Spatiotemporal Patterns of Cultivated Land Quality Integrated with Multi-Source Remote Sensing: A Case Study of Guangzhou, China

: Scientiﬁcally revealing the spatiotemporal patterns of cultivated land quality (CLQ) is crucial for increasing food production and achieving United Nations Sustainable Development Goal (SDG) 2: Zero Hunger. Although studies on the evaluation of CLQ have been conducted, an effective evaluation system that is suitable for the macro-regional scale has not yet been developed. In this study, we ﬁrst deﬁned the CLQ from four aspects: soil fertility, natural conditions, construction level, and cultivated land productivity. Then, eight indicators were selected by integrating multi-source remote sensing data to create a new CLQ evaluation system. We assessed the spatiotemporal patterns of CLQ in Guangzhou, China, from 2010 to 2018. In addition, we identiﬁed the main factors affecting the improvement of CLQ. The results showed that the CLQ continuously improved in Guangzhou from 2010 to 2018. The area of high-quality cultivated land increased by 13.7%, which was mainly distributed in the traditional agricultural areas in the northern and eastern regions of Guangzhou. The areas of medium- and low-quality cultivated land decreased by 8.1% and 5.6%, respectively, which were scattered throughout the whole study area. The soil fertility and high productivity capacity were the main obstacle factors that affected the improvement of CLQ. Simultaneously, the obstacle degree of stable productivity capacity gradually increased during the study period. Therefore, the targeted improvement measures could be put forward by applying biofertilizers, strengthening crop management and constructing well-facilitated farmland. The new CLQ evaluation system we proposed is particularly practical at the macro-regional scale, and the results provided targeted guidance for decision makers to improve CLQ and promote food security.


Introduction
Cultivated land is the most important natural resource that plays a vital role in ensuring food security and the sustainable development of human society [1]. However, cultivated land is currently suffering from frequent natural disasters [2,3], soil nutrient loss [4,5], soil acidification, and heavy metal pollution [6,7] due to climate change and high-intensity utilization [8]. The degradation of cultivated land quality (CLQ) not only caused serious eco-environmental problems but also posed great threats to food security [9]. Therefore, how to improve CLQ and increase food production have become a research

Study Area
Guangzhou (112°57′-114°03′ E, 22°26′-23°56′ N) is the capital city of Guangdong Province and the largest city in southern China (Figure 1). Guangzhou contains 11 districts, including Yuexiu, Haizhu, Liwan, Tianhe, Baiyun, Huangpu, Nansha, Panyu, Huadu, Conghua, and Zengcheng. The minimum terrain altitude of Guangzhou is 0 m, the maximum is 1181 m, the mean is 107 m, and the median is 34 m. The terrain in Guangzhou is high in the northeast and low in the southwest, with mountains and hills in the north, hilly basins in the middle, and coastal alluvial plains in the south. The cultivated land is dominated by paddy fields, and rice is the main crop, with the rice planting area accounting for 88% of the total cultivated land area [41]. Other crops, such as vegetables, only account for approximately 10%. The soil type of the cultivated land is lateritic red soil (Ferralsol, World Reference Base for Soil Resources 2015), which is characterized by strong acidity and low nutrient content. The mean annual temperature (MAT) in Guangzhou is 21.9 °C, and MAT is higher in the south and lower in the north. The mean annual precipitation (MAP) in Guangzhou is about 1800 mm, and MAP is lower in the south and higher in the north, and the precipitation is mainly concentrated from April to September. We chose Guangzhou as the study area because, in 2018, the Guangzhou municipal government proposed to build the "Vegetable Basket" production and circulation service system of the Guangdong-Hong Kong-Macao Greater Bay Area [42] with Guangzhou as the hub, aiming to provide more and better edible agricultural products. However, a large amount of cultivated land in Guangzhou has been converted into built-up land due to urbanization [43]. In this context, the CLQ in Guangzhou will be a key factor in achieving the goal of providing more and better edible agricultural products to a large extent. Therefore, it is urgent and important to carry out a spatiotemporal evaluation of the CLQ in Guangzhou to promote the efficient improvement of the CLQ.

CLQ Evaluation Method 2.2.1. The Definition of CLQ
This study defined the CLQ from four aspects, including soil fertility, natural conditions, construction level, and cultivated land productivity. Soil fertility is the foundation of crop growth and is the most critical indicator used to measure the CLQ [44,45]. The natural conditions of cultivated land can reflect some of the background properties that are closely related to the CLQ [27,33]. Cultivated land is a semi-artificial and semi-natural ecosystem; thus, the impact of construction level on the CLQ cannot be ignored [15,17]. The productivity capacity of cultivated land is the most direct manifestation of the CLQ [46] and should be taken into account.

