search

mapme-initiative / mapme.biodiversity

Efficient analysis of spatial biodiversity datasets for global portfolios
https://mapme-initiative.github.io/mapme.biodiversity/dev
GNU General Public License v3.0
33 stars 7 forks source link

Conflict Data #44

Closed Jo-Schie closed 1 year ago

Jo-Schie commented 2 years ago
Jo-Schie commented 2 years ago

@melvinhlwong Potential Sources:

Jo-Schie commented 2 years ago

I will temporarly close this issue as "not currently planned". We can reopen when we start to work on it. I also created an overview page of all potentially new indicators here.

goergen95 commented 1 year ago

Conflict Event Data

Definition of reasearch question

What type of event with a given acceptable spatio-temporal accuracy occured "within" an asset and what are the number of casulties?

Definition of the indicator

Different data sets encode different type of events, ranging from inter-state battles, clashes with rebels, terrorist attacks to peaceful political protests. Thus instead of speaking of conflicts, here we refer to the occurences of specific types of events. The types as well as the accuracy of the assignment of the spatio-temporal location of an event is usually encoded in the data set. For example, ACLED assigns an event where the precise location is unknown to the capital of the country it occured in. Because the proposed indicator should represent to the greatest extent possible what happend within an asset, a filtering procedure to ommit events with low accuracies of the localization needs to be put in place. Based on a monthly time-scale the occurence of events should be aggregated based on the type. Additionally, the total number of casulties (in case of violent events) or protestors (in case of peaceful protests), is aggregated as well. For the scope of this indicator in the mapme.biodiversity package, more fine-grained data is ommited (e.g. on the types or roles of parties involved).

Possible data-sources

Spatio-temporal resolution

All above mentioned data sets are distributed with point locations. The spatial accuracy, however, may vary between events, depending if the actual location could be inferred. If not, normally a surrogate location is put in place (e.g. the capital of a country or district). The temporal resolution is usually on a day-specific scale. Some events might cover a whole time period. If the temporal occurrence or duration of an event cannot be inferred, similar to the spatial location, usually a surrogate timestamp is given.

Temporal Extent:

Accessibility

ACLED and GTD need users to fill out a form to download the data manually. GDELT is multi-terrabyte data set the requires working with GoogleBigQuery for which users need a registered Google account. Only UCDP currently seems to support a download option which does not require setting up an account or user interaction with the data source.

goergen95 commented 1 year ago

Pinging @melvinhlwong and @Jo-Schie.

fBedecarrats commented 1 year ago

Great prospect! 2 reflections:

  1. should it be part of {mapme.biodiversity} or of a {mapme.conflicts}, {mapme.socioeconomic}, etc. package? I guess it also depends on how the computation size differs from other indicators already included in {mapme.biodiversity} (it might be quite similar to fire indicators...)
  2. I think it might be of interest of other researcher I am close to (in particular https://www.thomascalvo.com, who I could try to bring on board on this).

Very exciting in any case.

goergen95 commented 1 year ago

Hi, concerning your suggestion number 2 any feedback would be welcome :) for number 1 we might discuss this during the next meeting?

goergen95 commented 1 year ago

@melvinhlwong do you have any suggestions for these resources?

goergen95 commented 1 year ago

Moving on with UCDP database and a simple event count as discussed with @melvinhlwong.

Jo-Schie commented 1 year ago

Event count will include types of conflict and which other metadata?

goergen95 commented 1 year ago

Not sure yet, I started with getting the resource routine. I will inform you here once I start working on the indicator.

goergen95 commented 1 year ago

As target variables I would propose to aggregate fatality counts based on the following two levels:

The trickier question is which filters are to be supported to users. I see the following options (in the order of decreasing importance):

goergen95 commented 1 year ago

Or we might simply return all events matching the spatio-temporal extent and leave the analysis/aggregation to users.

goergen95 commented 1 year ago

At branch 44-conflict-data a first draft is available. Feedback would be really appreciated!

goergen95 commented 1 year ago

Pinging @melvinhlwong. Do you currently have a use case for testing the indicator?