Large-Scale and High-Resolution Crop Mapping in China Using Sentinel-2 Satellite Imagery

agriculture

Article Large-Scale and High-Resolution Crop Mapping in China Using Sentinel-2 Satellite Imagery

Yulin Jiang 1, Zhou Lu 2, Shuo Li 1, Yongdeng Lei 1, Qingquan Chu 1, Xiaogang Yin 1 and Fu Chen 1,*

1 College of Agronomy and Biotechnology, China Agricultural University, Key Laboratory of Farming System, Ministry of Agriculture and Rural Aﬀairs, Beijing 100193, China; [email protected] (Y.J.); [email protected] (S.L.); [email protected] (Y.L.); [email protected] (Q.C.); [email protected] (X.Y.) 2 Institute of Geographic Sciences and National Resources Research, Chinese Academy of Sciences, Beijing 100193, China; [email protected] * Correspondence: [email protected]

Received: 24 August 2020; Accepted: 25 September 2020; Published: 26 September 2020

Abstract: Large-scale, high-resolution mapping of crop patterns is useful for the assessment of food security and agricultural sustainability but is still limited. This study attempted to establish remote sensing-based crop classification models for specific cropping systems using the decision trees method and monitored the distribution of the major crop species using Sentinel-2 satellites (10 m) in 2017. The results showed that the cropping areas of maize, rice, and soybean on the Northeast China Plain were approximately 12.1, 6.2, and 7.4 million ha, respectively. The cropping areas of winter wheat and summer maize on the North China Plain were 13.4 and 16.9 million ha, respectively. The cropping areas of wheat, rice, and rape on the middle-lower Yangtze River plain were 2.2, 6.4 and 1.3 million ha, respectively. Estimated images agreed well with field survey data (average overall accuracy = 94%) and the national agricultural census data (R2 = 0.78). This indicated the applicability of the Sentinel-2 satellite data for large-scale, high-resolution crop mapping in China. We intend to update the crop mapping datasets annually and hope to guide the adjustment and optimization of the national agricultural structure.

Keywords: crop pattern; multiple indicator; classiﬁcation model; cropping system; Sentinel-2

1. Introduction Ensuring global food security is one of the greatest challenges for scientists around the world [1]. As the country with the largest population in the world, China’s food security is not only important to economic development and social stability but also to global food patterns [2,3]. China has enacted a series of measures to ensure food security during recent decades, including food self-sufficiency policies and environmentally friendly agricultural development strategies [4–6]. These policies limit the scale of water, pesticide, fertilizer, and arable land use, which increases the difficulty of new round adjustment strategies for agricultural structure [7–9]. Updated information on crop pattern mapping can provide significant scientific evidence for estimating agricultural production and food security. Mapping major crop patterns in China has great implications for policymaking in agricultural sustainability [10,11]. Despite the development of global land cover products, the specific information on crop types at large scales is still limited [12]. Additional studies should be performed to enrich crop mapping data to assist in the optimization of agricultural distribution. Earth observation satellites can be effective tools for land use mapping and crop classification due to their ability to quickly and efficiently collect information in the field [13,14]. The mapping

Agriculture 2020, 10, 433; doi:10.3390/agriculture10100433 www.mdpi.com/journal/agriculture Agriculture 2020, 10, 433 2 of 16 methods have been rapidly developed over the past 20 years [15,16]. From the methods based on single imagery [17,18] to time-series imagery [19–21], satellite mapping can fully capture temporal features. Comparative studies have proven that the methods that use multiple features are more suitable for regional mapping of areas with complicated planting structures [22]. Furthermore, the combination of remote sensing data and agricultural statistical data can realize large-scale crop pattern mapping with a relatively coarse resolution [23,24]. Future crop mapping aims to create “a map of crops” that covers more crop types and enlarged mapping areas [15,25]. On the other hand, many studies have proven that it is difficult to extract each crop type across large areas simply by applying the temporal profiles of the vegetation indices [12,26]. The combination of the multiple spectral indicators and unique signals of the target crops at specific times is necessary for large-scale, high-resolution crop pattern mapping. At present, moderate-resolution imaging spectroradiometer (MODIS) data are widely used in crop mapping because of their high revisit frequency. However, the coarse spatial resolution (250 m–1 km) of the data limits their suitability for monitoring patchy and fragmented cropland, especially in the middle and south of China [27,28]. Landsat imagery with a higher spatial resolution (30 m) has also been widely used in recent years. However, it is difficult to collect enough cloud-free images for time-series analysis because of the low temporal resolution (16 days) of the data [29]. The GF-1 satellite can provide high-spatial-resolution (16 m) data and dense time-series (4 days) to cover the key growth phases of different crops. However, the limited spectral bands of the GF-1 data cannot provide enough information for multiple classification indices because different crops in the same region may share similar spectral characteristics [30]. Since 2017, with the combination of the Sentinel-2A and 2B satellites, the ESA has provided high-resolution (10 m) images with worldwide coverage and a 5-day revisit interval. These data make it possible to significantly advance routine global monitoring of crop patterns at no cost, especially in the field of vegetation research, through enhanced definition in the red-edge spectral reflectance [31–33]. There have been some successful applications in crop classification using Sentinel-2 in recent years. For example, Immitzer et al. used Sentinel-2 images to map six summer crop species in Lower Austria and to map seven different deciduous and coniferous tree species in Germany [34]. Hegarty-Craver et al. used imagery acquired from unmanned aerial vehicles to create a high-fidelity ground-truth dataset, and then implemented the dataset for classifying cropped land using Sentinel-1 and Sentinel-2 data [35]. Zhu et al. developed a method for tea plantation detection and mapped the tea planting area in the Henan Province in China using Sentinel-2 images [36]. Despite the progress of the satellite and monitoring methods, little emphasis has been placed on large-scale crop mapping. Most efforts have focused on the classification of land cover types, such as farmland, forest, and grassland. Few large-scale mapping studies have attempted to map specific crop types and the associated land use practices at the national scale, especially for the rapid land cover changes that commonly occur in annual crop rotations [27]. Since 1997, the National Agricultural Statistics Services (NASS) of the US Department of Agriculture (USDA) has produced the annual cropland data layer (CDL), which is a raster-format, georeferenced, and crop-specific land cover map [37]. Wardlow and Egbert used time-series MODIS 250 m data for large-area crop-related LULC mapping over the US Central Great Plains [27]. However, large-area crop monitoring is difficult in cropland with multiple cropping systems or small plot sizes [10,13]. Extreme variation of cropping system occurred in China from the north to the south. The single-cropping system was widely adopted in the Northeast Plain, while the triple-cropping system was adopted in the south of China. The degree of fragmentation in agricultural fields is also totally different. Many crop pattern mapping studies in China occurred in recent years [15]. These studies tried to perform large-scale crop mapping using moderate resolution images [12] or perform high-resolution crop mapping in the typical area using different satellite or monitoring methods [11,13]. However, few studies have focused on both high-resolution and large-scale crop pattern mapping. The objective of this study was to investigate the applicability of the Sentinel-2 satellite data for large-scale, high-resolution crop mapping in the main grain-producing areas of China, to build different remote sensing-based models for crop classification in typical cropping systems that could be applied Agriculture 2020, 10, x FOR PEER REVIEW 3 of 17 Agriculture 2020, 10, 433 3 of 16 regions from 2017 to 2018, and to evaluate the classification accuracy of the crop mapping. The toprimary large areas,research to mapquest theion distribution in this study of was the as major follows: crops W inhat the thematic main grain-producing accuracy can be regionsachieved from for 2017large- toscale 2018, and and high to- evaluateresolution the crop classification pattern classification accuracy of in thedifferent crop mapping. regions with The the primary use of Sentinel research- question2 images? in this study was as follows: What thematic accuracy can be achieved for large-scale and high-resolution crop pattern classification in different regions with the use of Sentinel-2 images? 2. Materials and Methods 2. Materials and Methods 2.1. Study Area 2.1. Study Area Considering the first-order agroecological zones in China, which were defined based on the climatic,Considering soil, and the landform first-order characteristics agroecological [38 zones], we in chose China, the which Northeast were defined Plain based(NEP), on North the climatic, China soil,Plain and (NCP) landform, and middle characteristics-lower Yangtze [38], we River chose plain the Northeast(MYRP) as Plain the st (NEP),udy area North (Fig Chinaure 1). PlainThis region, (NCP), andwhich middle-lower has an area Yangtzeof approximately River plain 2 million (MYRP) km as2 the, is the study main area grain (Figure-producing1). This region,area in whichChina, hascovers an areaalmost of approximately60% of the total 2 farmland million km in2 ,China is the, mainand produces grain-producing more than area 60% in China, of the coversnational almost grain 60% yield. of theConsidering total farmland the terrain in China, complexity and produces on the more MYRP, than we60% chose of the the national plain grain area yield.as the Considering study area but the terrainmonitored complexity the entirety on the of MYRP,we the NEP choseand NCP. the plain The areacropping as the system study area in each but monitored zone was thetotally entirety different. of the NEPThe single and NCP.-cropping The cropping system was system widely in each adopted zone was in the totally NEP, different. while the The double single-cropping-cropping sy systemstem was widely adoptedadopted ininthe the NEP, NCP. while Both thethedouble-cropping double-cropping system system was and widely the triple adopted-cropping in the system NCP. Both existed the double-croppingin the MYRP. system and the triple-cropping system existed in the MYRP.

