A Geo-Sampling Model To Analyse Micro Level Historical Agricultural Production Data For Mid-19th Century Southeast European and Anatolian Regions

poster / demo / art installation
  1. 1. M. Erdem Kabadayi

    Koç Üniversitesi (Koc University)

  2. 2. Petrus Johannes Gerrits

    Koç Üniversitesi (Koc University)

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.

With this poster we would like to present and discuss our geo-sampling method devised to make selective and efficient use of a wealth of undigitized, underutilised micro level archival data from the 1840s for five regions (covering around 2,000 villages with a total of around 100,000 households and 500,000 people) in today’s Bulgaria, Northern Republic of Macedonia, and Turkey.
While the construction of land use suitability models is a common practice conducted in different fields, an attempt of bringing historical Ottoman archival data together with the spatial methods has not been done before, to the best of our knowledge.

In the last two years we have geo-located 1797 villages in five regions shown above. We are working as a research team on this task and by early 2019 we are confident that we will have completed our fifth region, centred around Bitola, in our project, and will reach a total number of around 2,000 villages in our dataset. We have already created weighted Voronoi polygons for all the villages in our five sets based upon their relative population sizes. Below you can see the example for our Plovdiv region.

For every village in the dataset we have already manually extracted total number of households and total population from a series of population registers from the 1840s and entered in to our geo-database.
For almost all of the villages in the set, we have also very detailed information on agricultural production, recorded per household in an Ottoman empire tax survey from 1845. The data on agricultural production from this 1845 survey include a detailed product mix (grains cultivated, fruits and vegetables grown), total area of cultivation, and total value of agricultural production. Moreover animals (pack, drought, and all other purposes) kept in households are also listed according to type, number, value, and generated income.
Evidently, we are facing a challenging and complex sampling problem to reach a representative sample of villages to be able to extract data and analyse agricultural production and animal breeding to understand dynamics of the primary and most important sector of economic activity in these five regions in the mid-19
th century. To tackle massive number of total villages and huge regional diversity we have divided our regions to four 4 sub-units. Now we are dealing with 20 units and we would like to conduct the same sampling strategy for each of the 20 sub-units.

Our geo-spatial sampling model has three major components:
1- Soil depth and soil quality, which can be downloaded for three regions centred around Bitola, Plovdiv, and Ruse from ESDAC (European Soil Data Centre:
https://esdac.jrc.ec.europa.eu/) for Bulgaria). We do also have soil data for two regions in Turkey.

2- Suitability for cultivation: using available SRTM data with 30 meter spatial resolution, we have created our Digital Elevation Model and using the our devised weighted Voronoi polygons we can analyse the agricultural suitability of our village-polygons based upon their relative positioning according to elevation, slope, ruggedness, and aspect.
3- Connectivity: We have map-mined reliable and extremely detailed information on transport facilities from the Third Military Mapping Survey of Austria-Hungary military map for a large territory covering our three regions from 1900s. Below you can see the extent of Southeast Europe for which we have already vectorised the road infrastructure.

We are using a less yet for our sampling purposes sufficiently detailed and already geo-referenced map for two regions in Turkey from 1899 (

We weighted soil depth and quality with 50%, agricultural suitability with 35% and connectivity with 15% in our model. A preliminary example of Plovdiv region with five categories of suitability can be seen below.

In our poster we will operate with our sampling strategy to group village polygons in to five groups of suitability and then we will test the representativeness of our model by conducting a five percent sampling strategy.

If this content appears in violation of your intellectual property rights, or you see errors or omissions, please reach out to Scott B. Weingart to discuss removing or amending the materials.