The Guanzhong area is located in the middle of Shaanxi Province, ranging from 34°31′-35°35′N to 106°36′-110°32′E (Figure 1). It has a maximum height of 3654 m in its southwest and a minimum height of 308 m on the plain. The Weihe River runs through the Guanzhong area from west to east and converges into the Yellow River. The Guanzhong area is formed mainly by river alluvium and loess accumulation. The annual average temperature ranges from 6℃ to 13℃, and the annual precipitation is between 500 mm and 800 mm. With its flat terrain, mild climate, fertile soil, and abundant water resources, Guanzhong is the most conducive area for machinery farming and irrigation in Shaanxi. As an important grain producing area in Shaanxi Province, the cropland in the Guanzhong area currently accounts for more than 30% of the land area. The extensive croplands are mainly found in the Guanzhong Basin, located in the central and eastern parts of the Guanzhong area. The Guanzhong Basin primarily consists of plains and loess tablelands. Woodland and grassland are mostly found in the high-altitude areas around croplands, including the west, southwest, and north of the study area (Liu et al., 2010).

From the Qing Dynasty (1644-1911) to the present era, the administrative boundaries have been changed several times. In the Qing Dynasty, the study area was included in six prefectures (Xi’an, Fengxiang, Tongzhou, Qianzhou, Binzhou, and Fuzhou). By 1911, the Guanzhong area was governed by a higher administrative unit, called Guanzhong dao. After 1949, the names and jurisdictions of the prefecture-level administrative units in the Guanzhong area changed significantly. The boundaries of counties did not change significantly, but there were cancellations, mergers, and renaming of counties (Fu and Zheng, 2007; Fu et al., 2013).

2.2.1 Cropland data

The cropland data sources represented three periods: the Qing Dynasty (1644-1911), the Republic of China (1911-1949), and after 1949, covering the period from 1650-2016. Most data sources provide cropland data at the county level (Table 1).

Cropland data in the Qing Dynasty were obtained from gazetteers. Based on the book A General Catalogue of Gazetteers in China (BAOCAS, 1985), there are approximately 198 volumes of gazetteers on the Guanzhong area from 1644 to 1949. We collected 161 volumes of gazetteers from the National Digital Library of China (http://www.nlc.gov.cn/) as well as the books Collection of Chinese Gazetteers (PPH, 2007) and Series of Chinese Gazetteers (CWP, 1966), including county, prefectural, and provincial gazetteers. Since the records of the gazetteers in the Republic of China were mostly the continuation of those in the Qing Dynasty, the Republic of China’s gazetteers were also essential sources of cropland data in the Qing Dynasty. Gazetteers recorded tax-paying cropland data and tax revenue before the year of publication, which was vital information about the cropland area of each county. Gazetteers of different levels of administrative units had different ways of recording tax-paying cropland data. County gazetteers recorded the size of various tax-paying cropland types and their changes over time. Prefectural and provincial gazetteers mostly summarized county gazetteers’ records and only roughly recorded each county or prefecture’s total tax-paying cropland area.

For the period of the Republic of China, cropland data were obtained from two sources. The book The Fifth Agriculture and Commerce Statistics Table (MAC, 1919) recorded the number of peasant households and gardens, planting area of crops, amount of cropland area, abandoned cropland area, and cropland area affected by disasters in each county. The cropland area data collected from this book are statistics. Another source is survey data from the book of Land Utilization in China (Buck, 1937). Funded by the Institution of Pacific Relations, Buck investigated the land use of China from 1929 to 1933 and published this book based on the investigation of land area, cropland area, planting area, etc. Compared with statistics from The Fifth Agriculture and Commerce Statistics Table, Buck’s survey data on cropland are more scientific and closer to each county’s “real value”.

For the period after 1949, we collected both statistics and survey data around the 1980s and 2016. In the 1980s, statistics and survey data were sourced from Summary of Rural Economic Statistics by County in China (NBS, 1989) and Datasets of Land and Resources of China (CISNR, 1989), respectively. The statistics for 2016 were obtained from the Chinese Statistical Yearbooks (https://data.cnki.net/Yearbook/) of each prefecture in the Guanzhong area. Survey data in 2010 and 2016 were acquired from The Second National Land Survey (http://tddc.mnr.gov.cn/). We also used the total cropland area of the Guanzhong area around 1950 from the study of Liu (2015) and the data in 2010 to analyze the temporal changes in the total cropland area.