Indicator Selection
Based on the proposed definition of CLQ, eight indicators were selected by integrating multi-source remote sensing data to construct a comprehensive CLQ indicator system (Table 1). The indicators used to evaluate soil fertility are soil fertility size and stability. Soil fertility size represents whether the current soil fertility is good, while soil fertility stability focuses on the state of soil fertility over a period of time. Soil fertility is affected by many factors, but the growth status of the vegetation can be used as a direct reflection of soil fertility [47]. Normalized Difference Vegetation Index (NDVI) is obtained from remote sensing images and can accurately reflect the growth of vegetation; thus, it is often referred to as the "soil fertility index" and has been used to fully characterize the soil fertility of cultivated land [37,48]. In this study, we used the mean of NDVI and coefficient of variation (CV) of NDVI of cultivated land for three consecutive years to characterize the size and stability Remote Sens. 2022, 14, 1250 5 of 20 of soil fertility, respectively. For soil fertility size, we first calculated the average NDVI within one year and then calculated the average NDVI for three consecutive years. Due to weather conditions, high-quality remote sensing images (cloud cover was less than 10%) within one year were concentrated from September-November. September-November is a critical period for crop growth and harvesting in Guangzhou [49]. Therefore, the average NDVI during this period can effectively characterize the soil fertility of cultivated land. If the mean of NDVI is larger and the CV of NDVI is smaller for three consecutive years, then the soil fertility of cultivated land is better.
To evaluate the natural conditions of the cultivated land, we chose the slope and topsoil texture, which are the most frequently used and representative indicators due to their wide availability [50,51]. Previous studies have found that the slope is closely related to CLQ; that is, the greater the slope, the worse the CLQ [30]. Moreover, topsoil texture (0-30 cm) is most relevant to soil quality and crop growth, which is widely used in many previous CLQ evaluation studies [11,40]. If the slope is smaller and the topsoil texture is fine, then the natural conditions of the cultivated land are better.
To facilitate field production activities and meet the needs of agricultural mechanized production, farmers usually build a sound field road system and carry out land consolidation projects [19]. Therefore, in this study, road accessibility and centralized connectivity were used to represent the construction level of the cultivated land. Road accessibility refers to the distance from the cultivated land to the nearest road [52]. Centralized connectivity refers to whether the cultivated land is connected and to which degree of connectivity. In this study, the CONTIG landscape index was used to represent the centralized connectivity of cultivated land. The CONTIG index is a contiguity index, which can effectively assess the spatial connectedness of objects [53]. If road accessibility is smaller and centralized connectivity is larger, then the construction level of the cultivated land is better.
The productivity capacity of cultivated land was evaluated from the perspectives of the high productivity capacity and the stable productivity capacity. Previous studies have shown that net primary productivity (NPP) derived from MODIS 17 A3 (MOD17A3) can effectively characterize the productivity level of cultivated land [54,55]. The advantage of using NPP is that time-series data are readily available, and they can serve as a common comparable unit across different crop types [55]. In this study, we used the mean of NPP and the CV of NPP for three consecutive years to represent the high productivity capacity and stable productivity capacity of cultivated land, respectively. If the mean of NPP is large and the CV of the NPP is small for three consecutive years, then the cultivated land productivity is better.

Calculation of the CLQ index
The CLQ index was used to evaluate CLQ and mainly consists of three steps: (1) indicator selection and weight determination; (2) grading and scoring indicators; (3) calculation of the weighted sum of different indicators [10]. The impact of different evaluation indicators on the CLQ should be different; that is, their weight values are different [32]. Therefore, it is critical to set appropriate weight values for the evaluation indicators. In this study, we consulted ten experts from Guangdong Academy of Agricultural Sciences and South China Agricultural University in the form of questionnaires. They ranked the importance of the selected indicators. We combined the expert survey results with an analytic hierarchy process (AHP) to calculate the weight coefficients. In addition, the weight results passed the consistency test (consistency ratio value less than 0.1) [56]. We divided each evaluation indicator into three grades: high, medium, and low (Table 1). The grading of slope refers to China's national standard: the regulation for gradation of agricultural land quality [25]. The grading of the topsoil texture refers to the classification standard of the Harmonized World Soil Database (http://www.fao.org/soils-portal/, accessed on 10 October 2020). The grading of road accessibility refers to a previous study on CLQ evaluation [57]. The grading of the remaining five indicators is based on Jenks classification, which can reduce the intra-class variance and maximize the inter-class variance [58]. We used the scoring Remote Sens. 2022, 14, 1250 6 of 20 method to score the high, medium, and low grades as 3, 2, and 1, respectively. Finally, we calculated the CLQ index as follows: where CLQI is the CLQ index; W i is the weight of the evaluation indicator; Y i is the score of the evaluation indicator; n is the number of evaluation indicators (n = 8 in this study).