Figure 1. The spatial distribution of the elevation andand surveysurvey sitessites ((aa),), and and the the cropland cropland and and Sentinel-2 Sentinel- images2 images (b ()b in) in the the main main grain-producing grain-producing area area of of China,China, includingincluding thethe NortheastNortheast Plain (red frame), North China PlainPlain (blue(blue frame),frame), and and middle-lower middle-lower Yangtze Yangtze River River plain plain (purple (purple frame). frame). The The rectangle rectangle in thein the right-bottom right-bottom corner corner represents represents China’s China’s Nansha Nansha Islands Islands and and the the nine-dash nine-dash line. line.

2.2. Datasets 2.2. Datasets In this study, wewe collectedcollected aa totaltotal ofof 6,2666,266 Sentinel-2Sentinel-2 imagesimages (almost(almost 2.32.3 TBTB ofof computationalcomputational storage) from from February February to to October October of of 2017 2017 and and 2018, 2018, which which covered covered the the whole whole study study area area and andthe key the keyphenological phenological periods periods of the of target the target crop cropspecies species (Figure (Figure 1). All1). images All images were werecollected collected from fromthe open the- open-sourcesource data of data the of European the European Space Space Agency Agency (http://scihub.copernicus.eu) (http://scihub.copernicus.eu and) and we we set set a a 20% 20% standard to control the cloud cover. The Sentinel-2Sentinel-2 sensor provides 13 bands at a 10 10-m-m resolution and contains three bands at thethe redred edge,edge, whichwhich cancan betterbetter classifyclassify didifferentﬀerent typestypes ofof vegetation.vegetation. All data were acquired under low cloud conditions and were georeferenced to the UTM UTM-WGS84-WGS84 projection projection system. system. In total, we chose more than 100100 actual sampling points in the study area (Figure1 1a).a). TheseThese samplingsampling points were selected by local agricultural experiment stations and agricultural observation stations stations.. The training dataset was used to determine the general decision trees and to adjustadjust the thresholds in subregions. W Wee also established another dataset for accuracy evaluation. We randomly chose a total of 113 test countiescounties and 34,99534,995 sample points within the study area under the principle of uniformuniform spatial distribution.distribution.

Agriculture 2020, 10, x FOR PEER REVIEW 4 of 17

AgricultureWe2020 ,also10, 433 used the 90-m resolution SRTM DEM and SRTM SLOPE data, which were 4obtained of 16 from the Geospatial Data Cloud (http://www.gscloud.cn), as the terrain layer to assist the crop mapping research. We used 91Weitu software to choose 0.5-m resolution Google Map temporal We also used the 90-m resolution SRTM DEM and SRTM SLOPE data, which were obtained from images from February to October of 2017 and 2018 during crop-specific growing periods (crop the Geospatial Data Cloud (http://www.gscloud.cn), as the terrain layer to assist the crop mapping jointing stage or mature stage), which were used as the data source for the accuracy test. We chose research. We used 91Weitu software to choose 0.5-m resolution Google Map temporal images from national statistical data obtained from the National Statistical Bureau of China (NSBC) February to October of 2017 and 2018 during crop-specific growing periods (crop jointing stage or (http://www.stats.gov.cn) to establish the target crop species and to evaluate the classification mature stage), which were used as the data source for the accuracy test. We chose national statistical accuracy. data obtained from the National Statistical Bureau of China (NSBC) (http://www.stats.gov.cn) to establish the target crop species and to evaluate the classification accuracy. 2.3. Methodology 2.3. MethodologyWe chose multiple indicators for the crop mapping and built remote sensing-based models for cropWe chose classification multiple indicatorsin the different for the croppingcrop mapping systems. and built The remote general sensing-based technical process models included for crop the classificationfollowingin procedures the different (Fig croppingure 2): systems.The selection The general of the technical target crop process species, included data the collection following and procedurespreprocessing, (Figure 2 design): The selection of the phenology of the target-based crop indicators, species, data crop collection area mapping and preprocessing,, and accuracy designevaluation. of the phenology-based indicators, crop area mapping, and accuracy evaluation.

Figure 2. Overview of the methodology in the crop pattern mapping research. Figure 2. Overview of the methodology in the crop pattern mapping research. 2.3.1. Selecting Target Crop Species 2.3.1. Selecting Target Crop Species The cropping systems in the different zones were totally different. We used national statistical data from 2015The to calculatecropping thesystems ratio ofin thethe areasdifferent of the zones major were crops totally that weredifferent. sown We in theused different national regions statistical (Tabledata1). from We calculated 2015 to calculate nine major the kinds ratio of of crop the areaareas (rice, of the maize, major wheat, crops soybean, that were rape, sown peanut, in the cotton, different potato,regions and (Table coarse 1). cereals) We calculated and selected nine targeted major kinds crops of in crop each ar studyea (rice, area. maize, On the wheat, Northeast soybean, Plain, rape, a single-croppingpeanut, cotton, system potato is, used and duecoarse to thecereals) limited and temperatures selected targeted and rainfall crops resources. in each study Maize, area. soybean, On the andNortheast paddy rice Plain, were the a singlemain crops-cropping in this system area, accounting is used due for 98.11% to the of limited the total temperatures planting area. and On rainfall the Northresources. China Plain, Maize, most soybe farmsan, choseand paddy a winter rice wheat-summer were the main maize crops double-croppingin this area, accounting system. for The 98.11% area of of thethe wheat total planting and maize area. can On cover the 90.26% North ofChina the farmlandPlain, most in thisfarms zone. chose On a thewinter middle-lower wheat-summer Yangtze maize Riverdouble plain,-cropping both double-cropping system. The systemsarea of the and wheat multiple and cropping maize can systems cover exist 90.26% because of the of thefarmland sufficient in this climaticzone. resources. On the Paddy middle rice,-lower wheat Yangtze and rape River were major plain, crops both in double this zone,-cropping accounting systems for 93.03% and of multiple the totalcropping planting area.systems Thus, exist we because chose maize, of the soybean, sufficient and climatic paddy resources. rice as the targetPaddy crop rice, species wheat onand the rape NEP; were wheatmajor and maizecrops asin targetthis zone, crop speciesaccounting on the for NCP; 93.03% and paddyof the rice,total wheat, planting and area. rape asThus, target we crop chose species maize, on thesoybean MYRP., and paddy rice as the target crop species on the NEP; wheat and maize as target crop species on the NCP; and paddy rice, wheat, and rape as target crop species on the MYRP. Table 1. Percentage of areas of target crop species sown in diﬀerent zones in 2015. Table 1. Percentage of areas of target crop species sown in different zones in 2015. % Maize Soybean Paddy Rice Wheat Rape Total Northeast Plain% 64.95 14.43Maize Soybean 18.73 Paddy - Rice Wheat - Rape 98.11 Total North ChinaNortheast Plain Plain 40.63 -64.95 14.43 - 49.6318.73 -- 90.26- 98.11 Middle-lower Yangtze North China Plain- 40.63 61.15- 15.85- 16.0349.63 93.03- 90.26 River plain Middle-lower Yangtze River plain - 61.15 15.85 16.03 93.03

