Description
Accurate maps of past land use are necessary for studying the impact of anthropogenic land-cover Changes on climate and biodiversity. The main source of pre-industrial human land-use is agriculture. Archaeological data could provide a way of estimating the extent of past agricultural land-use. To investigate the possibilities and limitations of available archaeological data we have conducted a pilot study using Swedish data from 2000 B.C. (stone age) to present day. The Swedish National Heritage Board maintains a large database ofarchaeological nds, dated and classied into dierent historic uses. These data can be seen as presence only data, i.e. a lack of archaeological nds is never reported. A point-process model can be used to estimate´the relative abundance of agricultural land use in Sweden since the late stoneage (absolute levels are not estimatable for presence only data). A complication with the archaeological data is that dating ranges from very exact (radiocarbon or dendrochronology) to very imprecise. To model the data we use a log-Gaussian
Cox process where the latent process is modelled using a spatio-temporal Gaussian Markov random eld. The exact date of the archaeological nds are seen as unknwon and estimated together with the latent eld and the unknown parameters using Markov chain Monte Carlo, incorporating a Metropolis-Adjusted Langevin Algorithm. The method is evaluated using a simulation study, where we also investigate how the amount of imprecies dates aects estimation, before being applied to the archaeological data.
| Period | 2016 Jul 18 |
|---|---|
| Event title | The 26th Annual Conference of the International Environmetrics Society |
| Event type | Conference |
| Location | Edinburgh, United KingdomShow on map |