Debriefing: Conflict Early Warning - Early Action Practitioners Workshop


The NYU Center on International Cooperation recently held the Conflict Early Warning/Early Action Practitioners Workshop. This 4 day event featured several panel discussions and pitches relevant to the environment-security sector; from both a research and practitioner perspective. All of the sessions were recorded and available on the conference agenda. Summaries and relevant links from some of the sessions attended by DANTE team members are below.

Establishing best practices in climate security modeling (UNEP) and Weathering Risk: A Climate and Security Risk and Foresight Assessment (PIK).

This session brought together 3 project coordinators of environment-security modeling projects that provide early warning support for policy makers. Marie Schellens (United Nations Environment-Security Analyst) presented Strata: Earth Stress Monitor. Strata is an online mapping tool designed to identify environment-security hotspots through a simple coupled threshold model that analyzes metrics across human geography, displacement, unrest, conflict, and maladaptation. 

In their framework, each data layer has a corresponding early warning threshold that, when surpassed, will flag the region for a particular level of risk. Risk levels then increase with each threshold that is surpassed. The simplified modeling framework makes Strata a flexible and extensible platform. The developers intend to extend the platform in a way that allows users to include their own datasets, thresholds of interest, and easily integrate the Strata dashboard with their own visualization systems.

The next speaker was Barbora Sedova from the Potsdam Institute for Climate Impact Research (PIK). She outlined Weathering Risk; a multilateral German collaborative project designed to assist risk informed planning, increase capacity for action, and promote operational responses for resilience and peace. Weathering Risk is a complex 5 stage project that is adaptive to the needs of its users by offering multiple complexity levels in response to available funding, geographic scope, and available data. The 5 steps include:

  1. Climate Impact Analysis to identify potential impacts, risks, and regions. This also includes an Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) that reviews climate scenarios.

  2. Contextual analysis consisting mostly of a literature review.

  3. A randomForest machine learning model used to select metrics of interest.

  4. Development of foresight for potential climate scenarios.

  5. Identification of potential responses.

This ambitious project is currently in development and aims to be fully transparent with their methodologies.

The final speaker was Corey Pattison from World Bank West Africa. He spoke about the Climate Change Fragility and Violence project’s efforts in West Africa. Corey is coordinating 3 localized modeling projects in the Lake Chad, Sahel, and Guinea regions. These modeling tools are designed to support early responders interested in predicting conflict by strengthening joint datasets and integration in the regions, mapping these data, implementing case studies for greater contextual information, and developing community based tools for digital data collection and information sharing.

Their methodology includes 4 primary steps:

  1. Generate PCA components for sub-regions from a large pool of datasets covering exposure to conflict, sensitivity to disturbance, and adaptive capacity.

  2. Classify sub-regions into vulnerability clusters based upon the PCA results using Bayesian hierarchical clustering.

  3. Model these classifications against ACLED and UCDP conflict data for 2000-2019.

  4. Identify true and false positive cases with high perceived value.

  5. Commission case studies to provide greater contextual information and subsequent model refinement.

The pool of candidate metrics for the model includes some interesting measures: travel time to nearest conflict, short term precipitation change, travel time to surface water, medium to short term forest losses, road length and density, travel time to health services, and nighttime lights data.

Ultimately, the World Bank hopes to use these analysis by integrating them into numerous development projects through assisting preparation and design of investment types and populations, integration with KMP, knowledge support to local stakeholders, and development of the CDD app.

Predicting Conflicts: The Water, Peace, and Security Partnership Early Warning Tool

The World Resources Institute (WRI) chaired this session to introduce global and regional early warning conflict tools they’re developing as part of the Water, Peace and Security (WPS) partnership. WPS’ objectives are to develop innovative tools, raise awareness, increase local capacity, and foster peace by supporting dialogue and conflict resolution.

A confluence of events are placing extreme pressures on water resources. This includes factors like more variability in water availability due to climate change, poor governance, and adaptive measures that may have long term damaging effects (deeper wells, damning). In response, WPS set out to develop global and regional early warning tools to help manage increasing pressures on global water resources.

WPS’ global tool predicts conflict 12 month out using a randomForest machine learning model. They initially hoped to develop a model that only predicted water conflict, although they quickly realized that was too narrow of a scope and expanded to all conflicts. In their framework, the spatial unit of analysis is 2nd level administrative units, the response metric is 10 conflict fatalities (as reported by ACLED), and over 80 datasets serve as potential independent variables fed into the randomForest algorithm; including several real time climate and water security datasets. For a more detailed look at their methodologies, please refer to their online description.