2.2.2 Administrative boundaries

The digital administrative division maps were obtained from three sources. The administrative division map at the county level in 1911 was obtained from The China Historical Geographic Information System (CHGIS; http://yugong.fudan.edu.cn), whereas those in 1980 and 2015 were sourced from the National Earth System Science Data Center (http://www.geodata.cn/) and National Catalogue Service for Geographic Information (http://www.webmap.cn/main.do?method=index), respectively.

2.2.3 Geographical conditions

We also used spatial data of land use, digital elevation model (DEM), climate, and river. The satellite-based land-use data in 1980 and 2000 and the climatic potential productivity value were provided by the National Earth System Science Data Center (http://www.geodata.cn/). The 1 km×1 km DEM used in this study was from the Resource and Environment Data Cloud Platform (http://www.resdc.cn/). The digital river data in 1820 was obtained from CHGIS.

The outline of the relationship between different cropland data sources and the methods for the reconstruction of cropland area change from 1650 to 2016 in the Guanzhong area is summarized in Figure 2. Based on gazetteers, we used methods of registered tax-paying cropland data collection, time points selection, missing data interpolation, and data calibration to obtain each county’s cropland area before 1910. The cropland area from statistics was compared with survey data and calibrated. Then, the cropland area data with a uniform measurement standard at the county level was produced. The main factors (altitude, slope, climate, and rivers) affecting the spatial distribution of cropland were selected, and cropland area data at the county level were allocated to 1 km×1 km grid cells.

2.3.1 Registered cropland data collection

In the Qing Dynasty, the Manchu government divided cropland into multiple forms for taxation. Different forms of tax-paying land paid different amounts of tax per unit area. There were four common forms of registered tax-paying land in the gazetteers in the Guanzhong area, including civilian (Min Di), soldier-cultivated (Tun Di), renamed (Gengming Di), and education (Xue Tian) lands. Civilian land referred to tax-paying land cultivated by private civilians. In the Ming Dynasty (1368-1644), the government established military departments (Wei/Suo) and provided them some land for cultivation to supply food for local troops, which was called soldier-cultivated land. In the early Qing Dynasty, several military departments were disbanded. Their soldier-cultivated land was changed to tax-paying land and allocated to private civilians. Renamed land was owned by royal family members and officials in the Ming Dynasty. At the beginning of the Qing Dynasty, Manchu warriors overthrew the Ming Dynasty, and the government gave renamed land to private civilians to cultivate free of charge. Still, civilians paid taxes after gaining income from farming. Education land was used for supporting schools, but it was cultivated by civilians who rent it. These four common forms of registered tax-paying land were cropland. Thus, their areas were all included in the total cropland area of each county.

There were many records about the change in the administrative ownership of the soldier-cultivated land in gazetteers. Regarding the soldier-cultivated land that was put under the administration of other counties, it was not included in the county’s total cropland area calculation. The abandoned land area was also not included in the total cropland area because no crop was grown on the forsaken land, and it was exempt from taxes. After the step of registered cropland data collection, we collected 431 records related to tax-paying cropland area data from gazetteers in the Guanzhong area, covering 46 counties in the Qing Dynasty. The data sources in the Republic of China and after 1949 were statistical and survey data, which recorded the cropland area in tables. Therefore, the total cropland area of each county was collected directly.

2.3.2 Time points selection and data interpolation

Because not all counties have collected tax-paying cropland data each year in the Qing Dynasty, we selected several years when there were enough counties with tax-paying cropland data as time points for studying cropland change in the Guanzhong area. Figure 3 shows the number of counties that had tax-paying cropland data from 1644 to 1911. Specifically, 17 (37%), 41 (89%), 10 (22%), 18 (39%), and 46 (100%) of the 46 counties had tax-paying cropland data in 1650, 1734, 1741, 1779, and 1827, respectively. In other years, the number of counties that had tax-paying cropland data was below 10. Thus, we selected 1650, 1730, 1780, and 1830 as four time points and chose the tax-paying cropland data of the neighboring 15 years to supplement the data in 1650, 1730, and 1780. Augmented by the data in the neighboring years, 36, 42, and 37 counties with tax-paying cropland data were available in 1650, 1730, and 1780, respectively. The missing data were interpolated using the data in neighboring time points and correlations of tax-paying data in each time point.