2.3.2.2.3.2. Designing Designing Phenology-Based Phenology-Based Indicators Indicators A combination of multiple spectral indicators and unique signals of targeted crops at speciﬁc times was necessary for large-scale high-resolution crop pattern mapping. Many indicators have

Agriculture 2020, 10, 433 5 of 16 been established in the application of agricultural remote sensing [15]. Considering the different cropping systems and the mixed crop density in different regions, we chose a total of seven common indicators and designed a new indicator to better classify the crop patterns by investigating the sampling points. The normalized difference vegetation index (NDVI) is widely used in to monitor vegetation coverage [39]. The NDVI time-series (∆NDVI) for the crop growing period can be used in the classification of crops and other land cover types. The normalized difference water index (NDWI) and normalized difference building index (NDBI) can distinguish between water and buildings in remote sensing images [40]. The land surface water index (LSWI) more accurately classify different crops, especially paddy rice [41]. Other indicators used in our research included the red-edge parameter (REP) and the blue-green (B-G) ratio [42,43]. We also designed a new indicator called the normalized difference rice index (NDRI), which can better identify the paddy rice on the Northeast Plain. The calculation formula of NDRI was as follows: ρswir ρred NDRI = − (1) ρswir + ρred where ρswir and ρred represent the shortwave infrared band and red band, respectively. The indicators used in the different regions are shown in Table2.

Table 2. Different indicators used for crop classification in the different study regions.

Region Crops Indicators Northeast Plain (NEP) Maize, Soybean, Rice NDVI, REP, NDRI North China Plain (NCP) Maize, Wheat NDVI, ∆NDVI Middle-lower Yangtze River plain (MYRP) Wheat, Rice, Rape NDVI, ∆NDVI, LDSI, NDBI, NDWI, B-G

2.3.3. Mapping Crop Area The choice of classification algorithm is usually based upon a series of factors, such as the availability of software, ease of use, performance, and expected overall accuracy. The decision trees (DTs) method has been widely used in land cover classification during past years, considered the advantages of being computationally fast, making no statistical assumptions, and the ability to handle data from different measurement scales [15,16]. On the other hand, software to implement DTs is readily available over the Internet. DTs construction involves the recursive partitioning of a set of training data, which is split into increasingly homogenous subsets on the basis of tests applied to one or more of the feature values. Thus, we chose the decision tree classifier for this task. We chose the univariate decision tree algorithm and built trees in each target study region considering different cropping systems. For each study region, we labelled the training sample, calculated the value of spectral indicators in specific phenological period, and compared the differences of spectral characteristics occurring in different crop species. We tried to determine the threshold manually by maximizing the dissimilarity or minimizing the similarity of the descendant nodes, and obtained a preliminary decision tree of classifications in three agroecological zones. However, according to our investigation, we found that the crop phenological features could be different even in the different provinces. The time difference of the Sentinel-2 images causes different threshold values. Building a decision tree based on a single threshold value for each study region could make a certain error. Therefore, we set 10% of value difference as the standard of subregion division and divided the Northeast Plain (NEP), North China Plain (NCP), and middle-lower Yangtze River plain (MYRP) into 13, 16, and 10 subregions, respectively, according to the differences in crop phenological features and image times (Figure3). The goal of subdivision was to better identify the threshold of decision trees and improve the overall accuracy. We built different decision trees for the crop classification of different cropping systems and slightly adjusted the threshold values in different subregions because of the varying image times and management practices. Meanwhile, for each pixel of cropland, a time series dataset of the targeted spectral indicator Agriculture 2020, 10, x FOR PEER REVIEW 6 of 17 andAgriculture slightly2020 adjusted, 10, 433 the threshold values in different subregions because of the varying image times6 of 16 and management practices. Meanwhile, for each pixel of cropland, a time series dataset of the targeted spectral indicator was developed. We implemented the specific decision tree in specific was developed. We implemented the specific decision tree in specific subregions and obtained the subregions and obtained the large-scale crop map by combining of all the subregions. large-scale crop map by combining of all the subregions.

Figure 3. Distribution of the subregions on the Northeast Plain (a), North China Plain (b), and middle-lower Figure 3. Distribution of the subregions on the Northeast Plain (a), North China Plain (b), and middle- Yangtze River plain (c). Different colors represent different subregions. lower Yangtze River plain (c). Different colors represent different subregions. 2.3.4. Accuracy Assessment 2.3.4. Accuracy Assessment In this study, we calculated the whole crop area directly by remote sensing. We evaluated the Sentinel-2-derivedIn this study, cropwe calculated maps based the on whole ground crop truth area data directly and agricultural by remote sensing. census data. We Weevaluated referenced the Sentinelthe ‘National-2-derived specifications crop maps of inspection, based on groundacceptance truth and data quality and assessment agricultural of census digital surveying data. We referencedand mapping the products’ ‘National (GB-T18316-2001) specifications of andinspection, selected acceptance more than 10%and ofquality total countiesassessment as evaluationof digital surveyingtargets. Data and from mapping a total products’ 113 survey (GB counties-T18316 with-2001) around and selected 400 sample more pointsthan 10% of each of total were counties collected. as evaluationWe randomly targets. chose Data these from test a counties total 113 and survey sample counties points with under around the principle 400 sample of uniform points of distribution, each were collected.which covered We randomly Northeast, chose North, these and test South counties China. and Wesample identified points the under crop the pattern principle through of uniform visual distribution,interpretation which using 0.5 covered m-resolution Northeast, Google North Maps, and temporal South images. China. The We spatial identified consistency the crop of the pattern crop throughpattern maps visual with interpretation the Google Maps using image 0.5 m- resultsresolution were Google assessed Maps for thetemporal test counties. images We. The calculated spatial consistenthe user’s,cy producer’s, of the crop andpattern overall maps accuracies. with the TheGoogle kappa Maps index image was results also computed were assessed [44]. for the test counties.Due toWe the calculated limitation the of fielduser’s, survey producer’s data and, and error overall of manual accuracies. visual The interpretation, kappa index an was accuracy also computedassessment [44 was]. also conducted with the national agricultural census dataset. A dataset of crop areas at the countyDue to levelthe limitation was calculated of field based survey on data remote and sensing error of derived manual cropvisual maps. interpretation, Estimated an sown accuracy areas assessmentwere then compared was also conducted to those from with the the national national agricultural agricultural census census datasets. dataset. We A conducteddataset of crop regression areas atanalysis the county for each level crop was and calculated chose the based r-squared on remote value sensing to evaluate derived the crop accuracy maps. of Estimated results. We sown set 95% areas as werethe confidence then compared coeffi cientto those and from then the tested national the p agricultural-value. census datasets. We conducted regression analysis for each crop and chose the r-squared value to evaluate the accuracy of results. We set 95% as3. the Results confidence coefficient and then tested the p-value.

