 |
|
|
Slide 29 
|
Slide 30 
|
|
Slide 31 
|
Slide 32 
|
|
Slide 33 
|
Slide 34 
|
|
Slide 35 
|
Slide 36 
|
|
Slide 37 
|
Slide 38 
|
|
Slide 39 
|
Slide 40 
|
|
Slide 41 
|
Slide 42 
|
|
Slide 43 
|
The source notes are of particular importance and cannot be found in any printed
historical atlas. They consist of extracts from the historical sources used to
determine administrative changes, point locations, and boundaries. Historical
texts, with bibliographic citations, are quoted directly from primary and secondary
sources in Chinese and are accompanied by commentaries in which the editors
provide written justifications for their choices. In this way the core research being
done to create the database is exposed to the users for their reference, making
use of the power of the relational database in way that clearly differentiates the
CHGIS datasets from a printed atlas, or digitized versions of a printed atlas.
For reasons discussed below, for the last millennium it is possible to add
settlements (usually market towns) outside of the hierarchy of administrative
seats to the database. CHGIS Version 1 (released Apr 2002) included 7,700
towns for the year 1820 CE. With CHGIS Version 2 (released Oct 2003) this
number increased to 8,600 towns, while an additional 17,400 towns for the year
1911 CE were added, along with 2,000 localities extracted from Russian
Historical maps. By Version 3 (released Apr 2005) the number of 1911 towns
reached 27,400, and we hope to finish the project with approximately 40,000
towns and villages of historic significance in the database. [slide 44]
Slide 44
Using the CHGIS datasets
Although when completed CHGIS will supersede any existing historical
administrative geography of China in terms of temporal range, spatial accuracy,
and documentation of sources, it is more than a mere improvement on the
traditional printed atlas. I shall illustrate this in two ways, by examining what we
might learn from the datasets themselves and by showing what can be learned
by joining other kinds of data to the CHGIS base datasets.
Looking at administrative change
A continuous time series reveals far more information about the administrative
system in the aggregate through space and time than any other usable resource.
A year-by-year administrative atlas allow us to see changes in the spatial
distribution of the administrative system over time, without, however, accounting
for those changes. It thus has the potential for generating interesting historical
problems.
My first example has to do with changes in administrative density. In maintaining
a field administration the imperial state was torn between two goals. On one
hand it wanted to fully incorporate territory into the civilized realm, which required
an expensive investment in civil institutions and military security. On the other
hand it wanted to maximize its revenues, which required that its investment be
reduced to the minimum necessary to secure a surplus that could be transported
to the center.6 The basic units of the field administration (i.e. administrative units
managed by officials sent from the center), the counties and prefectures, were
expected to pursue both goals but at the very least they were expected to pay for
themselves, even if their revenue value to the center was minimal. Thus when we
see changes in the density of the field administration we know we are seeing a
symptom of something. In this first case we are looking at eastern Guangdong
(the Canton area) in the 8th, 11th, and 16th century [slides 45-46] and concluding
that it is much more part of the early 8th century Tang state than the 11th century
Song, although it becomes important again in Ming.
Slide 45
Slide 46
This only defines a problem; it does not explain very much. At first glance the explanation has to do with
Canton's status as the major port for the Southeast and South Asia sea trade,
where goods were offloaded and transferred to internal rivers for transport to the
north. In the 11th century that role was fulfilled by Fujian ports. The shift in ports
can be attributed to the Huang Chao rebellion, which sacked Canton in 879 and
massacred as many as 120,000 of the 200,000 inhabitants, the bulk of whom
were Southeast Asian, Indian, Persian, and Arab merchants.7 However, when we
look at the data more closely we find that the decline in administrative density
had begun prior to the rebellion. In fact a 13 percent decrease took place
immediately after the An Lushan rebellion of 755, which sacked the capital of
Chang'an. This suggests an alternative account: that an embattled court saw
local government in Guangdong as an unnecessary cost. The successful attack
on Canton in 879 followed a declining government presence. In any case
although we can see the renewal of local government in Guangdong in the 15th
century as a revival, we can also understand why its Ming inhabitants traced their
history back to migration south during the collapse of the Song in the 13th century
and had little sense of social continuity with Tang.8
Adding Data to CHGIS Datasets
The great promise of CHGIS lies in using it to analyze relationships in other kinds
of historical data that have spatial attributes, that that can be joined to CHGIS
locations.
