Development Seed

Subscribe to Development Seed feed
Updated: 2 days 9 hours ago

Let's Make a Better HOT OSM Export Tool

Tue, 04/15/2014 - 19:00



As part of the Sprint Day at State of the Map US this year, a few of us -- myself, Kate Chapman from the Humanitarian OpenStreetMap Team and Chris Henrick from NYC -- collaborated on making a better interface for the HOT OSM Export Tool. While this export tool makes it easy to download ready-to-use data formats from selected OpenStreetMap extracts, its current interface can be confusing to those who are new to the tool.

Collabortin' #sotmus #sprintday at American Red Cross

Chris and I worked with Kate to understand the workflow and various features of the export tool and reimagined a better UI/UX for it. Our suggestions focused on simplifying the interface and making the workflow more apparent to the users, especially to those who are new. You can find our suggestions in the hot-export repo on GitHub, consisting of two wireframes of an improved UI.

A simpler interface for HOT OSM Export Tool. Details in GitHub repo.

Under the hood, the export tool converts the OSM data that the user choose to extract into desirable formats. (Check out the on GitHub for a complete spec.) This means if you are a GIS manager at a disaster response agency you can immediately use the available data with the software of your choice. If you are an advanced user you can specify custom presets, sql transformation files, as well as translation files as you create a new export.

The tool is written in Ruby on Rails, originally by folks at geofabrik, and HOT would love some help with the development. Talk to Kate @wonderchook or comment on the GitHub issue to get in touch. If you have any questions about design, you can reach us at @jue_yang and @chrislhenrick.

Scrubbing Afghanistan Election Data for OpenStreetMap

Fri, 04/11/2014 - 19:00



Afghans lining up at a polling station in Herat, Afghanistan on Saturday, April 5. This polling station also happens to be a prominent mosque.

Last Saturday's election in Afghanistan took place in 2,000 schools, 3,000 mosques, and countless other municipality buildings. These exact locations were made available by Afghanistan's Independent Election Commission (IEC), and represent a rich data set of religious and education institutions across the country. We're passionate about building data as infrastructure and we're beginning to work with the OpenStreetMap (OSM) community to import this data into OSM.

Satellite image above Jama Masjid of Herat. Green dots represent polling center locations across the city.

Integrating this data into OSM makes it part of a permanent data infrastructure so it's available for everything from development projects to emergency preparedness and response. The recent mapping response to the Ebola outbreak shows again the importance of geographic information for responders. This data will soon be available to humanitarians, future election observers, and developers.

We're just beginning the import process. Here's a look at the beginning of the merge process in JOSM.

Red dots show POIs over curent OSM data in Jalalabad.

State of the Map US

If you are at State of the Map US this weekend, look for Jue, Ian, Nate, and myself at the conference. We love to talk about using OSM and other open data in government and development projects.

Image: Hoshang Hashimi/Associated Press, via Mashable

Tracking the Ebola Response in Guinea

Thu, 04/10/2014 - 20:00



The outbreak of Ebola haemorrhagic fever has been unprecedented in scale. Over 151 cases and 95 deaths have been reported from Guinea alone. In this humanitarian crisis, the need for information is crucial not only to plan operations, but to coordinate with partners and educate the general public. This afternoon, we're launching with ReliefWeb an application that pulls vetted reports and analysis, and combines this with filtered Twitter and historical disaster data to create a usable disaster-tracking dashboard. It utilizes the Reliefweb API that we deployed in partnership with Phase2.

The ReliefWeb Disaster Tracker is a simple prototype that shows how open data can help improve understanding of a disaster. The early hours of any crises are critical times for preparedness and humanitarian response, and having access to information is crucial. We built this application so that it can be quickly rolled out by an independent developer during the early moments of a disaster.

The application is also open source, and we want to encourage developers to use it and improve it.

Combining data to improve access

There is a lot of useful data out there that can help disaster responders, policymakers, and media to understand a crises, but it's time-consuming to find them all. Data becomes more useful when it's in context. Alongside ReliefWeb reports and updates, we've baked in external data from Twitter, EM-DAT (the OFDA/CRED database of disasters), and UNDP's Human Development Index to give a snapshot of a country's situation.