Grading the CLQ index
The CLQ index ranges from 1.00 to 3.00. The larger the index is, the better the CLQ. In this study, the CLQ was divided into three grades, including high quality, medium quality, and low quality, according to the following criteria: (2)

Multi-Source Data
Detailed information about the data used in this study is presented in Table 2. The cultivated land data for Guangzhou were obtained from China's National Land Use and Cover Change (CNLUCC) dataset. The dataset was generated using a uniform classification method with visual interpretation based on Landsat images, and it was provided by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences. The overall accuracy of the dataset was >90% [59]. The NDVI was calculated to evaluate the soil fertility of the cultivated land based on the Landsat Thematic Mapper/Operational Land Imager (TM/OLI) data in ENVI 5.3. However, due to the quality of Landsat remote sensing image being not good (cloud cover was greater than 10%) in 2018, we chose a high-quality Sentinel-2 image with higher spatial resolution (10 × 10 m) as a substitute. To maintain the same resolution as the Landsat images, the Sentinel-2 image was resampled to 30 m.
Using ArcGIS (version 10.6; Software for Spatial Analysis; Environmental Systems Research Institute, Redlands, CA, USA, 2019), the slope of the cultivated land was extracted based on the Digital Elevation Model (DEM) data, and the distance from each cultivated land pixel to the nearest road was calculated. In this study, the Fragstats (version 4. was used to calculate the Contig landscape index. The calculation of the mean and CV of the NPP was conducted in ENVI 5.3. To unify the evaluation unit, all data were resampled to 30 m.

Validation Dataset
In this study, the validation data consisted of the comprehensive evaluation results of field-measured indicators, including soil organic matter, total nitrogen, available phosphorus, available potassium, pH, slope, topsoil texture, soil thickness, soil bulk density, soil parent material, irrigation guaranteed rate, drainage conditions, yield. Due to the limitations of the acquisition of the validation data, we took the Conghua, Zengcheng, and Nansha districts as examples to verify the results. There were 75 ground monitoring points in Conghua district, 89 ground monitoring points in Zengcheng district, and 65 ground monitoring points in Nansha district. The number and spatial location of the ground sampling points in the three districts were the same for all study years ( Figure 1).
To further verify the evaluation results of the CLQ in the entire study area, we chose the spatiotemporal distribution data for well-facilitated farmland (i.e., high-quality cultivated land formed through land consolidation and construction within a certain period) in Guangzhou as an additional validation dataset. Well-facilitated farmland construction is considered to be a strategic measure to ensure China's national food security [57,60]. The purpose of well-facilitated farmland construction is to improve the comprehensive quality of the cultivated land; thus, its spatial distribution data consisted of a validation dataset suitable for the needs of this study. According to China's National Overall Plan for Well-Facilitated Farmland Construction, when the construction of well-facilitated farmland is completed, the CLQ grade will be significantly improved. Therefore, we reasonably believe that the CLQ in a well-facilitated farmland construction area is significantly better than that in a non-well-facilitated farmland construction area. We further verified the accuracy of the evaluation results for the study area by comparing the CLQ index values (obtained in this study) of a well-facilitated farmland and non-well-facilitated farmland.
Guangzhou started implementing the well-facilitated farmland construction project in 2012, and by the end of 2018, 309 well-facilitated farmland construction projects had been completed ( Figure 1). Therefore, we used the spatial distribution data for the wellfacilitated farmland constructed in 2015 and 2018 to verify the evaluation results of the CLQ for 2015 and 2018, respectively.

Obstacle Factor Diagnosis Model
In this study, the obstacle factor diagnosis model was used to identify the main factors affecting the improvement of the CLQ. The obstacle factor diagnosis model identifies the obstacle factors based on three basic variables: the factor contribution degree, indicator deviation degree, and obstacle degree [28]. The greater the obstacle degree of the evaluation indicator is, the stronger its negative effect on the improvement of the CLQ. The equation for calculating the obstacle degree is as follows: where O i is the obstacle degree of the ith indicator; W i is the factor contribution degree of the ith indicator, which is equal to the weight of the indicator; I i is the deviation degree of the ith indicator, which is equal to the difference between the highest value and the actual value of the indicator. Here, I i = 3.00 − Y i .