At the beginning of the reign of Guangxu (1877-1878), a severe drought (Dingwu Disaster) hit the Yellow River basin in China, which lasted for 3-4 years and caused a considerable reduction in the cropland area. Shaanxi Province was one of the disaster-stricken centers. Many gazetteers recorded the effect of the severe drought on the cropland area in the Guanzhong area. Therefore, to quantitatively analyze the effect of the severe drought in the Guanzhong area on cropland area, it is necessary to select a time point at approximately 1877-1878. Around 1880, 15 out of 46 counties (33%) had tax-paying cropland data. Gazetteers recorded that the proportion of abandoned cropland in northwestern Guanzhong area was approximately 6%-20%, whereas it was approximately 0.41%-0.94% southeastern and 0.14%-3.6% in the central plain area from 1830 to 1880. Based on the research of Zhang (2018), the abandoned cropland recorded in gazetteers of the reign of Guangxu may include the cropland that was abandoned in the early Qing Dynasty. Moreover, Shaanxi Province was not seriously affected by the drought (Xia, 1992). We assumed that neighboring counties had the same proportion of abandoned cropland caused by the disaster. Then, we interpolated the missing tax-paying cropland data in 1880 based on the ratio of abandoned cropland areas in their neighboring counties.

Based on the collected cropland area data in each county in the Republic of China and after 1949, we selected 1910, 1930, 1980 and 2016 as time points. No missing data were interpolated from 1910 to 2016.

2.3.3 Data calibration

The conventional Chinese land-assessment principles are very complicated. The registered tax-paying cropland area data in the Ming and Qing dynasties were very different from the real cropland area (Ho, 1959). The area unit of mu used in tax-paying cropland registration is not a land area unit in modern times, but a land tax unit. The tax-paying cropland area data recorded in gazetteers are also affected by land assessment principles and land tax units. Since the tax-paying cropland area data recorded in gazetteers were obtained based on the real cropland area, they will be closer to the real cropland area after analysis and calibration. In previous studies, main factors such as various definitions of mu, discount of mu, land tax quotas, unregistered land, and exaggerating phenomenon were considered to calibrate the collected tax-paying cropland data from gazetteer (Wei et al., 2015; Ye et al., 2009). These factors led to the difference between the registered tax-paying cropland area data and the real cropland area. We calibrated our collected tax-paying cropland data based on the characteristics of the registered cropland area and the trend of cropland area change in the Guanzhong area (He et al., 2003).

There was no description of the unit of cropland area in gazetteers in the Guanzhong area. Qǐng, mǔ, and fēn were three units for tax-paying cropland recorded in gazetteers, with 1 qǐng = 10 mǔ = 100 fēn. However, the definition of mǔ in gazetteers (Qing mǔ) is different from that in the present day (Modern mǔ). The relationship between them is 1 Qing mǔ ≈ 0.9216 modern mǔ = 0.9216/1500 km². As there were no different definitions about the Qing mǔ in different counties, we converted the unit of tax-paying cropland area from Qing mǔ to km². The gazetteers in the Guanzhong area recorded discounts of mu for civilian land in 12 counties. We found ways of converting the true mu to tax-paying mu in three counties, i.e., Hancheng, Xianyang, and Yijun. After analysis and calculations, the discount rates of mu for civilian land in Hancheng, Xianyang, and Yijun counties were determined as 1.11, 1.01, and 2.76, respectively. We used them to convert the collected tax-paying mu to true mu. For the nine counties without any record about the ways of mu converting, we calibrated the tax-paying mu based on the conversion rate in their neighboring counties, statistics, and survey data in the Republic of China in corresponding counties.