Filtered data from Twitter enables a quick view of information coming from the organizations working on the ground to respond. To pull data from Twitter, we deployed our Twitter Server codebase, which runs continually on a small cloud server.

Open data

Nearly all of the data used in the application is open and accessible through public API's. Demographic data is made available through UNDP's Human Development Index API, Twitter data is made available through their REST API, and the ReliefWeb reports are made available through their new API. Data from EM-DAT is not completely open and does not have a method for programmatically integrating the data.

ReliefWeb Opens 20 Years of Humanitarian Data

Thu, 04/10/2014 - 16:00



Today, ReliefWeb released a public REST API that provides access to over 545,000 reports, maps, and job listings relating to humanitarian crises and response. This data, previously only available on the website, is now open for easy use in other websites, apps, and research projects.

ReliefWeb is the go-to source for humanitarian information within the United Nations. We worked with ReliefWeb and the team at Phase2 on the API that will make decades of humanitarian data open in machine-readable formats.

Data when it matters

In a crisis, speed and accessibility of information are critical. The ReliefWeb API allows responders, journalists, and policymakers to access vetted, authoritative data when they need it, in a standard format (JSON). It's built on an Elasticsearch datastore and has put a focus on speed. Providing open, fast, and programmatic access to humanitarian content can lead to better understanding of crises and can drive innovation.

Open data will allow the development of tools that make official reports on disaster response more useful. The API already powers ReliefWeb’s mobile app and it will make it easier to combine official data with data from other sources.

What’s to come

As we blogged last week, we are building an application that utilizes the ReliefWeb API to combine official disaster reports with other data sources. A simple Twitter server provides filtered disaster-related tweets in combination with ReliefWeb reports and data.

We’re excited to launch with ReliefWeb and Phase2 today. To get started building on this data, explore the developer docs and check out Phase2's post on the API application details. For an example of the new API in action, look for a public repo with live tracking the ebola epidemic in Guinea.

Better Ballot Allocation

Wed, 04/09/2014 - 19:00



On Saturday, Afghanistan went in large numbers to vote for a new president and provincial representatives. While the elections were widely seen to be a success, there were numerous reports of shortages in ballot papers.

By analyzing the placement of polling centers along with high resolution population information, we can project where people are likely to vote. This polling coverage map shows stations likely to have a shortage of ballots in blue, and stations likely to have a surplus in orange.

Overburdened Stations

This map also shows the impact of closing 352 polling stations in the week leading to elections. It shows where closed polling centers, displayed as black dots, have burdened nearby polling centers. For example, the polling center at Sangani Primary School has 3 stations to accommodate over 80,000 people. That's nearly 27,000 people per polling station. Polling stations were each issued 600 ballots.

Bright orange polygons, with many more ballots than likely voters, are areas to watch for potential ballot box stuffing.

How we built this map

We built this map from open data. There is a rich set of data available at Afghanistan Open Data Project, its GitHub repo, and NDI's Afghanistan Election Data.

For this map we used 2014 Polling Center locations and created voronoi polygons around each polling center using QGIS. The voronoi polygons show the each area for which the polling center is the closest option. This represents the swath of the country likely to go to that polling center.

To see how many people live within each Voronoi polygon, we used the WorldPop Afghanistan population layer which provides population data to 100m grids. We used the QGIS zonal statistics module to compute the population within each polygon.

Finally, we translated raw population numbers into the burden on each polling center. Polling centers have one or more polling stations. Each polling station has 600 voting ballots. Just over 53% of the population is registered to vote. On average, a polling center would need one station for every 1,115 people in the area it serves. We used TileMill to style polygons blue for areas with more than 1,115 people per station, or orange for stations with fewer than 1,115 people per station.

The methodology section provides additional information on the data and processing required to make this map.