3.3.1. Results Remote Sense-Based Crop Classification Models in the Typical Cropping Systems 3.1.1. Classification Model on the Northeast Plain 3.1. Remote Sense-Based Crop Classification Models in the Typical Cropping Systems Considering the limited temperature and precipitation resources, a single-cropping system was 3.1.1.widely Classification adapted on Model the NEP. on Thethe Northeast crop growing Plain season usually starts in April and ends in October. SimilarConsidering growth periods the limited for di fftemperatureerent crops increasedand precipitation the difficulty resources, of the cropa single classification-cropping withsystem the was use widelyof a single adapted indicator. on the We NEP. chose The the crop NDVI, growing REP, and season NDRI usually for this starts area, in which April canand better ends in distinguish October. Similarthe diff erentgrowth crop periods types. for We different found thatcrops using increased the value the difficulty of the NDVI of the in crop May classification and July could with easily the usesuppress of a thesingle interference indicator. of We buildings, chose the water, NDVI, and forest. REP, and The NDRIvalue of for the this REP area, in the which maize can sampling better distinguishsites during the August different was highercrop types. than thoseWe found of other that crops, using which the value could of be the used NDVI to identify in May theand maize July couldplanting easily areas. suppress We also the found interference that the NDRI of buildings, in the rice water sampling, and sitesforest. during The value August of werethe REP lower in than the maizethose insampling the maize sites and during soybean August sampling was sites, higher which than could those be of used other to identifycrops, which the paddy could rice be plantingused to identifyareas (Figure the maize4). The planting general decisionareas. We tree also on found the NEP that is the shown NDRI in Figurein the5 .rice sampling sites during August were lower than those in the maize and soybean sampling sites, which could be used to

Agriculture 2020, 10, x FOR PEER REVIEW 7 of 17

Agriculture 2020, 10, x FOR PEER REVIEW 7 of 17 identify the paddy rice planting areas (Figure 4). The general decision tree on the NEP is shown in AgricultureidentifyFigure 5. the 2020 paddy, 10, 433 rice planting areas (Figure 4). The general decision tree on the NEP is shown7 of in 16 Figure 5.

Figure 4. Comparison between the temporal profiles of the normalized difference vegetation (NDVI), Figurered-edge 4. Comparisonparameter (REP), between and thenormalized temporal difference profiles of rice the normalizedindex (NDRI) di ffforerence the different vegetation land (NDVI), cover Figure 4. Comparison between the temporal profiles of the normalized difference vegetation (NDVI), red-edgetypes on the parameter Northeast (REP), Plain. and Red normalized lines represent difference the classification rice index (NDRI) threshold. for the different land cover typesred-edge on theparameter Northeast (REP), Plain. and Red normalized lines represent difference the classification rice index (NDRI) threshold. for the different land cover types on the Northeast Plain. Red lines represent the classification threshold.

Figure 5. Decision tree for crop classification on the Northeast Plain. T1-4 represent threshold values of Figure 5. Decision tree for crop classification on the Northeast Plain. T1-4 represent threshold values each procedure, with slight differences under subregions. of each procedure, with slight differences under subregions. Figure 5. Decision tree for crop classification on the Northeast Plain. T1-4 represent threshold values 3.1.2.of Classification each procedure, Model with slig forht the differences North China under Plain subregions. 3.1.2. Classification Model for the North China Plain The NCP is a major winter wheat-summer maize double-cropping system in China. 3.1.2. Classification Model for the North China Plain FarmersThe usuallyNCP is a plant major winter winter wheat wheat from-summer October maize to double June and-cropping plant summersystem in maize China. from Farmers June tousually lateThe September. plantNCP winteris a The major wheatvalue winter of from the wheat NDVI October-summer was to highest June maize andin double the plant crop-cropping headingsummer period,system maize whichinfrom China. June can Farmers be to used late tousuallySeptember. distinguish plant The the winter value crops wheatof from the NDVI from other October landwas highest cover to types, June in the and such crop plant as heading buildings summer period, and maize water.which from Tocan diJune befferentiate used to late to betweenSeptember.distinguish the The cropsthe cropsvalue and other fromof the types other NDVI of land vegetation,was cover highest types, such in the as such forestcrop as andheading buildings grass, period, we and used water. which the ∆ TocanNDVI differentiate be betweenused to thedistinguishbetween crop heading the the crops crops and and harvest from other otherperiod, types land which of covervegetation, was types, significantly such such as as higher fo buildingsrest than and those grass,and at water. we other used To times differentiate the (Figure ΔNDVI6). Thebetween general the decision crop crops heading and tree other for and the types harvest NEP of is shown vegetation,period, in which Figure such was7 . as significantly forest and grass, higher we than used those the at ΔNDVI other betweentimes (Fig theure crop 6). The heading general and decision harvest tree period, for the which NEP iswas shown significantly in Figure higher 7. than those at other times (Figure 6). The general decision tree for the NEP is shown in Figure 7.

Agriculture 2020, 10, x FOR PEER REVIEW 8 of 17 Agriculture 2020, 10, 433 8 of 16 Agriculture 2020, 10, x FOR PEER REVIEW 8 of 17

Figure 6. Comparison between the temporal profiles of normalized difference vegetation (NDVI) and FigureNDVI time6. ComparisonComparison series (ΔNDVI) between between for the thethe temporaltemporal different prof profileslandiles cover of of normalized normalized types on thedifference diff erenceNorth vegetation China Plain. (NDVI) Red lines and NDVIrepresent time the series classification (ΔNDVI) (∆NDVI) threshold. for the didifferent fferent landland covercover typestypes onon thethe NorthNorth ChinaChina Plain.Plain. Red lines represent the classificationclassification threshold.

Figure 7. Decision tree for crop classification on the North China Plain. T1-4 represent the threshold Figure 7. Decision tree for crop classification on the North China Plain. T1-4 represent the threshold values of each procedure with slight differences among the subregions. Figurevalues of7. Decisioneach procedure tree for with crop slight classification differences on amongthe North the China subregions. Plain. T1-4 represent the threshold 3.1.3.values Classification of each procedure Model for with the slight Middle-Lower differences among Yangtze the River subregions. Plain 3.1.3. Classification Model for the Middle-Lower Yangtze River Plain Both the double-cropping system and the triple-cropping system occurred on the MYRP. 3.1.3. Classification Model for the Middle-Lower Yangtze River Plain FarmersBoth usually the double plant-cropping single rice system (from and April the to tripleOctober),-cropping early systemrice (from occurred April to on July), the MYRP. or late riceFarmers (fromBoth usually July the todouble plant October)- croppingsingle in therice summersystem (from April and and thetoplant October), triple winter-cropping early wheat rice or system rape(from in April occurred the winter. to July), on We theor chose lateMYRP. rice the FarmersSentinel-2(from July usually imagesto October) plant for late single in Aprilthe rice summer or (from early and April May plant thatto October), winter were taken wheat early during orrice rape (from the in rice Aprilthe transplanting winter. to July), We or chose period late ricethe to (fromidentifySentinel July- the2 imagesto rice October) cropping for late in area.Aprilthe summer We or foundearly and May that plant thethat value winterwere oftaken wheat the NDVIduring or rape in the April in rice the between transplanting winter. T1 We and choseperiod T2 could the to Sentinelremoveidentify - thethe2 images influencerice cropping for late of the April area. water, orWe early wheat, found May andthat tha forest. thet were value The taken of characteristics theduring NDVI the in riceof April thetransplanting NDBIbetween and T1 NDWIperiod and T2 into Aprilidentifycould wereremove the used rice the tocropping influence distinguish area. of the the We water, rice found from wheat that other, andthe types, forest.value such ofThe asthe characteristics buildings NDVI in and April of rape the between (Figure NDBI and8T1). Forand NDWI the T2 couldclassificationin Apr removeil were of usedthe winter influence to distinguish wheat of andthe water,the rape, rice thewheat from images, other and from forest. types, middle The such characteristics ofas latebuildings March andof that the rape were NDBI (Figure taken and duringNDWI8). For inthe Apr jointingclassificationil were period used of wereto winter distinguish chosen wheat in the ourand rice research.rape, from the other Theimages valuetypes, from ofsuch the middle as NDVI buildings of in late March andMarch wasrape that totally (Figure were di 8).ff takenerent For thebetweenduring classification the the jointing vegetation of periodwinter and werewheat the other chosen and land rape, in coverour the research. typesimages (Figure Thefrom value8 ).middle We of used theof late thisNDVI March characteristic in March that werewas to extract totallytaken vegetation.duringdifferent the between jointing Considering the period vegetation the were color chosen and variation the in other our in landresearch. the cover rape The flowers types value (Fig and ofure the 8). NDVI other We used crops,in March this we characteristic was found totally that differenttheto extract value between ofvegetation. the B-G the in vegetationConsidering the rape sampling and the the color other sites variation land was significantlycover in types the rape (Fig lower ureflowers than 8). We thoseand used the in this wheat other characteristic andcrops, forest. we Thistofound extract feature that vegetation. the could value be of usedConsidering the B to-G identify in the the rape thecolor sampling rape variation cropping sites in was areas.the significantlyrape We flowers also found lower and that thethan other the those value crops, in wheat of thewe foundand forest. that theThis value feature of the could B-G bein theused rape to samplingidentify the sites rape was cropping significantly areas. lower We thanalso foundthose in that wheat the and forest. This feature could be used to identify the rape cropping areas. We also found that the