In the first case [slide 47] I have taken data on the quota assigned stations for
collecting commercial tax (the transit tax of 3% of value and sales tax of 3%) in
1077 and distinguished between tax quotas for prefectural seats, county seats,
and market towns in the circuit of Liangzhe (modern Zhejiang, southern Jiangsu,
and eastern Anhui provinces). The measure is strings of 1000 copper cash
(guan). Black symbols are used for market towns, yellow for county seats, and
red for prefectural seats. Symbols of the same size represent equal amounts.
We see towns with tax quotas greater than their county seats and county seats
with quotas greater than their prefectural seats, and there are administrative
centers with no quotas at all, indicating an absence of significant commercial
activity.9 The conclusion to be reached from this is important: by the 11th century
the hierarchical network of economic central places had separated from the
hierarchical network of administrative seats. This was the first great commercial
revolution in China's history and the moment when government learned how to
tax the commercial economy instead of trying to control and command it as it had
in early imperial eras.
Slide 47
My second case draws on a prosopographical database of over 20,000 officials
from the 10-13th century, [slides 48-50] including 8,000 holders of the highest civil
service examination degree.10
Slide 48
Slide 49
Slide 50
By joining these to the CHGIS datasets we can
accurately depict where these officials came from and compare their spatial
distribution over time [slides 51-2].
Slide 51
Slide 52
One preliminary result has not, as far as I
know, been noticed: over time the geographic representativeness of the higher
echelons of the bureaucracy (i.e. those people whose biographical details were
most likely to be preserved) narrowed to the point that three circuits dominated
the government (Liangzhe, Fujian, and Jiangxi) just at the point that the Mongols
were gathering power (they took north China from the Jurchens in 1234 and
successfully invaded Song in the south in 1270) and the Song government was
trying (ultimately without success) to increase its access to the fiscal resources it
needed for national defense.
In all these cases I have been employing the most basic GIS techniques:
querying the database by year, joining data, and arriving at a preliminary analysis
by looking for patterns in the mapping of the data in time and space. Even at this
unsophisticated level one of the contributions of GIS is apparent: it allows us to
keep track of large quantities of data and to see them simultaneously in
relationship to each other. Statistical analysis could have done much of this - and
indeed the next step would be to quantify some of these findings - but some of it
would have required deriving information from a map. More fundamentally, I think
GIS provides an inducement for historians to give more attention to thinking
spatially: to see space, like time, as a "conceptual and analytic framework within
which data can be integrated, related, and structured into a whole"; to store
structured data in forms that others can share and analyze; and interpret and
explain spatially structured information.11
Spatial Ontology in Historical GIS
However, if we are to use GIS in history we need to ask how our
conceptualization and representation of space in the present relates to past ways.
This is a problem of spatial ontology. I shall treat one aspect of this, the "points
versus polygons" question, which is fundamental to the way CHGIS represents
the administrative hierarchy. CHGIS provides "points" for up to seven levels of
administrative seats; it provides polygons for the counties in 1911 based on
mathematical cartography but also for the prefectural level units throughout
history. For the first millennium of imperial history the roughly 300 prefectures
were the most important units in the field administration as they mediated
between the capital and the roughly 1,200 counties.12
CHGIS depicts prefectural units as territorial units with clearly marked boundaries
[slide 53].
Slide 53
In my view this is anachronistic and misleading for too much of
Chinese history. In China, from early times, the administrative seat defined the
administrative unit, rather than a clearly bounded territorial unit. In the United
States, administrative units are likely to have names different from the town in
which the administrative seat of government is located. But this has not been the
typical assumption in China's history. To illustrate: if the state of Nebraska were
in China its administrative seat would be called Nebraska, we would be meeting
in Nebraska rather than Lincoln and, as we see on all early Chinese maps and
read in early historical sources the state would be depicted as a point location on
the map not as a polygon. In China the national capital was the seat of the court,
the center of power and wealth, and for much of history a planned city. It was the
"pivot of the four quarters," standing against the uncultivated wilds; just as the
light of the "son of heaven" in the central plain was in the midst of the forces and
deities in the darkness of the surrounding mountains.13 Administrative capitals
replicated this notion of centrality. The depiction of administrative units, which
obviously had jurisdiction over surrounding villages, is evident in these 11th-12th c.