Accuracy Verification of the CLQ
The results of CLQ based on field measurements (measured) and the remote sensing evaluation (evaluated) are shown in Figure 2. In the Conghua district, there was a good spatial consistency between the measured results and the evaluation results. High-quality cultivated land was mainly concentrated in the central and western regions, and mediumand low-quality cultivated land was widely distributed throughout the entire region. In the Zengcheng district, there is a relatively good spatial consistency between the measured results and the evaluation results. High-quality cultivated land was concentrated in the southeastern and northern regions, medium-quality cultivated land was widely distributed, and low-quality cultivated land occupied a relatively small proportion. In the Nansha district, there is a slightly larger spatial difference between the measured results and the evaluation results in 2010. However, in 2015 and 2018, the evaluation results were in good agreement with the measured results. Overall, the verification results can largely indicate that the resultant CLQ based on remote sensing evaluation was reliable.
where Oi is the obstacle degree of the ith indicator; Wi is the factor contribution degree of the ith indicator, which is equal to the weight of the indicator; Ii is the deviation degree of the ith indicator, which is equal to the difference between the highest value and the actual value of the indicator. Here, Ii = 3.00 − Yi.

Accuracy Verification of the CLQ
The results of CLQ based on field measurements (measured) and the remote sensing evaluation (evaluated) are shown in Figure 2. In the Conghua district, there was a good spatial consistency between the measured results and the evaluation results. High-quality cultivated land was mainly concentrated in the central and western regions, and mediumand low-quality cultivated land was widely distributed throughout the entire region. In the Zengcheng district, there is a relatively good spatial consistency between the measured results and the evaluation results. High-quality cultivated land was concentrated in the southeastern and northern regions, medium-quality cultivated land was widely distributed, and low-quality cultivated land occupied a relatively small proportion. In the Nansha district, there is a slightly larger spatial difference between the measured results and the evaluation results in 2010. However, in 2015 and 2018, the evaluation results were in good agreement with the measured results. Overall, the verification results can largely indicate that the resultant CLQ based on remote sensing evaluation was reliable. The grade difference between the remote sensing evaluation results and the measured results of CLQ is shown in Figure 3. From 2010 to 2018, in the Conghua district, the The grade difference between the remote sensing evaluation results and the measured results of CLQ is shown in Figure 3. From 2010 to 2018, in the Conghua district, the average 49% of CLQ had a grade difference of 0, the average 40% of CLQ had a grade difference of 1 (including 1 and −1), and the average 11% of CLQ had a grade difference of 1 (including 2 and −2). In the Zengcheng district, the average 48% of CLQ had a grade difference of 0, the average 40% of CLQ had a grade difference of 1 (including 1 and −1), and the average 12% of CLQ had a grade difference of 1 (including 2 and −2). In the Nansha district, about 40% of CLQ had a grade difference of 0 in 2010, which was relatively low. In 2015 and 2018, the average 47% of CLQ had a grade difference of 0, the average 39% of CLQ had a grade difference of 1 (including 1 and −1), and the average 14% of CLQ had a grade difference of 2 (including 2 and −2). This indicated that the remote sensing evaluation results of CLQ were in good agreement with the measured results. average 49% of CLQ had a grade difference of 0, the average 40% of CLQ had a difference of 1 (including 1 and −1), and the average 11% of CLQ had a grade diffe of 1 (including 2 and −2). In the Zengcheng district, the average 48% of CLQ had a difference of 0, the average 40% of CLQ had a grade difference of 1 (including 1 an and the average 12% of CLQ had a grade difference of 1 (including 2 and −2). Nansha district, about 40% of CLQ had a grade difference of 0 in 2010, which wa tively low. In 2015 and 2018, the average 47% of CLQ had a grade difference of average 39% of CLQ had a grade difference of 1 (including 1 and −1), and the averag of CLQ had a grade difference of 2 (including 2 and −2). This indicated that the r sensing evaluation results of CLQ were in good agreement with the measured resu The comparison of the CLQ results for the well-facilitated farmland and the well-facilitated farmland is shown in Table 3. In 2015 and 2018, the CLQ index of the facilitated farmland construction area was significantly greater than that of the non-wel itated farmland construction area. Moreover, the CLQ of the well-facilitated farmlan struction areas was mainly high quality (about 50%), while the CLQ in the non-well-faci farmland construction area was mainly medium quality (about 40%) in these two years. results are consistent with the reasonable hypothesis made in this study, indicating th resultant CLQ evaluation using integrated remote sensing data is reliable.  The comparison of the CLQ results for the well-facilitated farmland and the nonwell-facilitated farmland is shown in Table 3. In 2015 and 2018, the CLQ index of the well-facilitated farmland construction area was significantly greater than that of the nonwell-facilitated farmland construction area. Moreover, the CLQ of the well-facilitated farmland construction areas was mainly high quality (about 50%), while the CLQ in the non-well-facilitated farmland construction area was mainly medium quality (about 40%) in these two years. These results are consistent with the reasonable hypothesis made in this study, indicating that the resultant CLQ evaluation using integrated remote sensing data is reliable. The CLQ index in Guangzhou increased steadily from 2010 to 2018 (Figure 4), indicating that the CLQ in Guangzhou continuously improved. From 2010 to 2018, the high-quality cultivated land increased by about 10,000 ha, and the proportion increased by 13.65%. The medium-quality and low-quality cultivated land decreased by 12,000 and 8000 ha, respectively, and their proportions decreased by 8.10% and 5.55%, respectively (Table 4).
Spatially, from 2010 to 2018, the high-quality cultivated land in the eastern and northern parts of Guangzhou expanded continuously, and the medium-and low-quality cultivated lands decreased significantly, which indicated that the CLQ in these two regions continuously improved. In the western and southern parts of Guangzhou, the high-quality cultivated land initially decreased and then increased. Correspondingly, the mediumand low-quality cultivated land initially increased and then decreased. The CLQ in these two regions improved slightly overall. The CLQ in the central region was low, and there was no significant change during the study period ( Figure 5).  (Table 4).  Spatially, from 2010 to 2018, the high-quality cultivated land in the ea ern parts of Guangzhou expanded continuously, and the medium-and l vated lands decreased significantly, which indicated that the CLQ in th continuously improved. In the western and southern parts of Guangzho ity cultivated land initially decreased and then increased. Corresponding and low-quality cultivated land initially increased and then decreased. T