Agriculture 2020, 10, 433 9 of 16 Agriculture 2020, 10, x FOR PEER REVIEW 9 of 17

LSWIvalue in the of wheatthe LSWI sampling in the siteswheat was sampling higher sites than was that higher in forest, than which that in was forest, used which to identify was used the to wheat croppingidentify areas. the wheat The general cropping decision areas. The tree general for the decision MYRP tree is shown for the inMYRP Figure is shown9. in Figure 9.

FigureFigure 8. Comparison 8. Comparison between between the the temporal temporal profilesprofiles of of the the normalized normalized difference difference vegetation vegetation (NDVI), (NDVI), normalizednormalized difference difference water water index index (NDWI), (NDWI), land land surface surface waterwater index (LSWI) (LSWI) and and blue blue-green-green ratio ratio (B- (B-G) for theG) diforff theerent different land coverland cover types types on theon the middle-lower middle-lower YangtzeYangtze River River plain. plain. Red Red lines lines represent represent the the Agriculture 2020, 10, x FOR PEER REVIEW 10 of 17 classificationclassification threshold. threshold.

Figure 9.FigureDecision 9. Decision tree fortree crop for crop classiﬁcation classification on on thethe middle-lowermiddle-lower Yangtze Yangtze River River plain. plain. T1-7 represent T1-7 represent the thresholdthe threshold values values of each of each procedure procedure with with slight slight di differencesﬀerences among among the the subregions. subregions.

3.2. Spatial Distribution of the Major Crops in the Main Grain-Producing Regions We monitored the areas of the major crop sown in the three grain-producing areas from 2017 to 2018. The results showed that the crop planting area on the NEP was mostly concentrated on the Songnen Plain, Liaodong Plain, and Sanjiang Plain. We estimated that the total areas sown with maize, rice, and soybeans on the NEP were 12.1, 6.2, and 7.4 million ha, respectively. The maize cropping area was mainly distributed in the central-western portion of the NEP, while a few areas were located on the Sanjiang Plain. The rice cropping areas were mainly concentrated on the Sanjiang Plain in the Heilongjiang Province, while a few areas were located on the Liaodong Plain. The soybean cropping area was primarily distributed in the Jilin and Heilongjiang provinces in the northern-central region of the NEP (Figure 10a–c).

Agriculture 2020, 10, 433 10 of 16

Figure 10. SpatialSpatial distribution distribution maps maps of of the the major crops on the Northeast Plain ( a–c), middle-lowermiddle-lower Yangtze River plain (d––ff),), andand NorthNorth ChinaChina PlainPlain ((gg––hh)) duringduring 20172017 andand 2018.2018. The crop planting area on the NCP covered almost the whole region except the mountains and The crop planting area on the NCP covered almost the whole region except the mountains and urban areas. We estimated that the total areas sown with winter wheat and summer maize on the NCP urban areas. We estimated that the total areas sown with winter wheat and summer maize on the were 13.4 and 16.9 million ha, respectively. The winter wheat cropping area was mainly distributed in NCP were 13.4 and 16.9 million ha, respectively. The winter wheat cropping area was mainly the piedmont region of the Taihang Mountain, Luxi Plain, and Huang-Huai Plain. Few areas of winter distributed in the piedmont region of the Taihang Mountain, Luxi Plain, and Huang-Huai Plain. Few wheat were planted in the Fenhe Valley in the western part of the NCP. The maize cropping area on areas of winter wheat were planted in the Fenhe Valley in the western part of the NCP. The maize the NCP covered almost the whole area, including the Jiaodong Peninsula and some mountain areas cropping area on the NCP covered almost the whole area, including the Jiaodong Peninsula and some (Figure 10g–h). mountain areas (Figure 10g–h). Rice is a major crop on the MYRP, including early rice, single rice, and late rice. We estimated Rice is a major crop on the MYRP, including early rice, single rice, and late rice. We estimated that the total areas sown with rice, winter wheat, and winter rape on the MYRP were 6.4, 2.2, and 1.3 that the total areas sown with rice, winter wheat, and winter rape on the MYRP were 6.4, 2.2, and 1.3 million ha, respectively. The rice cropping areas were mostly distributed in the northern part of the million ha, respectively. The rice cropping areas were mostly distributed in the northern part of the MYRP and the plains areas in the south. The winter wheat planting area was mostly concentrated MYRP and the plains areas in the south. The winter wheat planting area was mostly concentrated in the northern part of the MYRP, while the rape planting area was distributed in the western and southern areas of the MYRP (Figure 10d–f).

3.3. Spatial Agreement with the Google Maps Image Results We evaluated the crop map estimated from the Sentinel-2 imagery based on the 0.5-m resolution Google Maps image results for the test counties. The results showed that the Sentinel-estimated crop map was consistent with the interpreted images from Google Maps, especially on the NCP. The overall accuracy on the NEP was 0.93 with a kappa index of 0.90 (Table 3). The identification accuracy for maize was relatively high in comparison with those of the other crops. The high accuracy occurred on the NCP with an overall accuracy of 0.96 and a kappa index of 0.92 because of the low interference from the other crop types, especially in the winter wheat cropping areas (Table 4). We achieved a relatively high accuracy on the MYRP considering the patchy and fragmented cropland in this area.

Agriculture 2020, 10, 433 11 of 16 in the northern part of the MYRP, while the rape planting area was distributed in the western and southern areas of the MYRP (Figure 10d–f).

3.3. Spatial Agreement with the Google Maps Image Results We evaluated the crop map estimated from the Sentinel-2 imagery based on the 0.5-m resolution Google Maps image results for the test counties. The results showed that the Sentinel-estimated crop map was consistent with the interpreted images from Google Maps, especially on the NCP. The overall accuracy on the NEP was 0.93 with a kappa index of 0.90 (Table3). The identiﬁcation accuracy for maize was relatively high in comparison with those of the other crops. The high accuracy occurred on the NCP with an overall accuracy of 0.96 and a kappa index of 0.92 because of the low interference from the other crop types, especially in the winter wheat cropping areas (Table4). We achieved a relatively high accuracy on the MYRP considering the patchy and fragmented cropland in this area. The overall accuracy on the MYRP was 0.93, with a kappa index of 0.90 (Table5). The results fully demonstrate the advantage of high-resolution images in crop mapping research in areas with complex land features.

Table 3. Accuracy assessment using the interpreted images from Google Maps for the NEP.

Maize Soybean Rice Others User Accuracy Maize 4557 103 24 180 0.94 Soybean 92 2800 69 148 0.90 Rice 73 37 1269 40 0.89 Others 159 116 14 4506 0.94 Producer 0.93 0.92 0.92 0.92 accuracy Overall 0.93 accuracy Kappa 0.90

Table 4. Accuracy assessment using the interpreted images from Google Maps for the NCP.

