Reconstruction of the cropland cover changes in eastern China between the 10th century and 13th century using historical documents

To evaluate and improve datasets of anthropogenic land cover change used in local and global climate models, great efforts were made to reconstruct historical land use, including the LandCover 6k project which dedicated to reconstructing human land use over the past 10,000 years. In this study, we utilized historical records, including taxed-cropland and cropland measurement areas, and data on the number of households in eastern China between the 10th century and 13th century in concert with coefficient calibration, model allocation, and per capita cropland estimation to reconstruct areas of provincial cropland for 22 provinces over five time periods. Our reconstructions indicate that total cropland areas of eastern China for AD 1000, 1066, 1078, 1162, and 1215 are 34.74 × 106 ha, 49.42 × 106 ha, 51.62 × 106 ha, 35.21 × 106 ha, and 51.21 × 106 ha, respectively. And the cropland area fluctuated because of dynasty shift and went through three phases. Cropland expansion and contraction mainly occurred in the middle and lower reaches of the Yangtze and Yellow Rivers as well as the Huaihe River Basin, while in some regions far away from battlefields, including northeastern and southern China, cropland area expanded continuously throughout the study period.


Collection of cropland area data
These data include taxed-cropland area and cropland measurement data. It is noteworthy that while the taxed-cropland area recorded in historical documents do not correspond to actual areas [1,2] but an indicator of agricultural taxes. However, as these values nevertheless bear close relationship with real cropland areas, they are the raw materials that form the basis for our estimates. In contrast, cropland measurement areas are real as these were recorded by officials appointed by the government to increase national taxes. These measurements, however, were only carried out in limited areas, and so data are sporadic. We therefore utilized several available measurements for cropland areas in this study in order to estimate the relationship between taxed-cropland and real cropland areas as accurately as possible.
National and provincial level taxed-cropland area data were mainly obtained from the Collections of Historical Governmental Archives (《文献通考》). This volume contains provincial taxed-cropland area data (19 Lu) for AD 1078 as well as national taxed-cropland area data for AD 976, AD 997, AD 1066, and AD 1078 and thus encompasses the period between the end of the10 th century and the early 13 th century.
County level taxed-cropland area data was derived mainly from Collections of Local

S1.2 Household and population data
The household and population data estimates used in this study were derived from Wu and Ge [3] . These workers calibrated and estimated provincial household and population data for a number of periods during the Song, Liao, and Jin dynasties (i.e.,

S2 The Song Dynasty
This regime comprised the Northern and Southern Song dynasties. Of these, the first was established in AD 960 with its capital at Kaifeng Fu (a 'Fu' was a city unit during the Song Dynasty; Kaifeng Fu is the present-day city of Kaifeng and surrounding areas). Subsequent to AD 1126, the war between the Northern Song and Jin dynasties resulted in occupation of the north of the Huaihe River by the latter. And the Northern Song Dynasty moved its capital to Lin'an Fu (the present-day city of Hangzhou in Zhejiang Province) and established the Southern Song Dynasty in AD 1127.

S2.1 Accuracy assessment of cropland area data
Following the establishment of the Song Dynasty, a system of taxation and corvee was implemented by the government to increase national revenue and to consolidate its political power. The number of people in each household as well as their incomes and registered properties were used as taxable units in ancient China, and so were  afforded considerable attention by the local government [4] . Of these, the cropland area owned by each family formed the main basis of taxation during the Song Dynasty and so abundant historical records exist from this period. However, as land policy switched to allow free exchanges and mergers during the Song Dynasty [5] , large areas of cropland were not registered to avoid agricultural taxes; thus, the recorded area of taxed-cropland at this time will be lower than actual area.

S3 The Liao Dynasty
The had finished by AD 1004 [6] . The region to the north of Yanshan Mountain then developed into a farming-pastoral zone that was self-sufficient throughout the period of the Liao Dynasty. The south of this region of China remained the dominant farming area throughout this period; its topographical environment and level of agricultural development were similar to Hedong and Hebei Lus during the period of the Northern Song Dynasty [7] .