maps [slides 54-55] where the prefectures are named points rather than
territories.14
Slide 54
Slide 55
This was not because a national map of limited size impeded the
depiction of boundaries. Surviving county and prefectural maps from the
thirteenth century represent prefectures and counties as administrative seats. An
example from 1261 [slide 56] depicts physical features (mountains, rivers, lakes)
graphically and the prefectural and county seats planimetrically, as square,
enclosed cities. The subcounty administrative units (the "township" or "canton"
xiang, territorial units without fixed seats) are merely labeled.
Slide 56
The difference between this traditional style of square map and a much later
addition to the repertoire, the map that depicts an administrative unit as isolated
bounded territory, is evident in these examples, which come from the 1892
edition of a local gazetteer. Here [slide 57] is a traditional page-filling square of a
county: it shows the seat and physical features and labels of subcounty cantons.
It characteristically magnifies the county seat and locates the county relative to
the surrounding counties by cartouches along the perimeter at the eight compass
points (N, NE, E, etc.), reading "north to Yiwu county," and so on. The other map
[slide 58], from the same source, shows roads connecting villages to the county
seat, but for the first time in the case of this county represents the administrative
unit as bounded space.15
Slide 57
Slide 58
Although there were precedents, depicting the county as an isolated bounded
territory becomes popular in the nineteenth century and is probably due to the
influence of European cartography.16 The traditional sense of a hierarchy of
administrative seats has continued in China. Even today, a villager likely sees the
village as subordinate to the county, by which the county seat is meant, rather
than as being "in" the county.17 The administrative hierarchy to which a given
settlement belonged mattered because the government saw itself as governing
through a hierarchy of administrative centers. And this is how traditional
geographies such as Li Jifu's Maps and Treatises on Commanderies and
Counties in the Yuanhe Reign Period from 806-814 and Wang Cun's Treatise on
the Nine Regions in the Yuanfeng Reign Period from 1078-85 conceptualized
the administrative hierarchy.18 They treat administrative units as points in space,
giving the direction and distance from the prefectural seat to surrounding
prefectural seats and from the county seat to the prefectural seat [slides 59-60].
The appropriate way to represent this would be as a hierarchical system of
networked points [slides 61-62].
Slide 59
Slide 60
Slide 61
Slide 62
Chinese historians and administrators could conceive of boundaries and of space
as bounded territory. Beginning in the eleventh and twelfth centuries greater
attention came to be paid to the territorial extent of prefectures in both
geographies and maps.1919 Nevertheless, although surveying techniques for
measuring direct distance were known,20 extant maps and geographies did not
yet attempt to depict or describe administrative units as bounded space. The
focus on the administrative seats where officials sent from the center resided, the
limited resources of government, and above all the nature of the tax system,
which registered population first and land second, help explain the lack of interest
in delineating administrative boundaries. Land parcels were measured and their
productivity assessed in order to assign the tax burden of each parcels.
However, there was variation across regions: in the north China plain into the
eighth century, land was tracked for the purpose of the equitable distribution of
lands to taxpayers, whereas this appears not to have been done in the south.21
And there was temporal change: as land came to be redistributed through the
private market in all regions in the late 8th century, the tax system graded
households by wealth and collected land tax from the owner of each parcel.
However, correlating the land registers with the population registers depended on
recording all land transactions or regular cadastral surveys, neither of which
happened. Land registers could have provided a basis for defining administrative
units as bounded territory, but only if all land was farmed in some fashion and
made tax-worthy. Thus although the idea of creating maps from the bottom up to
serve as an integrated guide to land and population did occur to some there is no
evidence that such a system was put into effect.22 Given the private market in
land, over time the land held by households in a given registration group would
change so that to depict their registration unit as bounded space would make it
even more difficult to collect taxes. This suggests that unless there were natural
barriers such as rivers or mountain ranges which created clear lines of
jurisdiction, people at the time thought of the territorial boundaries that divided
administrative units as flexible zones. From an administrator's perspective the
real question was whom he had to collect taxes from. He had to know which
settlements fell under his jurisdiction and what property was held by the
inhabitants of those settlements; he did not need absolute boundaries.