Spatiotemporal Changes of CLQ in Unchanged/Lost/Gained Area
Spatiotemporal changes of CLQ in the unchanged area (cultivated land that had not been converted to other land use types during the study period) are presented in Table 5. In these three different periods, the quality grade of about 50% of the cultivated land did not change. The proportion of cultivated land with an improved quality grade (including improvement by 2 grades and 1 grade) was greater than that of the cultivated land with a decreased quality grade (including a decrease by 2 grades and 1 grade). This indicated that the quality of the unchanged cultivated land improved from 2010 to 2018.

Spatiotemporal Changes of CLQ in Unchanged/Lost/Gained Area
Spatiotemporal changes of CLQ in the unchanged area (cultivated land that had not been converted to other land use types during the study period) are presented in Table 5. In these three different periods, the quality grade of about 50% of the cultivated land did not change. The proportion of cultivated land with an improved quality grade (including improvement by 2 grades and 1 grade) was greater than that of the cultivated land with a decreased quality grade (including a decrease by 2 grades and 1 grade). This indicated that the quality of the unchanged cultivated land improved from 2010 to 2018.
Spatially, the unchanged cultivated land with no change in quality grade was widely distributed throughout the entire study area. The unchanged cultivated land with improved quality (including improvement by 2 grades and 1 grade) was mainly concentrated in the eastern, northern, and southern parts of Guangzhou. The unchanged cultivated land with decreased quality (including a decrease by 2 grades and 1 grade) was mainly concentrated in the western part of Guangzhou ( Figure 6). Spatially, the unchanged cultivated land with no change in quality grade was distributed throughout the entire study area. The unchanged cultivated land w proved quality (including improvement by 2 grades and 1 grade) was mainly c trated in the eastern, northern, and southern parts of Guangzhou. The unchanged vated land with decreased quality (including a decrease by 2 grades and 1 grad mainly concentrated in the western part of Guangzhou ( Figure 6). Spatiotemporal changes of CLQ in lost/gained area (including the lost cultivate in the previous year and the gained cultivated land in the following year during the periods) are presented in Table 6. In these three different periods, the lost cultivate (cultivated land converted to non-cultivated land) was mainly medium quality an quality, accounting for about 77%, while the gained cultivated land (non-cultivate converted to cultivated land) was mainly high quality and medium quality, acco for about 72%. Spatiotemporal changes of CLQ in lost/gained area (including the lost cultivated land in the previous year and the gained cultivated land in the following year during the study periods) are presented in Table 6. In these three different periods, the lost cultivated land (cultivated land converted to non-cultivated land) was mainly medium quality and low quality, accounting for about 77%, while the gained cultivated land (non-cultivated land converted to cultivated land) was mainly high quality and medium quality, accounting for about 72%. There was large spatial heterogeneity in the quality changes of the lost/gained cultivated land. From 2010 to 2015 and from 2010 to 2018, the lost medium-quality and low-quality cultivated lands were mainly located in the southern and western parts of Guangzhou. The gained medium-and high-quality cultivated lands were mainly concentrated in the eastern and central-northern areas (Figure 7a,c). From 2015 to 2018, the lost medium-and low-quality cultivated lands were mainly located in the western and northern parts of Guangzhou. The gained medium-and high-quality cultivated lands were mainly concentrated in the eastern and southern areas (Figure 7b).

