and climatic conditions suitable for growing soybean. This county is also well experienced in growing soybean [57]. As a result, soybean is the main crop and is distributed throughout the territory. For maize, the yield and quality are higher in plains than in hilly areas. Farmers preferred to grow maize in the plains for higher economic benefits [58]. Therefore, maize in Hailun County is mainly distributed in the central and southern plains. For rice, it is concentrated near water sources due to its high water demand. From 2010 to 2014, China raised the minimum rice purchase price for five consecutive years [59]. During this period, more farmers preferred to grow rice. Some dry fields along the Tongken River were converted to paddy fields. As a result, rice's distribution breadth and aggregation intensity along the Tongken River increased after 2015.

For soybean and maize, rotation improves the physicochemical properties of black soil to increase crop yield [60]. BSRNC has been rotating soybean and maize for the past 20 years. In 2016, the "Pilot Program for Exploring the Implementation of Cultivated Land Rotation and Fallowing System" proposed to conduct a pilot rotational fallowing system in Northeast China, and Hailun County was listed [11]. Therefore, the soybean and maize distributional breadth and aggregation intensity were complementary spatially across years.

### *4.2. Crop Planting Structure and Food Security*

The Chinese agricultural conflict has shifted from an insufficient crop yield to an irrational crop structure in recent decades. To optimize this structure, China's food security strategy aims to balance the food supply and demand [61]. Rational CPS adjustment contributes to optimizing the crop yield proportion to achieve this balance. Most relevant studies have found that state macro-regulation and changes in farmers' cropping concept are the two primary measures to ensure food security from the CPS perspective [9,42].

One is the macro-control of the state. The CPS evolution is closely related to China's food security policy. " National mid-to long-term food security plan (2008–2020)" has set the goal of stabilizing the food self-sufficiency rate above 95% by 2020 [62]. At present, rice, wheat, and maize yields basically meet the demand for self-sufficiency. Due to imported soybean's low price and high quality, China still accounts for more than 60% of global soybean imports [10]. The BSRNC contributes to 40% of China's total soybean yield and plays a vital role in the national soybean supply [63]. Therefore, a new round of county-based regulation of CPS in BSRNC helps to realign the crop yield proportion, thus guaranteeing food security.

The other is the change in farmers' cropping concept. According to the "Agricultural Supply-side Structural Reform," the CPS optimization requires a shift in farmers' cropping concept [64]. Food prices and planting subsidies directly influence farmers' cropping concept. From 2000 to 2005, the soybean price in Hailun County was about 1.5–2-times higher than that of maize. The subsidies for growing soybean were also much higher than for growing maize. As a result, the CPS transformed from a "single type of maize" to a "single type of soybean." Overall, the optimization of CPS in BSRNC requires reasonable adjustment by the state and changes in farmers' cropping concepts to safeguard food security.

#### *4.3. Crop Planting Structure and Policy Implementation*

China has implemented policies to optimize the CPS of BSRNC at three levels. On overall policies, the "Outline of the Medium and Long-term National Food Security Plan (2008–2020)" and "National Agricultural Sustainable Development Plan (2015–2030)" called for rational optimization of agricultural production layout to reach a food balance between supply and demand [62]. On specific policies, "Agricultural Supply-side Structural Reform" and " National Planting Structural Adjustment Plan (2016–2020)" required the rational CPS adjustment to achieve a reasonable crop yield proportion [17,65]. More specifically, "Guiding Opinions on the Adjustment of Maize Structure in the 'Sickle' Region" is required to reduce the excessive maize area, restore the shrinking soybean area, and achieve a proper crop rotation in BSRNC.

According to our results, the CPS adjustment over the past 20 years has largely met the requirements of these policies. However, there were still some deviations in the implementation of the policy. For example, the maize area far exceeded the soybean area in 2015. In this case, the new round of CPS adjustment in BSRNC should prioritize policy implementation at the county scale. The county-level geographic characteristics of CPS evolution in BSRNC contribute to monitoring the performance of these policies and guiding the direction of CPS adjustments.

We revealed the geographic characteristics of CPS dynamics in Hailun City from both temporal and spatial perspectives. Due to the limitations in data availability and precision, we did not perform a comprehensive and detailed CPS analysis. However, the geographic characteristics of CPS are both the result of spatiotemporal evolution and the prerequisite for CPS adjustment. Therefore, we will conduct longer time series and more detailed analyses of spatio-temporal characteristics and influencing factors to provide BSRNC with detailed CPS optimization recommendations in future studies.

### **5. Conclusions**

This research attempted to systematically characterize the crop area dynamics and crop distribution dynamics and determine the CPS types in Hailun County. The main findings are as follows: From 2000 to 2020, soybean and maize had the largest area and alternated as the most dominant crop. The area of rice and other crops was tiny. The area interconversion was mainly concentrated between soybean and maize. Relatively, soybean was the most widely distributed and aggregated crop. The soybean and maize distribution breadth and aggregation intensity were spatially complementary in different years. Rice had the smallest distribution range but showed a substantial aggregation of water sources. In addition, the CPS types were a single type of soybean or maize in 2000, 2005, 2010, 2015, and 2020. The CPS evolution pattern was the interconversion between a single type of soybean and maize. The CPS spatial distribution had similar characteristics to the crop spatial distribution.

This study provides a new perspective for CPS research in BSRNC: the spatio-temporal analysis based on county-level geographic characteristics. The results suggest a future CPS adjustment of the BSRNC is necessary, and such an adjustment requires a county-level optimization of the crop-area proportion and crop spatial aggregation. These findings inform the Chinese government's new round of CPS adjustments for BSRNC to safeguard food security.

**Author Contributions:** Conceptualization, Q.L. and G.D.; methodology, Q.L. and W.L.; validation, Q.L., W.L. and G.D.; formal analysis, W.L.; investigation, Q.L., W.L., G.D. and L.W.; resources, G.D.; data curation, W.L., H.W. and Y.L.; writing—original draft preparation, W.L.; writing—review and editing, Q.L., G.D., B.F. and S.Q.; visualization, W.L. and Y.L.; supervision, Q.L. and G.D.; project administration, Q.L.; funding acquisition, Q.L. and S.Q. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by National Natural Science Foundation of China, grant number 41901208, National Natural Science Foundation of China, grant number 42101217, China Postdoctoral Science Foundation, grant number 2021M700738.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy or other restrictions.

**Acknowledgments:** We express gratitude to the Northeast Agricultural University professionals who have participated in the research and survey.

**Conflicts of Interest:** The authors declare no conflict of interest.