Afghanistan Closes 748 Polling Centers

Sat, 04/05/2014 - 10:00



In response to ongoing violence and concerns over voter fraud, the IEC has closed 748 polling centers (10%) across the country. Of these, 352 cancelled polling centers were announced on Sunday. Whether or not this is the right move, it has real impact on the ability of Afghan voters to participate in their democracy. We've created a map that shows the most recently closed polling places and highlights districts left with zero or one polling center.

Baghran District

According to Afghanistan's Central Statistics Organization, Baghran District in Hilmand Province has a population of 79,300 people. Baghran District was designated 19 polling centers for 2009 elections. This year there are none.

Residents of Baghran have a few options. Voters can trek over mountains to the East or West to find small polling centers that can accommodate no more 600 male voters and 600 female voters. These polling centers are in a different province. Baghran residents who chose to vote there will be permitted to vote for President only, not for Provincial Council. Baghran voters' best bet is to travel two hours South through the valley to the town of Musa Qala. In Musa Qala they will find three polling stations equipped to handle up to 10,800 voters.

Dropped polling centers

Baghran is not alone. Twelve districts have zero or one polling center. Several as a result of the closings announced Sunday. A week ago today, Ghazni District (population 35,500) in Giro Province had been allocated 18 polling centers. These centers would have been capable of processing up to 28,800 voters. On Sunday, the IEC cancelled 17 of these polling centers, leaving just one center capable of processing up to 2,400 voters. Voters in Mandol Province in Nuristan (pop. 19,200) lost all seven of their polling centers and many others nearby.

The river valley of Northern Zabul is home to 68,500 people. Only two polling centers survived Sunday's closings.

Finding an Open Polling Center

To help Afghan voters find their nearest open polling station, we joined with other hackers to build Learn more about here.

Launching Afghanistan's First Polling Center Finder

Fri, 04/04/2014 - 10:00



Tomorrow morning Afghanistan voters will go to the polls to select a new President and representatives to Provincial Councils. Today, with the help of partners in Afghanistan and abroad, we're launching a simple tool called to enable Afghan voters to find their closest polling center. is a mapping tool that direct voters to their nearest open polling center. It's optimized for mobile phones and can easily be embedded on a Facebook page or another website. We made it easy to use and share to increase the number of people who might benefit.

Navigating to the polls

Aghanistan's Independent Election Commission (IEC) closed 748 polling centers (10%) over concerns about violence and fraud--352 of these closed stations were announced earlier this week. At the same time, the IEC removed the map of polling stations from its website. A final list of polling stations was only just published. Many voters will travel significant distances to reach a polling center and will accept real risk to do so. provides voters accurate and accessbile information about the open and closed polling centers in their area.

Built on Open Data was quickly created and distributed by a collection of concerned developers and organizations in Afghanistan and internationally, including Impassion Afghanistan and members of Lamba Labs. is possible because of open data. We drew down data from the IEC website and made it open and accessible. Even while the IEC map was down, and international group of developers could build in parallel.

Help Share

Please join us in sharing with voters in Afghanistan. Instructions to embed the site are on the about page. Let use know if you have any suggestions to improve the tool.

Afghanistan Elections: Observing the Observers

Thu, 04/03/2014 - 17:00



In advance of this Saturday's elections in Afghanistan, The National Democratic Institute (NDI) launched its 2014 Election Data site with pre-election maps showing all polling stations across the country, what is open and closed, and where observers are going to be deployed.

Observer Deployments

For the first time in Afghanistan, you can view the deployment plans of Afghanistan's largest election monitoring groups, giving a unified picture of election monitoring across every district.

Four Afghan election monitoring organizations will field 15,384 observers.

The deployments for the four largest domestic election monitoring groups are mapped, including: Afghanistan National Participation Organization (ANPO); the Afghanistan Youth National and Social Organization (AYNSO); the Transparent Foundation for Free and Fair Elections in Afghanistan (TEFA); and Free and Fair Election Foundation of Afghanistan (FEFA). This data helps to deter fraud and increase confidence in the observation effort.

Polling Centers

All 5,423 planned polling centers are mapped, including the 352 planned polling centers that were recently closed. The distribution of polling stations provides context for understanding the deployment of observers. A population layer shows where polling station closures are likely to have the greatest impact.