Obstacle Factors Affecting the Improvement of CLQ
Based on the obstacle factor diagnosis model, the obstacle degree of each evaluation indicator was calculated (Figure 8). The soil fertility size, high productivity capacity, and slope were the main obstacle factors to the CLQ in Guangzhou from 2010 to 2018. The sum

Obstacle Factors Affecting the Improvement of CLQ
Based on the obstacle factor diagnosis model, the obstacle degree of each evaluation indicator was calculated (Figure 8). The soil fertility size, high productivity capacity, and slope were the main obstacle factors to the CLQ in Guangzhou from 2010 to 2018. The sum of the obstacle degrees of the three indicators was about 50.0%. In addition, the obstacle degree of the stable productivity capacity gradually increased from 12.1% in 2010 to 17

CLQ Evaluation Method Integrating Multi-Source Remote Sensing
An efficient evaluation method is the basis for revealing the spatiotemporal patterns of CLQ. However, traditional evaluation methods need to acquire a large number of ground-measured soil data, which is time-consuming and laborious, and the evaluation results depend on a spatial interpolation method. Different spatial interpolation methods will affect the evaluation results of CLQ [10], and thus, the spatial information of the CLQ cannot be reflected objectively. Satellite remote sensing data have the advantages of wide coverage, abundant historical data, and objective spatial information, which are suitable for the spatiotemporal evaluation of the CLQ.
The existing abundant remote sensing data can effectively support the comprehensive evaluation of the CLQ. However, previous studies have only focused on a single attribute of the CLQ, such as soil fertility or soil moisture [35,37]. Few studies have integrated multi-source remote sensing data to comprehensively evaluate the CLQ. In this study, a comprehensive indicator system for evaluating the CLQ was developed by integrating Landsat, Sentinel-2, and MODIS NPP data products. These remote sensing data and related products satisfy the spatial scale and temporal continuity requirements to reveal the spatiotemporal changes in the CLQ.
The verification for the accuracy of the results obtained in this study showed that the results of CLQ evaluation calculated by integrating multi-source remote sensing data were reliable. Our results on the spatial distribution of CLQ were also consistent with previous studies [49]. This indicated that the CLQ evaluation method integrating remote sensing data is reasonable and feasible.

The Mechanism behind the Spatiotemporal Changes of CLQ in Guangzhou
Previous studies have emphasized that formulating a specific policy and plan is an important measure for improving the CLQ [61,62]. Guangzhou has paid considerable attention to the protection and improvement of the CLQ and has formulated and implemented a series of strict cultivated land protection policies in recent years. For example, Guangzhou city started to construct well-facilitated farmland in 2012 and released the  The obstacle degrees of the evaluation indicators differed among the 10 districts in Guangzhou. From 2010 to 2018, slope was the largest obstacle factor of CLQ in Conghua, Huangpu and Tianhe districts, with an average obstacle degree of 24.1%, 20.6% and 21.2%, respectively. High productivity capacity was the largest obstacle factor of CLQ in Nansha, Haizhu, Panyu and Liwan districts, with an average obstacle degree of 30.3%, 34.1%, 33.5% and 36.3%, respectively. Soil fertility size, slope and stable productivity capacity were the main obstacles to the CLQ in Zengcheng district, and their obstacle degrees were 16.3%, 15.4% and 15.1%, respectively. High productivity capacity and soil fertility size were the main obstacles to the CLQ in Huadu and Baiyun districts, and their obstacle degrees were 18.7%, 18.6%, 22.3% and 19.4%, respectively.