In the early Qing Dynasty, affected by the Manchu war, approximately half of its cropland was abandoned (Peng, 1990). Records of the gazetteers in the Guanzhong area showed that the rate of cropland abandonment in the early Qing Dynasty reached approximately 30%. Since there was no policy to encourage reclamation before 1650, phenomena of concealment and exaggeration in tax-paying cropland registration were not apparent. Land tax quota delayed the growth of registered tax-paying cropland areas in the Qing Dynasty, especially in 1730-1830. However, the study of He et al. (2003) showed that in the period of agricultural restoration, which was from the tenth year of the Shunzhi reign (1653) to the sixth year of the Qianlong reign (1741), all the abandoned cropland in the early Qing Dynasty had been reclaimed. After that, from the sixth year of the Qianlong reign (1741) to the reigns of Daoguang and Xianfeng (1820-1861), the cropland expanded in high-altitude areas. The total cropland area around 1820-1861 should be larger than the land tax quota in the late Ming Dynasty. Thus, we used calibration coefficients of 1.1 and 1.2 to compensate for the unregistered cropland affected by the land tax quota in 1730 and 1780, respectively. We then used calibration coefficients of 1.2 to calibrate the registered cropland area in 18 counties located in the Guanzhong Basin and 1.5 to calibrate the cropland area in 28 counties in high-altitude areas in 1830. After calibration, the total cropland area in 1730, 1780, and 1830 was 0.86, 0.94, and 1.02 times the land tax quota in the late Ming Dynasty. Statistics in 1910 were also affected by the factor of unregistered land. The correlation analysis between statistics and the survey data of each county in 1980 in the Guanzhong area (CISNR, 1989) showed that the survey cropland area data was 1.18 times of the cropland area data from statistics. Thus, a calibration coefficient of 1.18 was used to improve the cropland area data from statistics to the survey data standard in 1910. After calibration, the cropland areas of seven counties were found to be much smaller than those in 1880. Based on the cropland area change from 1880 to 1910 in their neighboring counties, the cropland areas of seven counties in 1910 were re-calculated. For statistics in 2016, the calibration coefficients were calculated based on the statistics and survey data at the municipality level. The statistics at the county level in the corresponding municipalities were also improved.

2.3.4 Cropland area allocation

Based on the studies of administrative division change in the Guanzhong area over the past 300 years (Fu and Zheng, 2007; Fu et al., 2013), we found that administrative boundaries of most counties did not change significantly before 1949, except for splitting, merging and renaming of a few counties. The administrative map in 1911 was used as the base map for cropland area data in 1650, 1730, 1780, 1830, 1880, 1910, and 1930. We adjusted the cropland area data for counties that were split or merged. The administrative map in 1980 and 2015 was used as the base map for cropland area data in 1980 and 2016, respectively.

To allocate the cropland area data at the provincial level in China into 10 km×10 km grids, Li et al. (2016) developed a cropland area allocation model. This model was also applicable to many regional studies (Paudel et al., 2017; Li et al., 2019; Wei et al., 2019). In the Guanzhong area, the cropland area reached the maximum around 1980, and most of the built-up lands were changed from cropland. Thus, the potential maximum cropland area data for cropland area allocation was developed from the 1 km×1 km satellite-based cropland and built-up data in 1980. Factors of elevation, slope, and the climatic potential productivity data at 1 km×1 km resolution were quantified and used to calculate the value of land suitability for the cultivation of grid i. Based on cropland area data at the county level in the Guanzhong area and the cropland area allocation model (Li et al., 2016), we allocated the cropland area at the county level in the Guanzhong area into 1 km×1 km grids for the time points of 1650, 1730, 1780, 1830, 1880, 1910, 1930, 1980, and 2016.

Affected by natural disasters, wars, and politics, the total cropland area in the Guanzhong area fluctuated over the past 360 years (Figure 4). The changes could be classified into four phases.

The first phase was from 1650 to 1830. The Guanzhong Basin, which is close to the Yellow River and has a flat terrain, was almost wholly cultivated in the Ming Dynasty (1368-1644). However, the Manchu wars in the late Ming Dynasty caused numerous deaths and escapes. As a result, about 30% of the cropland was abandoned. Around 1650, the total cropland area stood at 1.47×10⁴ km². After Manchu took over Beijing and the Qing government suppressed the anti-Qing forces in Shaanxi Province, land reclamation in the Guanzhong area began. The government implemented a series of policies to encourage land reclamation. The total cropland area increased steadily from 1650 to 1830, peaking at 2.09×10⁴ km² in 1830.