S3.1 Taxation systems
The Liao Dynasty government enforced two different tax regimes in the areas to the north (i.e., Shangjing, Zhongjing, and Dongjing Daos) and south (i.e., Nanjing and Xijing Daos) of Yanshan Mountain. In the north, taxes and corvee were levied on the basis of the number of Ding (i.e., men aged between 15 years old and 50 years old) in each household, while the tax system of the Northern Song Dynasty was followed in the south, with taxes and corvee collected based on household Ding and associated properties. These differences meant that the government paid more attention to  recording Ding numbers than cropland areas; indeed, Ding inventories were implemented on numerous occasions during this period [8] and were recorded for the

S3.2 Population compositions
Wu and Ge [3] analyzed Ding numbers in addition to other historical population records in order to estimate provincial agricultural and total nonagricultural populations during the Liao Dynasty (Fig. S1). And according to the findings of existing research [9] as well as historical records from the Monograph on Camp Guarding from the History of the Liao Dynasty (《辽史•营卫志》) show that the non-agricultural population at this time was mainly distributed within Shangjing and Zhongjing Daos. We therefore estimated provincial cropland in this study using population data. Figure S1. Provincial agricultural and nonagricultural population data for the mid-to-late Liao Dynasty.

S3.3 Human demands for cropland
To the north of Yanshan Mountain, farmers cultivated crops to support their basic life styles. Indeed, the presence of both agricultural and nonagricultural populations in this region led to both cultivation and grazing and a difference in demand for cropland area between the two. In the case of the agricultural population, the cropland demand for each person can be estimated using their grain demand for one year (i.e., 237.6 kg  for adults and 118.8 kg for minors) [10] , crop yield per Song-mu (area unit of the Song Song-mu (about 0.35 ha) in Dongjing Dao. At the same time, the cropland demand of each person within the non-agricultural population was just one-tenth that of the agricultural population [7] .

S4 The Jin Dynasty
The Jin Dynasty was founded in AD 1115 by the Jurchen people, a minority who subsisted by fishing and hunting in northeastern China. The capital of this dynasty was located in Huining Fu within Shangjing Lu, the present-day city of Harbin in Heilongjiang Province. This dynasty successively occupied the territories of the Liao Dynasty as well as the northern area of the Northern Song Dynasty, expanding its area of influence as far as the Huaihe River in the north by AD 1127 (Fig. 1b).

S4.1 Population compositions
In order to estimate separately provincial area under Mengan Mouke and ordinary prefectural households, we estimated the number of households subject to two systems. We initially extracted data on the number of provincial households subject to Mengan Mouke across the whole of this region in AD 1183 as well as for Shangjing Lu in AD 1193. These data were then combined with the population growth rate between AD 1183 and AD 1207 (7.1‰) [11] , ratios of Mengan Mouke households in Shangjing, Dongjing, and Beijing Lus (6:3:1) [7] , and the frequency distribution of these households across northern China [12] (Fig. S2). Previous work has shown that the number of such households can be converted to an overall population estimate by applying a conversion coefficient of 10 [3] . These historical records were used to calculate the number of Mengan Mouke households in some Lus in northern and northeastern China. The number of ordinary prefectural households was calculated by using the total number of households and Mengan Mouke households.

S4.2 Cropland allocation method for Mengan Mouke households
Although just provincial household numbers are present in historical documents from the Jin Dynasty, records relating to historical land use policies and the system of  taxation in Mengan Mouke also yielded important information. In Mengan Mouke, for example, cropland was allocated by the government, while the so-called 'Cattle and Tool Tax' stipulated that this should be allocated according to population level [13] .

S4.3 Per capita cropland area for ordinary prefectural households
Land occupation by ordinary prefectural households at this time was based on private ownership [13] .Thus, the government collected agricultural as well as summer and autumn taxes (similar to the Song Dynasty) on the basis of the cropland area occupied by each ordinary household. This meant that local governments paid more attention to the registration and inventories of cropland area; during the Shizong (between AD 1161 and AD 1189) and Zhangzong Terms (between AD 1189 and AD 1208) periods, for example, the Jin Dynasty enacted a strict 'General Survey and Evaluation' system in order to survey property and expropriate tax. Between AD 1186 and AD 1187, the government performed this evaluation process three times [13] ,