Winter Wheat Summer Maize Others User Accuracy Winter wheat 4635 - 125 0.97 Summer maize - 5670 305 0.95 Others 356 464 18557 0.96 Producer accuracy 0.93 0.92 0.98 Overall accuracy 0.96 Kappa 0.92

Table 5. Accuracy assessment using the interpreted images from Google Maps for the MYRP.

Wheat Rape Rice Others User Accuracy Wheat 2686 24 - 203 0.92 Rape 45 2961 - 81 0.96 Rice - - 3661 175 0.95 Others 193 150 294 5110 0.89 Producer 0.92 0.94 0.93 0.92 accuracy Overall 0.93 accuracy Kappa 0.90

3.4. Evaluation with Agricultural Statistical Data Due to the limitations of the field survey data, we also compared the remote sensing-estimated crop areas with those from the national agricultural census data at the county level. These two datasets Agriculture 2020, 10, 433 12 of 16 agreed reasonably well, with an overall r-squared value of 0.78. The confidence coefficient of all models could reach the 99% level. The coefficients of determination between the Sentinel-estimated dataset and the agricultural census data at the county level on the NEP, NCP, and MYRP were 0.83, 0.81, and 0.71, respectively (Figure 11). Considerable agreement was obtained for the maize and rice on the NEP and the wheat in the NCP, with r-squared values higher than 0.85. The accuracy of the estimated crop area on the MYRP was relatively low because of the interference of cloud and mountain elements. Fragmented cropland on the MYRP also increased the difficulty of the crop mapping, especially in the areasAgriculture sown 2020 with, 10 rape., x FOR PEER REVIEW 13 of 17

FigureFigure 11. 11.Correlations Correlations of of areas areas sown sown with with di ﬀdifferenterent crop crop species species between between the the Sentinel-2-estimated Sentinel-2-estimated and nationaland national census census data on data the on Northeast the Northeast Plain (Plaina–c), ( middle-lowera–c), middle-lower Yangtze Yangtze River River plain plain (d–f ),(d and–f), and North ChinaNorth Plain China (g– Plainh). (g–h).

4. Discussion4. Discussion ThisThis study study contributed contributed to to large-scale large-scale cropcrop mappingmapping using high high-resolution-resolution satellite satellite images images and and multiplemultiple indicators indicators in China.in China. Previous Previous research research has has mapped mapped the thespatiotemporal spatiotemporal dynamics dynamics of the of maize the andmaize cropping and cropping intensity intensity trends in trends China in atChina the national at the national scale by scale using by MODISusing MODIS (500 m) (500 images m) images [12,45 ]. Moderate-resolution[12,45]. Moderate-resolution images can images achieve can coverage achieve coverageof large areas of large and areas make and large-scale make lar monitoringge-scale possible.monitoring However, possible. limited However, resolution limited and resolution indicator and systems indicator could systems cause could errors cause in crop errors distribution in crop mappingdistribution research mapping [30]. research Multiple [30]. indicator Multiple systems indicator have systems been widely have been used widely in local-scale used in local studies-scale [46 ]. It isstudies possible [46]. to It achieve is possible almost to achieve completely almost accurate completely results accurate for small results areas for with small the areas development with the of quantitativedevelopment monitoring of quantitative technologies monitoring for technologies remote sensing. for remote However, sensing. few However, studies have few focusedstudies have on the extensivefocused use on ofthe multiple extensive indicator use of multiple systems indicator in largeareas. systems On in the large basis areas. of previous On the basis studies, of previous we created a practicalstudies, we application created a for practical crop mapping application research for crop at the mapping national research scale with at the a combination national scale of with multiple a indicatorcombination systems of multiple and high-resolution indicator systems satellite and images. high-resolution satellite images. High-resolution satellite data have played an important role in remote sensing-based agricultural monitoring. The cost of large-scale use of high-resolution imagery is prohibitive. Some recent studies have monitored crop distributions in China with data from the GF-1 satellite, which was launched by China to obtain high-resolution imagery (i.e., 2 m/8 m/16 m) and a short revisit cycle (i.e., 4 days). However, limited wave bands (four bands) increase the difficulty of crop classification, especially in ferruginous plots [28]. This study is also a pioneering research to prove that the Sentinel- 2 satellites could be useful for large-scale, high-resolution crop mapping research in China. We realized relatively high accuracy on the middle-lower Yangtze River plain, which has complex land cover features and fragmented cropland. The results fully demonstrated that the Sentinel-2 satellites,

Agriculture 2020, 10, 433 13 of 16

High-resolution satellite data have played an important role in remote sensing-based agricultural monitoring. The cost of large-scale use of high-resolution imagery is prohibitive. Some recent studies have monitored crop distributions in China with data from the GF-1 satellite, which was launched by China to obtain high-resolution imagery (i.e., 2 m/8 m/16 m) and a short revisit cycle (i.e., 4 days). However, limited wave bands (four bands) increase the difficulty of crop classification, especially in ferruginous plots [28]. This study is also a pioneering research to prove that the Sentinel-2 satellites could be useful for large-scale, high-resolution crop mapping research in China. We realized relatively high accuracy on the middle-lower Yangtze River plain, which has complex land cover features and fragmented cropland. The results fully demonstrated that the Sentinel-2 satellites, which have high resolutions, low revisit periods, multiple wave bands, and the advantage of low cost, can easily provide images for specific times and multiple indicator systems, which could be widely used in large-scale classification of land cover types. We achieved relatively high classification accuracy compared with the interpreted Google Maps images. Though the temporal difference between Google Maps and Sentinel-2 images may bring some errors, cropping system in the study area would not change frequently. Thus, we did not take this kind of error into consideration. Visual interpretation of Google Maps would also cause some mistakes considering the different individual standard, especially in some blurry areas. We marked these challenge sample points and made the final decision through team discussion in order to reduce the human error in this part. However, the accuracy assessment with the county-level national census data was unsatisfactory. Previous studies of large-scale crop mapping have usually conducted an accuracy assessment with census data at the provincial level [45]. The use of provincial data could ignore some errors in local areas and reach relatively high accuracy levels. We conducted an accuracy assessment at the county level, which could enlarge the discrete degree to a certain extent. On the other hand, national census data at the county level have hysteresis characteristics and are easily affected by human factors, which could cause errors in the accuracy assessment. Cloud cover is also an uncertain factor for monitoring studies. A higher cloud cover ratio indicated higher uncertainties in the estimated results. The lack of images for specific times during the crop growing period increases the difficulty of crop classification. For example, entire crop cycles might be missed due to persistent cloud cover [47]. In this study, there is limited influence caused by cloud cover because of the short revisit period of Sentinel-2 satellite. We could almost obtain more than three available images in a single month, especially in NEP and NCP, while we could almost obtain at least one available image in a single month in MYRP. On the other hand, although we set a 20% standard to control cloud cover when downloading the images, missing images in some areas invalidated our decision model, especially in the critical crop growing period. We calculated the whole cloud area in our research and found that cloud area occurring in the specific growing period was less than 5% of the whole area, which could cause certain errors for the detection of the true situation. Large-scale and high-resolution remote sensing monitoring is significant for the adjustment and optimization of agricultural production. Understanding the spatial and temporal variations of crop patterns is important for the creation of policy, assessment of food security, and development of sustainable agricultural practices [8,48]. We found that the sown area of the winter wheat on the NCP was much lower than that of the summer maize, especially in the Hebei Province. This result was also much lower than previous crop mapping research in 2009 [49]. The results showed that the crop rotation and fallow policy have been properly implemented because of groundwater overexploitation on the NCP [50]. Building remote sensing-based models in China for multiple cropping systems is also meaningful for nationwide crop mapping research [49]. Creating and periodically updating different crop distribution datasets is meaningful for estimating crop yield, analyzing changes in cropping intensity and adapting agriculture to climate change [45]. We estimated the main crop distribution in the major grain-producing areas of China and achieved relatively high accuracy in the plain areas. However, it is difficult to ensure accuracy for crop monitoring in mountain areas. We used 90-m resolution SRTM DEM and SRTM SLOPE data to remove Agriculture 2020, 10, 433 14 of 16 the mountain areas and assumed that few crops were distributed in the mountains. Some crops could be planted in the mountain areas with relatively low yields due to limitations of the landscape, especially on the middle-lower Yangtze River plain. We could focus on the classification of crops in mountain areas and build a reasonable model for improving the accuracy of crop mapping in the future. We could also refine the crop types (i.e., spring maize/summer maize) and build a classification method for other field crops in the future, such as cotton and potato. On the other hand, the decision trees method is an ordinary but efficient method in crop classification research [15,16]. There are also some alternative methods based on machine learning or deep learning in this field. For example, the random forest algorithm has been widely used in recent years. It is an ensemble learning algorithm that consists of multiple decision trees or classified regression trees and can reduce error caused by artificial subjective factor to a certain extent [34,36]. In the future, we will conduct comparative studies between the different classification methods and optimize our algorithm in different cropping systems. We built an open-access dataset on the internet (www.cross.ac.cn) to share the crop monitor results at county level and intend to update it annually. We hope our results can guide the adjustment and optimization of the national agricultural structure in the future.