The second phase was from 1830 to 1930. After 1830, the steady growth of the cropland area did not last for a long time. In the second phase, the “Tongzhi rebellion” among the Hui people in Shaanxi Province occurred. After the Qing government suppressed the rebellion, the most severe drought event of the 19th century, Dingwu Disaster, occurred and had devastating effects on agriculture and water supplies in the Guanzhong area. The total cropland area decreased to 1.93×10⁴ km² in 1880, which was less than that in 1780. The next 30 years still experienced a slight decrease, with 1910 arriving at 1.86×10⁴ km². From 1924 to 1930, devastating droughts, hail, and frost disasters caused severe damage to cropland, and war caused by warlords aggravated the impact of the disasters (Yuan, 1994). The total cropland area plunged to the lowest point of 1.81×10⁴ km² in 1930.

The third phase was from 1930 to 1980. During this period the total cropland area increased rapidly, peaking at 2.32×10⁴ km² in 1980. Due to the reduction of wars and the government’s encouragement of land reclamation, the fractional cropland area increased from 33.16% in 1930 to 41.86% in 1980.

The fourth phase was from 1980 to 2016. The total cropland area declined again to 1.67×10⁴ km² in 2016 because of urbanization, migration, and industrial structure changes. Moreover, survey data showed that areas comprising forests and fruit trees increased, whereas cropland and grassland distributions decreased in this period. The planting of fruit trees and changing cropland to forests also contributed to the decline in the cropland area.

Most of the cropland was distributed in the Guanzhong Basin, especially on both sides of the Weihe River and on the Yellow River’s west side (Figure 5). In 1650, a large amount of cropland was abandoned. Grid cells with more than 50% of the cropland area accounted for only 26.52% of the total number of grid cells. The cropland area fractions of most grid cells in the northern Guanzhong area, which belongs to the Loess Plateau, and northern Qinling Mountains were less than 30%. The years of 1730 and 1780 were in the period of land reclamation. The cropland area fraction increased in the Guanzhong Basin. However, the cropland area in the northern Guanzhong area and northern Qinling Mountains did not change significantly. Grid cells with more than 50% of the cropland area accounted for 33.35% in 1730 and 35.2% in 1780. The grid cells with more than 80% of the cropland area increased by 5%. As the population increased steadily, the abandoned cropland was completely reclaimed. Uncultivated land and mountainous areas with poor natural conditions also began to be cultivated by migrants. In 1830, the cropland expanded in the northern and western Guanzhong areas. Additionally, the cropland in the north of the Qinling Mountains increased slightly. Between 1730 and 1830, grid cells with more than 20% of the cropland area increased by 5.37%, mainly distributed in the northern and northwestern parts of the Guanzhong area. Grid cells with more than 50% and 80% of the cropland area accounted for 37.14% and 23.97% of the total number of grid cells in 1830, respectively.

Then, wars and disasters caused a sharp decline in the population of the Guanzhong area, and a vast amount of cropland was abandoned again. The cropland area in the overall Guanzhong area declined, either in the basin or the high-altitude areas. The devastating Dingwu Disaster ended around 1880. Grid cells with more than 50% and 80% of the cropland area declined to 35.62% and 16.69%. From 1880 to 1910, the government adopted several policies to encourage reclamation. Thus, the cropland area increased significantly in and around the Guanzhong Basin, mostly in Chang’an, Fuping, Heyang, Lintong, Jingyang, and Chunhua counties. However, the cropland area still decreased in the western, northern, and northeastern parts. Grid cells with more than 50% of the cropland area also decreased to 34.50% in 1910. In the following years, endless wars and frequent natural disasters prevented agricultural production in the Guanzhong area from fully recovering. In 1930, though the cropland area in Chang’an, Xianning, Weinan, Dali, Fufeng, and Baishui counties increased, the cropland area in Pucheng, Lintong, Chunhua, Fuping, and Baoji counties decreased a lot, which was impacted by one of the worst droughts in 1929 in northwestern China.

The end of wars, promotion of agricultural technology, and implementation of new policies caused an increase in the cropland area, which expanded dramatically in the northern, northeastern, and western parts of the Guanzhong area. Grid cells with more than 50% and 80% of the cropland area increased to 43.26% and 21.71% in 1980. However, urbanizations caused a decline in the cropland area in the central Guanzhong Basin.

