PORTFOLIO
A sampling of my technical work.
Nyagisozi High School, Nyanza, Rwanda
CONTENTS
Conflict Analysis Data Engineering and Web Map - Democratic Republic of the Congo
I created this web map using JavaScript and MapBox GL JS to visualize conflict events form the Democratic Republic of the Congo that Mercy Corps' Congo Analysis Team collects and annotates for the humanitarian community. The data lives and is updated in an Excel spreadsheet on SharePoint, so I created a fully automated and containerized pipeline to pull the data into AWS S3 periodically, clean the data, and process it for integration into the web map using Lambda functions, Elastic Container Service (ECS) and Docker. The map has been featured on their website and is used by the team to analyze events and their humanitarian impact.
At Mercy Corps, I developed this web application for a System Dynamics model we created to model household resilience among agro-pastoral communities in Somalia. The tool was created using JavaScript and the SD Everywhere package for JS. The "flight simulator" allowed humanitarian users in Somalia to model potential shock scenarios and impacts on the livestock sector and household impacts to inform program interventions.
Interactive Web Application for Livestock System Dynamics Model
Selected Report Contributions and Spatial Analysis from Mercy Corps
As the only geospatial person at Mercy Corps Headquarters, I drive strategy and data conversations, but I also do a fair amount of spatial analysis and mapping support to programs and other Mercy Corps and HQ departments for reports, proposals, or use in program implementation. My work has been featured in a number of reports for advocacy and crisis analysis for some of the key contexts where Mercy Corps works. Below is a selection of those reports and maps.
I reprised PostGIS solid waste routing work in Dar Es Salaam (see below) for Zanzibar on a contract with the World Bank supporting Open Map Development Tanzania (OMDTZ). The analysis and report supported World Bank efforts to invest in solid waste infrastructure and services on Unguja island. I created PostgreSQL database and routing table for the island using roads pulled from and edited in OpenStreetMap and ran various routing queries and analyses using an algorithm that included information like road surface type and class and ground validated estimated travel time. Select maps from the project are included below.
Solid Waste Routing for Infrastructure Planning in Zanzibar
Solid Waste Routing with PostgreSQL and OpenStreetMap in Dar es Salaam, Tanzania
This is a project done while I was working with the Humanitarian OpenStreetMap Team (HOT) in Tanzania during the Summer of 2019 in Dar es Salaam, Tanzania. The goal was to estimate how long it would take to transport solid waste from any point in the city to the only formal "dump" in the city, known as Pugu Hills. OpenStreetMap data was used for the road layer and routing model created in PostgreSQL using the pgRouting tools. The map below (best viewed full screen) is the final visual product of our work on the project and a blog post I wrote about the work, which was posted on the HOT website is included below that. The full Markdown for the blog tutorial as well as SQL code snippets can be found here.
From January to May 2020, I worked on a student consulting team on a project for the Wildlife Conservation Society (WCS), who wanted us to develop baseline trends for the 7 year Sustainable Wildlife Management Programme (SWM). The team worked on 9 sites in 10 countries and I managed all the analysis and map creation for the Democratic Republic of the Congo Ituri site. I used ArcGIS Pro, Terrset and QGIS to do the analysis for this project. Below are the slides that I prepared for the DRC project site.
Project Completed by Clark GIS Consulting Team
Baseline Social & Environmental Trends for Sustainable Wildlife Management (SWM) Programme
Download
Full Presentation
Detecting Informal Refugee Settlements with Medium Resolution Satellite Imagery Using Machine Learning
As a final project for Clark's Advanced Remote Sensing course, I worked with colleagues for Humanitarian OpenStreetMap Team's Mapping The Missing Millions project to create a classification schema for medium resolution Sentinel imagery using machine learning in Google Earth Engine. Most of the machine learning classification being done today is using high resolution imagery (1-5 m or less per pixel), which can yield incredibly accurate building footprints or roads. However, this hi-res imagery is very expensive and not always available globally. Using the Random Forest algorithm, we were able to highlight new settlement in refugee camps in 2020 based on training from settlements in 2016, which can be further extrapolated to highlight new settlement areas in places that are not well documented. Our poster detailing the project is below, and it can be downloaded by clicking on it. Example code for the final classification can be found here.