Open Data

The site provides all pre-election datasets for download, including polling center data from the Independent Election Commission of Afghanistan, and observer deployment plans from participating organizations.

Distance to Voter Registration Centers

Tue, 04/01/2014 - 10:00



Today was the last day to register in advance of Saturday's election in Afghanistan. Long lines and a last minute rush on registration was reported at some centers. Using data published by the Independent Election Commission of Afghanistan and population data from the UN, we can visualize travel time to registration centers. This map give an analysis of the distance from where people live to where they register to vote. Dark red areas show distances greater than 15km, while green and blue areas show shorter distances. White markers are voter registration centers.

Distance to voter registration centers across the country.

Also, check out Radio Free Europe/Radio Liberty's collection of pictures of registration centers around Kabul.

On Building The Lightest Twitter App Possible

Mon, 03/31/2014 - 20:00



There are heaps of well-written solutions for interacting with Twitter's REST API, but most are overly technical for non-developers. We wanted to build something that a disaster response technical team could roll out during a hurricane or major earthquake. So we created a lightweight Node.js program to mine Twitter and serve data that you can host anywhere. Today, we're tagging our new twitter-server application with v0.1.

Twitter-server v0.1 was designed for simplicity, it has a solid Express.js foundation, serves static files, and is now ready for development.

Serving Dynamic Content Through Static Files

We love static files. They're cheap to host and easy to maintain. With the proper safeguards, it becomes very difficult to take down your service. Twitter-server is built upon Node.js to serve dynamic content through static files. We decided to go this route because we didn't want to worry about scaling up and down during a disaster.

To serve content, Twitter-server queries Twitter every 10 seconds or so for a list of updates. This keeps the application within the bounds of Twitter's API limit. We also avoid the need for the user to sign into Twitter herself. The application can query a list of Twitter users, or do a keyword search.

The search parameters are currently hard-coded in a config file. This provides a foundation for making improvements. Our next step is to attach them to a Google Spreadsheet so they can be changed on the fly.

Once the Node.js application gets its response from Twitter, it spits out a simplified JSON file that is immediately pushed to an Amazon S3 bucket. Since it updates every 10 seconds, it continually over-writes the last file, so when new users visit your page, they get fresh Tweets.

For our purposes, the app lives on a free Amazon EC2 instance, where it happily chugs away. You could host it anywhere though, and of course you don't need to use S3 either.

All this work has been in collaboration with our friends at ReliefWeb and their upcoming improved API. Look for updates in the coming week and how this Twitter application can be used to track real-time disasters. Twitter-server is also open source, and the code is available on Github for you to use. A good chunk of the codebase is forked from @Kamicut's Seattle repo, which connects to Twitter's Streaming API and serves data through Websockets. Join the development!

Why We Opened Afghanistan's Election Data

Mon, 03/03/2014 - 18:00



Last week we launched Afghanistan Open Data Project, opening data to improve the quality and transparency of elections in Afghanistan. We hope to build a tradition and community around using data to improve the quality of all future elections in Afghanistan.

Data is the infrastructure on which we build better elections

During Afghanistan's 2009 and 2010 elections, Development Seed and NDI worked together to open up data in order to understand and describe those elections. That work was important in that it showed the extent and character of electoral fraud. We will continue this important work with NDI during 2014 elections.

Afghanistan Open Data Project is different. It is an effort to use data in advance of elections in order to improve them. We laid the first tracks, opening and visualizing data to show how it might be used in election planning and protection. Can this data make the election more inclusive? More efficient? More fair? Can better data lead to to a better distribution of polling stations? Can it power more targeted and effective efforts to improve participation by women? Can it provide political parties with data about vote patterns that rewards parties willing to actually go out and engage with voters?

We believe so.

Breakdown of 2014 polling stations by gender

Some donors and implementers still fundamentally misunderstand "open"

As we described the project to colleagues, most responded with support and enthusiasm. But we also received skepticism. One comment was that some other group was already "doing open" for these elections, as if someone had cornered the market of opening up information around elections, rendering our efforts duplicative.