CLQ Evaluation Method Integrating Multi-Source Remote Sensing
An efficient evaluation method is the basis for revealing the spatiotemporal patterns of CLQ. However, traditional evaluation methods need to acquire a large number of groundmeasured soil data, which is time-consuming and laborious, and the evaluation results depend on a spatial interpolation method. Different spatial interpolation methods will affect the evaluation results of CLQ [10], and thus, the spatial information of the CLQ cannot be reflected objectively. Satellite remote sensing data have the advantages of wide coverage, abundant historical data, and objective spatial information, which are suitable for the spatiotemporal evaluation of the CLQ.
The existing abundant remote sensing data can effectively support the comprehensive evaluation of the CLQ. However, previous studies have only focused on a single attribute of the CLQ, such as soil fertility or soil moisture [35,37]. Few studies have integrated multi-source remote sensing data to comprehensively evaluate the CLQ. In this study, a comprehensive indicator system for evaluating the CLQ was developed by integrating Landsat, Sentinel-2, and MODIS NPP data products. These remote sensing data and related products satisfy the spatial scale and temporal continuity requirements to reveal the spatiotemporal changes in the CLQ.
The verification for the accuracy of the results obtained in this study showed that the results of CLQ evaluation calculated by integrating multi-source remote sensing data were reliable. Our results on the spatial distribution of CLQ were also consistent with previous studies [49]. This indicated that the CLQ evaluation method integrating remote sensing data is reasonable and feasible.

The Mechanism behind the Spatiotemporal Changes of CLQ in Guangzhou
Previous studies have emphasized that formulating a specific policy and plan is an important measure for improving the CLQ [61,62]. Guangzhou has paid considerable attention to the protection and improvement of the CLQ and has formulated and implemented a series of strict cultivated land protection policies in recent years. For example, Guangzhou city started to construct well-facilitated farmland in 2012 and released the Action Plan for Cultivated Land Protection in 2014 to effectively improve the CLQ (http://ghzyj.gz.gov.cn/, accessed on 1 February 2021). By the end of 2018, the construction area of the well-facilitated farmland in Guangzhou accounted for 63.58% of the total cultivated land area, which was helpful in the overall improvement of the CLQ.
At the same time, socio-economic development also promotes the improvement of the CLQ [63]. Economic growth enables governments to invest more funds in agricultural production activities [64]. Moreover, increasing subsidies to farmers can motivate them to actively protect and improve the CLQ [65]. After 2015, Guangzhou's municipal government issued a series of agricultural machinery purchases and subsidy implementation plans, which greatly improved the level of agricultural mechanization and strengthened farmer awareness of protecting and improving the CLQ.
In addition, rational planning of land use and crop planting structure is also a potential driving mechanism for improving the CLQ in Guangzhou (http://g.mnr.gov.cn/, accessed on 11 January 2021). Agronomic measures are also crucial to the improvement of CLQ. For example, increasing the application of organic fertilizer and planting green manure can improve the soil structure and enhance the ability of soil to retain fertilizer and water [66,67]. The development of agrotechnology has provided a lasting impetus for the improvement of the CLQ. High-quality seeds and cultivated land management technology adapted to local conditions can promote the continuous improvement of CLQ [17,68].