Collaborators: Priscilla Ahn, Jiena He, Zhenhua Meng
This was a practical project for Clark University's Geospatial Analysis with R course and was a collaboration with Professor Lyndon Estes' Agroimpacts project to create a script that would automatically give QA/QC statistics on the digitization of agricultural plots created by Amazon Mechanical Turks workers. All of the coding was done in R and the HTML product below was created using RMarkdown. The full package code can be found on Github here.
Collaborator: Zhenhua Meng and Miles Weule-Chandler
How Well Do Crowdsourcers Map Individual Fields in Zambia?
Quantifying Future Crop Vulnerability in East Africa
This project fulfilled the final component of the Advanced Raster course at Clark University. My collaborator and I used Terrset's climate change modeling and land change modeling tools to create constraint and factor criteria for a multi-criteria evaluation that aimed to show the areas that would be most and least suitable for 3 key crops in East Africa: maize (corn), wheat, and sorghum. The results conclusively indicated that sorghum would be the most resilient crop long-term and wheat would be the least, with corn seeing less high-yield productivity zones.
Collaborator: Wei Hong Loh
Download
Full Report
After working in Tanzania with HOT, they expressed that they would like to have a way to visualize the vast amount of data collected through community mapping for their Ramani Huria project in an efficient way. I knew I would be doing a directed study in PostgreSQL and PostGIS in the Fall of 2019, so my school colleagues and I decided to tackle this as our final deliverable for the study. We set up an AWS ec2 instance where we loaded data into PostgreSQL, which could then be sent to a geovisualization server (GeoServer in this case for webmap visualization in Leaflet). This project is very much a work in progress and I hope to have a web hosted map that I can share here in the near future. Below is the documentation on our project and the JavaScript code for our preliminary webmap.
Collaborators: Bryce Stouffer, Sam Watson, Wei Hong Loh and Jordan Frey
Visualizing Ramani Huria's Crowdsourced Data with AWS, PostgreSQL and Leaflet
I created these maps for a friend and indigenous rights activist studying at Brandeis University who is from the Karrayyu tribe, a nomadic herding group in central Ethiopia. Lake Basaka (called Nogoba in the Oromo language) has been steadily growing in size since the establishment of a sugarcane plantation near its shores. On-the-ground accounts report a harsh smell and the spoiling of several water boreholes used by the Karrayyu and others, indicating that the contaminated lake water may be seeping into the water table. These maps help visualize the problem.
The Rising Waters of Lake Basaka (Nogoba) in
Oromia, Ethiopia
(Ongoing)
Analyzing Spatial Trends in Violence Against African Americans, 1890s and 2010s
These maps are part of a class project attempting to visualize and determine any relationship between lynchings of African Americans (1890-1900) and police shootings (2013-2018) in the American South. The project involved extensive data preprocessing, spatial statistical analysis and cartography, all using ArcGIS Pro
Collaborator: Iolanthe Brooks
Download
Full Report
Recommending New Primary Health Facilities in Kutapalong Rohingya Refugee Camp, Bangladesh
The maps at left are products of an analysis on the Kutupalong using Network Analyst to recommend new sites for primary and secondary health facilities. Skills utilized included ArcGIS Network Analyst, Spatial Statistics tools, and advanced geoprocessing techniques.
Collaborator: Miles Weule-Chandler
Download
Full Report
Detecting Sea Level Rise in The Maldives
These maps are part of a remote sensing project using indices like NDWI and Tasseled Cap plus other methods attempting to detect increase in sea level rise in the world's flattest country.
Collaborator: Jordan McCutcheon