That's not how Open works. It takes a community of organizations and individuals - each with unique motivations and contributions - to begin to reap many of the benefits of Open. This is just as true of open data as it is of open source software. We need more Open, not monopolies of Open.

This project is open. All the data is open for reuse and the project as a whole is open for contributions. If you are working in Afghanistan and have data, need data, or know how to work with data, please contribute to the effort.

Open Data Day Pre-Party Pics

Mon, 02/24/2014 - 18:00



Last Friday we kicked off Open Data Day DC with a pre-party at the MapBox Garage. Thanks to everyone who joined us. We had a blast and we hope you enjoyed it as much as we did! A big thanks to Julia's Empanadas for the delicious food.

Below are some photo highlights from the evening; it was the perfect chance to eat, drink, be merry, converse and rub elbows with like-minded open data enthusiasts (who says we can’t have it all?). Even the weather cooperated just in time to for the party kick-off.

Check out the photos from Derek on our Flickr album and Tatiana's Cloudup album and see if you can spot yourself, old friends, new friends and significant others.

As the weather gets better we will be planning more parties...

Better Data for Better Elections with the Afghanistan Open Data Project

Fri, 02/21/2014 - 18:00



Today we're launching the Afghanistan Open Data Project, an initiative of data and maps to better understand and prepare for Afghanistan elections. "Better data for better elections" is a theme we've advanced in several elections. The first step is to increase access to data. Five years ago, the 2009 Afghanistan presidential election data was locked in PDFs and using that data was reactionary and difficult. Today we're launching not just a portal but living experiment to prepare an open data infrastructure that supports free and fair elections.

Creating a data infrastructure before an election

With the election six weeks away, we're launching with a series of maps and data that begin to look at critical questions about preparing for an election. From access to polling centers, numbers and distribution of polling stations by gender, to voting irregularities and missing results, data can drive more professional and effective efforts by the election commission, election observers, media groups, poltical campaigns, and local citizen groups.

We hope that others will use, join, and out do this effort. Like telephones, open data has a network effect: the more people use it, the more powerful it becomes. If you are working in Afghanistan and have data, need data, or know how to work with data, please work with us.

Afghanistan Open Data Project map listing.

Visualizing settlement locations across Afghanistan. Time and distance are critical for enabling access to all citizens.

Snow cover data across Afghanistan in February combined with polling center locations. Election prepartions happening now may need to factor for access to remote areas with difficult weather conditions.

Hiring a Satellite Imagery Intern

Wed, 02/12/2014 - 18:00



We are growing. We're looking for an imagery and remote sensing intern to join the Development Seed team. You'll join a team that is pushing the bounds of how data and maps can improve disaster preparedness, humanitarian response, environmental protection, global health, urban design and managment, and the promotion of rights and democracy.

We're looking for someone to help us to push the bounds of how satellite imagery and sensor data can be processed, analyzed, and presented to inform the work of governments, NGOs, and development organizations. You'll work on really interesting problems like satellite tasking and image aquisition, automated satellite image processing, strategies for sensor and drone deployment, wrangling and integrating sensor data, and visualizing complex data.

You'll be joining a tight and growing team that gets things done. From strategists to designers and developers, we're a team that pushes the boundaries of how to make data make a difference. You will work with tools like GDAL and TileMill, while gaining experience with scripting and leveraging an open source toolbox to process imagery. You work closely with our team on strategic advising, rapid prototyping, and development of impactful web tools.

Qualities we're looking for:
  • Knowledge of remote sensing. Whether you've taken remote sensing courses or worked professionally doing imagery analysis, you bring specialized knowledge that contribute to the diversity of the team.
  • Ready to code. Processing will take work and it will take some significant code. Scripting in Python, Bash, and whatever else you need will be your tools.
  • Strong analytical ability You see the order in the chaos. You turn complex data into meaningful knowledge.
  • Strong communication skills. Github, email, and blogging are our tools we use to communicate and we're looking for someone who can contribute to both our internal and external communication strategies.
  • Self-taught learner. Many of the questions we will ask you haven't been answered. You'll figure it out. Along the way you will build the skills you need to get there.