After 1980, the cropland declined in the whole study area, especially in the cities of Xi’an and Xianyang. Because of the acceleration of urbanization, demands for urban land increased. Simultaneously, since cities could provide more employment opportunities and higher returns, many rural laborers entered the towns, making part of the cropland unusable. To increase income, people stopped growing food crops and replaced them with fruit trees and vegetables. All these factors caused the decline in the cropland area. In 2016, grid cells with more than 50% and 80% of the cropland area dropped to 27.91% and 3.59%, which were even less than those in 1730.

Our cropland data in 1650, 1730, 1780, and 1830 reflected the change of cropland cover in the early and middle Qing Dynasty. However, the decreases in the cropland area in 1880 and 1930 were caused by the temporary abandonment of cropland. The abandoned cropland was not cultivated for a short time or converted to other land cover types. Thus, our cropland area data in 1880 and 1930 reflected the impacts of wars and disasters on cropland, but not cropland cover for a long time.

From 1862 to 1877, the “Tongzhi rebellion” by the Hui occurred because of racial antagonism and class warfare (Hastings et al., 1916). The revolt began in 1862 in the Weihe River valley and spread rapidly throughout southeastern Shaanxi (Twitchett and Fairbank, 1980). Qwing to war conditions, many civilians died, and many people migrated to Shanxi and Inner Mongolia. The post-war consensus reflected a population reduction of 44.7% in Shaanxi (Cao, 2001). After Dungan Revolt, almost all cropland reclaimed by immigrants in the early Qing Dynasty in the northwestern part was lost (Hou, 1997). Then, a large-scale drought, known as the Dingwu Disaster, occurred in China from 1877 to 1878, which also caused the death of countless people and migrations. One of its centers was Shaanxi Province. The disaster temporarily prevented the post-revolt reclamation. However, the drought in central Shaanxi was fully relieved in April 1878 (Zhang and Liang, 2010). With the gradual return of refugees and policies that encouraged reclamation, the abandoned cropland was cultivated again (Hou, 1997).

Starting in 1927, wars and drought disasters across northern China occurred, causing severe famine and many deaths in rural areas. Among the disaster-affected provinces, Shaanxi Province was the most affected, with 3 million deaths by hunger, and 6 million people fled (Li, 2007). The Guanzhong area was the hardest-hit area in Shaanxi. In 1930, the population of all counties in Guanzhong decreased significantly compared with that in 1928 (Zheng, 2001). The amount of abandoned cropland area accounted for about 30%-100% in 19 counties in Shaanxi Province (Deng, 2012). Grain prices were abnormally high, while land prices plummeted; however, no farmer wanted to buy land (Zhao and Hou, 2012). With the appearance of rain in some counties in 1930, the drought ended by 1932 in the Guanzhong area (Zhang, 2014). However, affected by the drought, more than 100 km² of abandoned cropland in fertile Weihe River valley was still uncultivated in 1933 (Deng, 2012). Therefore, although the cropland was also abandoned temporarily in 1930, the barren cropland lasted longer than in 1880.

Three published studies provided the total cropland area data in the Guanzhong area over the past 300 years (Liu, 2015; Li et al., 2016; Klein Goldewijk et al., 2017); we compared them with our results (Figure 6). Liu (2015) collected the total tax-paying cropland area data from historical literatures without calibration. Thus, the total cropland area of Liu (2015) is much lower than ours, especially after 1730. Moreover, since different data sources were used by Liu (2015), the total cropland area from 1724 to 1741 varied greatly. The total cropland area data at the provincial level was allocated into grid cells of 10 km×10 km, depending on the land suitability for cultivation, and the Chinese historical cropland dataset (CHCD) was generated (Li et al., 2016). Because the Guanzhong area is the most suitable area for cultivation in Shaanxi Province, the cropland allocation method of the CHCD increased the total cropland area more than in this study.