Most prefectural and county gazetteers, which began to appear in large numbers
in the twelfth century, remained oriented toward the administrative seat, locating
mountains, bridges, reservoirs, markets, and religious sites by direction and
distance this one central point [slide 63].
Slide 63
Although some argued that the county
should be treated as a physiographic spatial unit rather than as a series of points
networked to the administrative seat, this did not become the dominant view until
the nineteenth century under the influence of European surveying techniques
and cartography.23 Although contemporary cartography follows the Western style
of depicting units as bounded space, the government is only now systematically
using modern surveying techniques to define county boundaries exactly.
This suggests to me that by depicting the administrative units as bounded
territory we are misconceiving the relationship of the imperial field administration
to the land and population. For the first 1000 years of imperial history, and well
into the last 1000 years for certain regions, we might be better served by thinking
of the administrative hierarchy of the Chinese empire as a colonial administration,
sent out into the countryside in search of revenue and plunked down in the midst
of the natives, to whose local elite it offers the opportunity of participation in the
imperial project. From this perspective the administrative seats are outposts,
strategically located at transportation hubs (where rivers meet) that divide up the
arable (and thus taxable) land between themselves. In fact the location of county
seats in the south follows this model fairly well [slide 64].
Slide 64
What does change—and I think this points in the direction of a new spatial
consciousness which starts to see administrative units as bounded territory as
well as networks of central places—is the emergence in the twelfth century of
elite communities with a high degree of local identity through much of the south.
As they begin to play a larger role in the life of the locality and learn how to
influence local officials, the colonial character of local government dissipates.
One sign of the de facto joint government of local elites/local officials is the
spread in the twelfth century of the "local gazetteer," a history and survey of a
county or prefecture that the local government and local elites compiled together.
Local gazetteers have extensive information on local elites, as well local
government, religious sites, markets, population, and physical geography and
infrastructure. The local gazetteers are practically all data—the challenge is to
extract the data in formats that allow for large-scale spatial analysis.
However, despite these objections, the historical geographers working on the
CHGIS project have been spending a great deal of time creating prefectural
boundaries or polygon layers, despite the fact that for the first 1000 years of the
database these boundaries simply do not have the same meaning that they do
for the second millennium. By doing this we have certainly created false
certainty—the boundaries simply cannot be identified with such exactitude—and I
fear we have given a fundamentally misleading impression of the process by
which the field administration and the empire operated. The justification for
creating boundaries, however, is that they have heuristic value: users expect to
be able to do density analysis and even if the boundaries are highly inexact they
serve this purpose.
Approaches to Density Analysis in an Historical GIS
A GIS that extends its temporal scope beyond coverage of accurate historical
maps will inevitably face the fact that we lack reliable information on boundaries
even if they ever were well defined (settlement points are much easier to
establish). What, then, should we do?. Here GIS technology can be of great help.
I shall illustrate this with two applications of GIS software. The first generates
heuristic boundaries for density analysis based on point data, boundaries which
can be refined as more point data is added to the database. The second creates
density analysis that does not normalize by the area of the administrative unit but
takes into account the uneven distribution of population and activity across a
landscape of mountains, plains, valleys and lakes.
Creating Heuristic Boundaries with Thiessen Polygons
Let us suppose that we have no information on boundaries at all, but that we
know the point locations of administrative seats. The simplest way of
approximating boundaries is to assume that all boundaries are equidistant
between two points. The resulting polygons are known as Thiessen polygons.
The results are very rough indeed, as we see in this case of counties in a
prefecture [slide 65]. However, we also know the point locations of a few towns
within each county. When we generate Thiessen polygons using this data the
boundaries become more complex.
Slide 65
Here we apply this approach to the province of Zhejiang. Thiessen polygons are
generated for the entire province, based on the locations of townships, and color
coded by county [slide 66]. We then compare this with the actual county
boundaries as of 1990 [slide 67]. We see that well over 90% of the county
territory is accurately defined with Thiessen polygons. When we aggregate up to
the prefectural level the accuracy is even higher [slides 68-69].
