Every year millions of people in the US are affected by power outages, disrupting the economy and daily life. Many of these outages are caused by events such as strong winds, heavy rains, thunderstorms, floods, tropical storms and hurricanes. At the University of Connecticut an outage prediction model (OPM) has been developed for forecasting outages during storms. The OPM has been operational since 2015 serving utilities in the Northeastern US. It uses variables describing weather events, infrastructure, land cover and elevation. Non-parametric machine learning (ML) ensembles generate the predictions. The first version of the model served Connecticut exclusively and was characterized by large uncertainty in predictions due to the dataset limitations of a small service territory and limited historical dataset. Over time, the model expanded to include utility territories in Massachusetts and New Hampshire, the dataset grew, the understanding between environmental forcing and outages improved, and probabilistic operational forecasts began to be produced. The relationship between UConn and the utility stakeholders has grown to where operational forecasts are now used as part of response planning to storm events by the utility. This work leverages knowledge from the UConn OPM and utilizes a similar ML framework in combination with non-utility-owned customer outage data to build a community OPM for predicting customer outages along the US Eastern Seaboard for large scale events. Proxies for proprietary infrastructure are used including road and publicly available transmission line data. Variables including tree type and ecoregion data are used to account for regional diversity of the larger domain. To validate the customer outage reference data, correlations are shown between customer outages and utility trouble spots in the Northeast where outage data from utilities is known. Model performance evaluated at county and state levels shows that the model is capable of predicting the peak number of customer outages with great accuracy, demonstrating promise for the ultimate goal of determining return periods of outages under current and future climate scenarios to help the public and utilities with resiliency and response planning.

The Blue Nile Basin, Ethiopia, whose inter-annual variability in local precipitation has resulted in droughts and floods that lead to economic and food insecurity, is the area of interest for our NSF-PIRE project, which aims to develop novel forecast technologies to mitigate the stresses to local communities. As part of the PIRE project, a Citizen Science Initiative (PIRE CSI) was established in June 2017, a project that trains high school students in hydrologic data collection under the guidance of classroom teachers and graduate students and professors from Bahir Dar University in four watersheds of interest, located south of Lake Tana, Ethiopia. Four MSc graduate students were selected from Bahir Dar University and trained nine high school students who were nominated taking into account gender and the proximity of their schools to the watersheds. High school students are currently collecting soil moisture data using TDR, river stage measurements using optical levels and groundwater levels using shallow water level meters. The data collection is supported by an app (B-WING), developed specifically for the needs of the project. College-ready activities are being planned for the high school students, i.e. inviting them to Bahir Dar University to analyze some of the data, present their work at a workshop, and familiarize themselves with the university experience. Recently, the PIRE CSI was extended to involve local farmers as “citizen scientists”, collecting soil moisture data using low-cost, soil moisture sensors developed in-house at the University of Connecticut, that have been installed in 12 locations and two soil depths (20 cm and 40 cm). The collected data will be used for the initialization and validation of the hydrological models developed in the region. The PIRE CSI promotes the empowerment of local communities and establishes long-lasting partnerships between scientists and stakeholders. It is believed that the co-generation of knowledge may contribute to higher rates of forecast adaption by the local farmers and may trigger the student’s interest in STEM and encourage their uptake of scientific careers. Acknowledgment: This material is based upon work supported by the National Science Foundation under Grant No. 1545874.