The cropland area changing trend of the Guanzhong area from the HYDE 3.2 dataset is close to that in this study, especially from 1730 to 1950 (Figure 6). However, HYDE 3.2 shows that the cropland area grew from 1930 to 1950, decreased from 1950 to 1980, and grew rapidly after 1980. This study reflects that the period between 1930 and 1980 experienced a sharp rise, and the cropland area quickly declined after 1980. Based on the analysis of Feng et al. (2005), China experienced a new land reclamation period from 1963 to 1966. The cropland area in the whole country increased until 1979. However, HYDE 3.2 significantly underestimates the cropland area in the Guanzhong area in 1980. The cropland area extracted from HYDE 3.2 was 1.74×10⁴ km² in 1980, which is less than that in this study. Conversely, in 2016, the total cropland area from HYDE 3.2 was 2.06×10⁴ km², which is more extensive than that in this study. Different definitions of cropland between HYDE 3.2 and this study may be among the reasons for the difference in 2016. The cropland in HYDE 3.2 included permanent crops, which were excluded in this study.

Since the total cropland area from HYDE 3.2 has good correspondence with our study, we compared our cropland maps with HYDE 3.2. We used Formula (1) to calculate the differences between HYDE 3.2 and this study for six time points (1830, 1880, 1910, 1930, 1980, and 2016). For HYDE 3.2, we selected 1820 in correspondence to 1830 in this study.

where Difference(i) is the difference between HYDE 3.2 and the data of this study, HYDE(i) is the cropland area fractions of grid i in HYDE 3.2, and GZ(i) is the cropland area fractions of grid i in this study.

Differences in cropland distribution between HYDE 3.2 and the data in this study are depicted in Figure 7. For the six selected time points, the differences of grid cells between HYDE 3.2 and this study were all between -79% and 40%, and the grid cells with a negative difference value accounted for approximately 34%-55% of the total grid cells. From 1820 to 1980, grid cells of HYDE 3.2 had lower amounts of cropland area fraction than those in this study in the Guanzhong Basin. However, in high-altitude areas around the Guanzhong Basin, HYDE 3.2 had higher amounts of cropland area fraction than this study, especially in the northern and northwestern parts of the Guanzhong area. Before 1930, HYDE 3.2 had a larger cropland area along Weihe River’s tributaries, Jinghe River and Qianhe River. Besides, there was little difference in the cropland area fraction of grid cells between different regions inside the Guanzhong Basin showed by HYDE 3.2. However, this study shows grid cells with high cropland area fractions concentrated in the middle of the basin, especially around Xi’an City. HYDE 3.2 showed the same spatial patterns of cropland in 1930 and 1980. However, this study showed that the number of grid cells with high cropland area fraction decreased in 1930 because of the drought, and cropland expanded to high-altitude areas around the Guanzhong Basin in 1980. The number of grid cells with differences of less than -20% and more than 20% accounted for 28% in 1980 and 23% in 2016. Since HYDE 3.2 and this study used different data sources and cropland area allocation models, there were significant differences between them. The difference in spatial resolution and accuracy of the data sources also increased the contrast. Assigning more weight to the grid cells around the river in the cropland area allocation model and allocating the total cropland area at the provincial level to grid cells increased the uncertainty of HYDE 3.2 before 1930 in the Guanzhong area.

There were still some uncertainties in this study. Since the cropland area information of only 15 counties was recorded in gazetteers, we interpolated the missing data in 1880 based on cropland records in neighboring counties, which may cause data errors.

Furthermore, we only found the administrative division map at the county level for 1911. However, there are few results on the changes in the administrative boundaries of each county in history. Thus, the collected cropland areas from data sources in some counties did not match the administrative division map of 1911. For example, the recorded cropland area in Fengxiang County before 1910 was larger than the land area provided by the administrative division map of 1911. We could only adjust the cropland area based on the cropland area fraction in its neighboring counties. Therefore, there is a need to obtain more information on the administrative changes of each county in history.

For the assessment of the accuracy of the cropland area allocation model, Li et al. (2016) compared their results, which were obtained using the same model, with satellite-based data obtained in 2000 for the whole country. The differences were mostly distributed between -20% and 20%, which reflected the reliability of the cropland area allocation model. Although the result of cropland area allocation in this study is reliable, the built-up land was not removed from the potential maximum cropland area map before 1980. A few cropland areas were allocated to grid cells that were supposed to be built-up land. Moreover, more information on historical facts relating to agricultural development in the Guanzhong area should be collected to improve the rationality of the reconstruction.