Slide 66
Slide 67
Slide 68
Slide 69
Thus rather than spending time producing boundaries our resources could better
be devoted to adding market and settlement points. Chinese local gazetteers
identify the location of markets. Since the gazetteers have been revised and
updated every century or so we can show that markets persisted once
established, even if we can rarely be sure of the date of establishment. Here I
have located markets from the fifteenth through the nineteenth century on a
gazetteer map [slide 70].
Slide 70
Density Analysis in Real Landscapes
Simple density analysis normalizes data by area. In this example I have taken
the registered household population for the Zhejiang area as of 1102, given in
the sources as a prefectural aggregates, and represented density on a
chloropleth map based on CHGIS prefectural polygons. [slide 71] As expected
the population is most dense around the Hangzhou Bay and north to the Yangzi
River. However, taking the circled prefecture as an example, we can see from
remote sensing imagery [slide 72] that the population was not spread uniformly
across the landscape, which means that the kinds of human activity that
historians tend to study were in fact concentrated in about half of the prefectural
area.
Slide 71
Slide 72
We have registered population data for China going back to AD 2 (60 million).
Much of this is only at the prefectural level. We would like to distribute this
population across the landscape as background to other data layer on land use,
taxation, warfare, education, etc. Given that the level of aggregation is the
prefecture, we want to do this automatically but with a high degree of reliability.
Guoping Huang of the Center for Geographic Analysis at Harvard and I have
developed a model for historical population distribution. The model makes certain
testable assumptions: that people will congregate on arable land with less slope,
that they need access to water for agriculture and cheap transportation, and that
larger settlements will emerge at the conjunctions of rivers. We began with a
"digital elevation model" of Zhejiang province [slides 73-74].
Slide 73
Slide 74
Using a DEM, seen
here as "hillshade," GIS software can measure slopes and determine where
rivers ought to be, moreover it can define the relationships between these rivers
(the "Strahler order" of the degree of tributary status) [slide 75]. The accuracy of
the result was tested against two sources: a 1:250K paper map from ca. 1950
(US Army Map Service) and a digital database of streams from 2000; the
simulated hydrography is more accurate than the digital database. The next step
was to identify land that could support intensive farming. Here we took a general
expectation that farm land would have a slope of less than 8°. We tested this
against the land use data in the 1950 maps and were satisfied with the accuracy.
[slide 76]. This gave us the basic information [slide 77].
Slide 75
Slide 76
Slide 77
We then assumed that population would be concentrated in the arable land thus
defined, that it would tend to congregate around river intersections and around
prefectural and county seats. [slide 78]. Based on this model we suppose that the
density of the registered population in 1102 for this prefecture looked something
like this. To test the predictive power of this model we tested it against population
data by township from 2000. The result [slides 79-82], supports the conclusion
that our model can be relied upon to give us a population density with a high
degree of accuracy.
Slide 79
Slide 80
Slide 81
Slide 82
And since most human activity follows the population, we
have now have a map which shows where most of China's history took place.
We are currently extending the application to the province of Zhejiang, but the
ultimate intent is to do this for the core areas of China so that all available
demographic data can be displayed [slide 83].
Slide 83
Some Suggestions
The creation of large-scale historical GIS promises to provide historians with a
framework for organizing and analyzing the very large amount of historical data
that has spatial attributes. Short of a major infusion of financial support creating
these historical GIS will be a long term project. I suggest we think of them as the
equivalent of the creation of databases of historical texts. And once they are
done they will be available for all to use. Another possibility is under development
(although it requires rich gazetteers with sufficient temporal information. This is to
spatially enable statistical datasets so that they can be filtered through the
gazetteer and represented on a map. In either case it seems to me that we need
to think beyond "national" historical GIS and think in regional if not world terms.
At the very least we need to think of the larger contexts in which historical polities
existed rather than reading nation state of modern history, and its concerns with
national sovereignty and identity, back onto the past.
Main |
About |
Public Lectures |
Project reviews
Tool reviews |
Essays |
Interviews |
Bibliography |
News |
Blog
Department of History, University of Nebraska-Lincoln
This work is licensed under a
Creative Commons
Attribution-Noncommercial-Share Alike 3.0 United States License.