5. Conclusions This study built three remote sensing-based models for typical cropping systems using multiple indicators and monitored the distribution of the major crops in the main grain-producing area of China using Sentinel-2 satellite data. We found that the areas sown with maize, rice, and soybean on the NEP were 12.1, 6.2 and 7.4 million ha, respectively. The maize and wheat planting areas on the NCP were 16.9 and 13.4 million ha, respectively. The total sown areas of the rice, wheat, and rape on the MYRP were 6.4, 2.2, and 1.3 million ha, respectively. The estimated images agreed well with ﬁeld survey data (average overall accuracy = 94%) and the national agricultural census data (R2 = 0.78). This proves the applicability of the Sentinel-2 satellite data for large-scale, high-resolution crop mapping in China. We intend to update the crop mapping datasets annually and hope to guide the adjustment and optimization of the national agricultural structure.

Author Contributions: Conceptualization and design, Y.J. and Z.L.; methodology, Y.J.; software, Z.L.; formal analysis, Y.J.; resources, S.L.; writing—original draft preparation, Y.J.; writing—review and editing, Y.L. and Q.C.; supervision, X.Y.; project administration, F.C.; All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by the National Key Research and Development Program of China, grant number (2016YFD0300201). Acknowledgments: We thank the research team at the Institute of Geographic Sciences and National Resources Research for their careful guidance and thank all reviewers and editors for their suggestions, which substantially improved the manuscript. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. He, G.; Zhao, Y.; Wang, L.; Jiang, S.; Zhu, Y. China’s food security challenge: Eﬀects of food habit changes on requirements for arable land and water. J. Clean. Prod. 2019, 229, 739–750. [CrossRef] 2. Wang, Q.; Liu, X.; Yue, T.; Wang, C.; Wilson, J.P. Using models and spatial analysis to analyze spatio-temporal variations of food provision and food potential across China’s agro-ecosystems. Ecol. Model. 2015, 306, 152–159. [CrossRef] 3. Wang, Y.; Wu, P.; Engel, B.; Sun, S. Comparison of volumetric and stress-weighted water footprint of grain products in China. Ecol. Indic. 2015, 48, 324–333. [CrossRef] 4. Kang, Y. Food safety governance in China: Change and continuity. Food Control. 2019, 106, 106752. [CrossRef] 5. Yang, H.; Liu, J.; Savenije, H. China’s move to higher-meat diet hits water security. Nature 2008, 454, 397. 6. Zang, Z.; Zou, X.; Xi, X.; Zhang, Y.; Zheng, D.; Sun, C. Quantitative characterization and comprehensive evaluation of regional water resources using the three red lines method. J. Geogr. Sci. 2016, 26, 397–414. [CrossRef] Agriculture 2020, 10, 433 15 of 16

7. Feng, Z.; Yang, Y.; Zhang, Y.; Zhang, P.; Li, Y. Grain-for-green policy and its impacts on grain supply in West China. Land Use Policy 2005, 22, 301–312. [CrossRef] 8. Liu, Z.; Yang, P.; Wu, W.; You, L. Spatiotemporal changes of cropping structure in China during 1980–2011. J. Geogr. Sci. 2018, 28, 1659–1671. [CrossRef] 9. Xie, H.; Cheng, L.; Lu, H. Farmers’ responses to the winter wheat fallow policy in the groundwater funnel area of China. Land Use Policy 2018, 73, 195–204. [CrossRef] 10. Xiao, X.; Boles, S.; Liu, J.; Zhuang, D.; Frolking, S.; Li, C.; Salas, W.; Moore, B. Mapping paddy rice agriculture in Southern China using multi-temporal MODIS images. Remote Sens. Environ. 2005, 95, 480–492. [CrossRef] 11. Yin, Q.; Liu, M.; Cheng, J.; Ke, Y.; Chen, X. Mapping paddy rice planting area in Northeastern China using spatiotemporal data fusion and phenology-based method. Remote Sens. 2019, 11, 1699. [CrossRef] 12. Qiu, B.; Huang, Y.; Chen, C.; Tang, Z.; Zou, F. Mapping spatiotemporal dynamics of maize in China from 2005 to 2017 through designing leaf moisture-based indicator from normalized multi-band drought index. Comput. Electron. Agric. 2018, 153, 82–93. [CrossRef] 13. Jia, K.; Wu, B.; Li, Q. Crop classification using HJ satellite multispectral data in the North China plain. J. Appl. Remote Sens. 2013, 7, 73576. [CrossRef] 14. Vintrou, E.; Desbrosse, A.; Bégué, A.; Traoré, S.; Baron, C.; Seen, D. Crop area mapping in West Africa using landscape stratification of MODIS time series and comparison with existing global land products. Int. J. Appl. Earth Obs. Geoinf. 2012, 14, 83–93. [CrossRef] 15. Hu, Q.; Wu, W.; Song, Q.; Yu, Q.; Yang, P.; Tang, H. Recent progress in research of crop patterns mapping by using remote sensing. Sci. Agric. Sin. 2015, 48, 1900–1914. 16. Tang, H.; Wenbin, W.; Yang, P. Recent progresses in monitoring crop spatial patterns by using remote sensing technologies. Sci. Agric. Sin. 2010, 43, 2879–2888. 17. Mathur, A.; Foody, G. Crop classification by support vector machine with intelligently selected training data for an operational application. Int. J. Remote Sens. 2008, 29, 2227–2240. [CrossRef] 18. Yang, C.; Everitt, J.; Murden, D. Evaluating high resolution SPOT 5 satellite imagery for crop identification. Comput. Electron. Agric. 2011, 75, 347–354. [CrossRef] 19. Cai, Y.; Guan, K.; Peng, J.; Wang, S.; Seifert, C.; Wardlow, B.; Li, Z. A high-performance and in-season classification system of field-level crop types using time-series landsat data and a machine learning approach. Remote Sens. Environ. 2018, 210, 35–47. [CrossRef] 20. Chang, J.; Hansen, M.C.; Pittman, K.; Carroll, M.; DiMiceli, C. Corn and soybean mapping in the United States using MODIS time-series data sets. Agron. J. 2007, 99, 1654. [CrossRef] 21. Thenkabail, P.; Biradar, C.; Noojipady, P.; Dheeravath, V.; Li, Y.; Velpuri, M.; Gumma, M.; Gangalakunta, O.; Turral, H.; Cai, X.; et al. Global irrigated area map (GIAM), derived from remote sensing, for the end of the last millennium. Int. J. Remote Sens. 2009, 30, 3679–3733. [CrossRef] 22. Biradar, C.; Thenkabail, P.; Noojipady, P.; Li, Y.; Dheeravath, V.; Turral, H.; Velpuri, M.; Gumma, M.; Gangalakunta, O.; Cai, X.; et al. A global map of rainfed cropland areas (GMRCA) at the end of last millennium using remote sensing. Int. J. Appl. Earth Obs. Geoinf. 2009, 11, 114–129. [CrossRef] 23. Leff, B.; Ramankutty, N.; Foley, J.A. Geographic distribution of major crops across the world. Glob. Biogeochem. Cycles 2004, 18, GB1009. [CrossRef] 24. Monfreda, C.; Ramankutty, N.; Foley, J.A. Farming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000. Glob. Biogeochem. Cycles 2008, 22, GB1022. [CrossRef] 25. Wu, B.; Li, Q. Crop planting and type proportion method for crop acreage estimation of complex agricultural landscapes. Int. J. Appl. Earth Obs. Geoinf. 2012, 16, 101–112. [CrossRef] 26. Qiu, B.; Luo, Y.; Tang, Z.; Chen, C.; Lu, D.; Huang, H.; Chen, Y.; Chen, N.; Xu, W. Winter wheat mapping combining variations before and after estimated heading dates. ISPRS J. Photogramm. Remote Sens. 2017, 123, 35–46. [CrossRef] 27. Wardlow, B.; Egbert, S. Large-area crop mapping using time-series MODIS 250 m NDVI data: An assessment for the U.S. Central Great Plains. Remote Sens. Environ. 2008, 112, 1096–1116. [CrossRef] 28. Zhou, Q.; Yu, Q.; Liu, J.; Wu, W.; Tang, H. Perspective of Chinese GF-1 high-resolution satellite data in agricultural remote sensing monitoring. J. Integr. Agric. 2017, 16, 242–251. [CrossRef] 29. Song, Q.; Zhou, Q.; Wu, W.; Hu, Q.; Lu, M.; Liu, S. Mapping regional cropping patterns by using GF-1 WFV sensor data. J. Integr. Agric. 2017, 16, 337–347. [CrossRef] Agriculture 2020, 10, 433 16 of 16