Corresponding Measures to Improve the CLQ in Guangzhou
How to improve CLQ in a targeted manner is one of the purposes of carrying out the evaluation of CLQ [10]. The obstacle factor diagnosis model can identify the main factors that affect the improvement of CLQ. In this study, the diagnosis results showed that the soil fertility size, high productivity capacity, and slope were the main obstacle factors affecting the improvement of the CLQ in Guangzhou, while the obstacle degree of the stable productivity capacity gradually increased during the study period.
Lateritic red soil is the main soil type in the cultivated land in Guangzhou. Its typical characteristics are low nutrient contents and poor ability to retain fertilizer [69], which lead to soil fertility being one of the main obstacle factors affecting the improvement of the CLQ in Guangzhou. Relevant studies have shown that stakeholders can improve soil fertility through reasonable crop rotation, application of biofertilizers, and biochar amendment [20,67,70].
The high productivity capacity had a great obstacle effect on the improvement of the CLQ, which may be the main reason that the yield per unit area of the grain crops in Guangzhou was significantly lower than that in Guangdong Province and China's national average level (Figure 9). This indicates that the production capacity of cultivated land in Guangzhou needs to be further improved. A large number of previous studies have shown that improving the utilization rates of water and fertilizer, rationally optimizing the spatial distribution of the crop production, and improving the agricultural production conditions can help increase the productivity of cultivated land [71][72][73]. Previous studies have found that the slope is closely related to CLQ; that is, the greater the slope, the worse the CLQ [30,74]. More than 70% of the cultivated land in Guangzhou is distributed in areas with slopes of > 5°, which seriously affects the CLQ in Guangzhou. Therefore, carrying out land leveling projects and ecological agricultural projects is highly recommended [75,76].
The obstacle effect of the stable productivity capacity to CLQ in Guangzhou gradually increased, which was mainly because the cultivated land experienced severe natural disasters from 2010 to 2018 ( Figure 10). Implementing conservation agriculture and strengthening crop management are effective measures for reducing the adverse impacts of natural disasters [17,77]. To ensure that cultivated land can be harvested after natural disasters, policymakers should continue to implement well-facilitated farmland construction and strengthen its maintenance and management after completion.  Previous studies have found that the slope is closely related to CLQ; that is, the greater the slope, the worse the CLQ [30,74]. More than 70% of the cultivated land in Guangzhou is distributed in areas with slopes of >5 • , which seriously affects the CLQ in Guangzhou. Therefore, carrying out land leveling projects and ecological agricultural projects is highly recommended [75,76].
The obstacle effect of the stable productivity capacity to CLQ in Guangzhou gradually increased, which was mainly because the cultivated land experienced severe natural disasters from 2010 to 2018 ( Figure 10). Implementing conservation agriculture and strengthening crop management are effective measures for reducing the adverse impacts of natural disasters [17,77]. To ensure that cultivated land can be harvested after natural disasters, policymakers should continue to implement well-facilitated farmland construction and strengthen its maintenance and management after completion. Previous studies have found that the slope is closely related to CLQ; that is, the greater the slope, the worse the CLQ [30,74]. More than 70% of the cultivated land in Guangzhou is distributed in areas with slopes of > 5°, which seriously affects the CLQ in Guangzhou. Therefore, carrying out land leveling projects and ecological agricultural projects is highly recommended [75,76].
The obstacle effect of the stable productivity capacity to CLQ in Guangzhou gradually increased, which was mainly because the cultivated land experienced severe natural disasters from 2010 to 2018 ( Figure 10). Implementing conservation agriculture and strengthening crop management are effective measures for reducing the adverse impacts of natural disasters [17,77]. To ensure that cultivated land can be harvested after natural disasters, policymakers should continue to implement well-facilitated farmland construction and strengthen its maintenance and management after completion.

Limitations and Future Work
Some potential reasons may lead to poor consistency between the remote sensing evaluation results and measured results of CLQ. First, inconsistent indicators lead to inconsistent results. In this study, the measured results were calculated based on thirteen field-measured indicators, while the remote sensing evaluation results were obtained from eight spatial attribute indicators, which may be the one reason for the inconsistency between the two results. Second, the remote sensing evaluation indicators of CLQ were more easily affected by factors such as crop varieties and climate (such as sun radiation and precipitations). In addition, one limitation of this study was that the spatial resolution of data was inconsistent. For example, the resolution of the NPP data was 500 × 500 m, that of the topsoil texture was 1 × 1 km, and that of the remaining data was 30 × 30 m. This can result in poor reliability of the evaluation results [10].
In the future, we will try our best to obtain indicators consistent with the measured data, so as to improve the accuracy of remote sensing evaluation of CLQ. The impact of crop varieties, land use and climate factors on the CLQ should be explored. Moreover, constructing a more comprehensive and representative evaluation indicator system is still the key of CLQ evaluation.

Conclusions
In this study, a new CLQ evaluation method was developed by integrating multisource remote sensing data to reveal the spatiotemporal patterns of the CLQ in Guangzhou from 2010 to 2018. The results showed that CLQ in Guangzhou continuously improved during the study period. The amount of high-quality cultivated land increased, while that of medium-and low-quality cultivated land decreased. CLQ in the eastern and northern parts of Guangzhou continuously improved, while that in the western and southern areas initially decreased and then improved. However, CLQ was low in the central region throughout the study period. Soil fertility size, high productivity capacity, and stable productivity capacity were the main obstacles for the improvement of CLQ in Guangzhou. Accordingly, we proposed a series of targeted measures, such as adopting reasonable crop rotation, strengthening crop management, and constructing well-facilitated farmland. The CLQ evaluation method proposed in this study is suitable for macro-regional scale research. The evaluation results provide clear guidance for further improvement of CLQ in Guangzhou.