30. Li, H.; Chen, Z.; Jiang, Z.; Wu, W.; Ren, J.; Liu, B.; Tuya, H. Comparative analysis of GF-1, HJ-1, and landsat-8 data for estimating the leaf area index of winter wheat. J. Integr. Agric. 2017, 16, 266–285. [CrossRef] 31. Cabezas-Rabadán, C.; Pardo-Pascual, J.E.; Palomar-Vázquez, J.; Fernández-Sarría, A. Characterizing beach changes using high-frequency sentinel-2 derived shorelines on the valencian coast (Spanish mediterranean). Sci. Total Environ. 2019, 691, 216–231. [CrossRef][PubMed] 32. Griffiths, P.; Nendel, C.; Hostert, P. Intra-annual reflectance composites from sentinel-2 and landsat for National-scale crop and land cover mapping. Remote Sens. Environ. 2019, 220, 135–151. [CrossRef] 33. Sánchez-Espinosa, A.; Schröder, C. Land use and land cover mapping in wetlands one step closer to the ground: Sentinel-2 versus landsat 8. J. Environ. Manag. 2019, 247, 484–498. [CrossRef][PubMed] 34. Immitzer, M.; Vuolo, F.; Atzberger, C. First experience with Sentinel-2 data for crop and tree species classification in Central Europe. Remote Sens. 2016, 8, 166. [CrossRef] 35. Hegarty-Craver, M.; Polly, J.; O’Neil, M.; Ujeneza, N.; Rineer, J.; Beach, R.; Lapidus, D.; Temple, D. Remote crop mapping at scale: Using satellite imagery and UAV-acquired data as ground truth. Remote Sens. 2020, 12, 1984. [CrossRef] 36. Zhu, J.; Pan, Z.; Wang, H.; Huang, P.; Sun, J.; Qin, F.; Liu, Z. An improved multi-temporal and multi-feature tea plantation identification method using Sentinel-2 imagery. Sensors 2019, 19, 2087. [CrossRef] 37. Boryan, C.; Yang, Z.; Mueller, R.; Craig, M. Monitoring US agriculture: The US department of agriculture, National agricultural statistics service, cropland data layer program. Geocarto Int. 2011, 26, 341–358. [CrossRef] 38. Liu, X.; Chen, F. Farming Systems in China; China Agriculture Press: Beijing, China, 2005. 39. Wu, W.; Yang, P.; Tang, H.; Zhou, Q.; Chen, Z.; Shibasaki, R. Characterizing spatial patterns of phenology in cropland of China based on remotely sensed data. Agric. Sci. China 2010, 9, 101–112. [CrossRef] 40. Firozjaei, K.; Alavipanah, K.; Liu, H.; Sedighi, A.; Mijani, N.; Kiavarz, M.; Weng, Q. A PCA–OLS model for assessing the impact of surface biophysical parameters on land surface temperature variations. Remote Sens. 2019, 11, 2094. [CrossRef] 41. Chang, Q.; Xiao, X.; Jiao, W.; Wu, X.; Doughty, R.; Wang, J.; Du, L.; Zou, Z.; Qin, Y. Assessing consistency of spring phenology of snow-covered forests as estimated by vegetation indices, gross primary production, and solar-induced chlorophyll fluorescence. Agric. For. Meteorol. 2019, 275, 305–316. [CrossRef] 42. Carbas, B.; Machado, N.; Oppolzer, D.; Ferreira, L.; Brites, C.; Rosa, E.; Barros, A. Comparison of near-infrared (NIR) and mid-infrared (MIR) spectroscopy for the determination of nutritional and antinutritional parameters in common beans. Food Chem. 2020, 306, 125509. [CrossRef][PubMed] 43. Zhang, M.; Su, W.; Fu, Y.; Zhu, D.; Xue, J.; Huang, J.; Wang, W.; Wu, J.; Yao, C. Super-resolution enhancement of sentinel-2 image for retrieving LAI and chlorophyll content of summer corn. Eur. J. Agron. 2019, 111, 125938. [CrossRef] 44. Congalton, R. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sens. Environ. 1991, 37, 270–279. [CrossRef] 45. Qiu, B.; Lu, D.; Tang, Z.; Song, D.; Zeng, Y.; Wang, Z.; Chen, C.; Chen, N.; Huang, H.; Xu, W. Mapping cropping intensity trends in China during 1982–2013. Appl. Geogr. 2017, 79, 212–222. [CrossRef] 46. Qiu, B.; Lu, D.; Tang, Z.; Chen, C.; Zou, F. Automatic and adaptive paddy rice mapping using landsat images: Case study in songnen plain in Northeast China. Sci. Total Environ. 2017, 598, 581–592. [CrossRef] 47. Gray, J.; Friedl, M.; Frolking, S.; Ramankutty, N.; Nelson, A.; Gumma, M. Mapping Asian cropping intensity with MODIS. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2014, 7, 3373–3379. [CrossRef] 48. Tang, H.; Wu, W.; Yang, P.; Li, Z. Systematic synthesis of impacts of climate change on China’s crop production system. J. Integr. Agric. 2014, 13, 1413–1417. [CrossRef] 49. Tao, J.; Wu, W.; Zhou, Y.; Wang, Y.; Jiang, Y. Mapping winter wheat using phenological feature of peak before winter on the North China plain based on time-series MODIS data. J. Integr. Agric. 2017, 16, 348–359. [CrossRef] 50. Jiang, Y.; Wang, X.; Ti, J.; Yin, X.; Lei, Y.; Chu, Q.; Chen, F. Winter wheat water-saving potential in groundwater overexploitation district of North China plain. Agron. J. 2020, 112, 44–55. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).