This position is a paid, full time, three-month internship, with the option to extend for longer.

To apply

Please send the following to

  • Note describing why you're interested in the internship at Development Seed.

  • Examples of your work. Portfolio, website, or whatever medium communicates your work best.

  • Resume

  • For extra points, find a satellite image from an open data archive of your favorite place or do something cool with the location of the image above.

Understanding access to services and resources

Tue, 02/11/2014 - 18:00



I spent some time this weekend looking at global accessibility data provided by the Global Environmental Monitoring Unit from the Joint Research Centre of the European Commission. The results are some stunning insight into rural populations and accessibility to services.

View the full screen map here.

Around the world, access to services matters. From traveling to the nearest health clinic, to time to market for goods or food, travel time and distance matter to the health and livelihood of communities.

This dataset looks at travel time to nearest city with 50,000 or more people. While this particular dataset is nearly a decade old, it still offers some remarkable insight. It shows the challenges that governments face providing services from transportation to health clinics to polling stations. Similar to the malaria mapping we've done the past, this map layered with additional information would provide NGOs and policymakers provide with useful strategic and operational insights.

Variability of access across central and south Asia. The dark area on the right is the northern part of Tibet.

Open garage for open data

Mon, 02/10/2014 - 09:00



We are hosting a pre-party for Open Data Day at 6:30pm on February 21st. Join us and get excited for the weekend-long DC hackathon (It's also time to open the MapBox Garage for its #firstGarageParty2014).

Let us know you're in by RSVP now. Can't make it? Find us over the weekend to pop some open source and open data kernels for better development.

If you haven't signed up for Open Data DC, check out or register directly.

Heidi Jaafar Joins Development Seed

Thu, 02/06/2014 - 18:00



Development Seed keeps growing! To help us manage that growth we welcome Heidi Jaafar to the Development Seed team!

Heidi has been in the trenches. She has worked with local and international development organizations across ten countries. Heidi can translate between donor and doer. She is going to help us to better serve our development partners and will manage our internal systems so that we can run at full speed.

Say mabrook to Heidi on twitter.

Derek Lieu Joins Development Seed

Wed, 02/05/2014 - 23:00



Derek Lieu just joined the Development Seed team and we couldn't be more excited. Derek tells rich stories with data. He has a deep background in data journalism and has the coding and information design chops to deliver robust data products. Derek is an open data advocate who strives to make government data more meaningful to citizens.

Derek will be jumping in to lead out technical builds and to help Development Seed to push the boundaries of making data make a difference.

Say hi to Derek on twitter.

Welcoming Heidi and Derek

Wed, 01/29/2014 - 18:00



We're excited to welcome Heidi and Derek to our team!

Heidi has experience working with international development across ten countries. She is passionate about empowering organizations to focus on their strengths. She'll help us do just that by managing contracts, human resources, logistics, and accounting.

Fun fact: you can tweet at Heidi in Arabic,

Derek javascript developer and information designer working in visualization and open data advocacy. He loves tools that make everyday tasks gratifying, like a pedal wrench that fits your palm just right, or a lovely tactile keyboard. To this end, he strives to write code that makes government data more accessible to ordinary people.

Say hello to Derek on twitter.

Insights from Radio Propagation Mapping

Wed, 01/29/2014 - 18:00



Geography matters a lot to how far radio signals carry. We recently worked with DigitalGlobe to map the reach of radio signals.

To determine the extent of radio influence on populations, DigitalGlobe GIS developers produced a geospatial model based on the Longley-Rice Irregular Terrain Model of radio propagation. The model uses terrain feature and radio measurement inputs to predict radio signal reduction.

These inputs are brought to life when the output raster is styled in TileMill to reveal the dramatic change in reception levels around real-world terrain features.

Here you can see how radio signals decay as they interfere with sand dunes:

Here is a close up of abrupt changes in signal strength around a mountain peak:

We can imagine a lot of other ways to use radio propagation modeling. Emergency preparedness and response efforts need to project radio coverage to identify dark spots. Using radio propagation maps overlaid on voter registration data, citizen groups can plan more effective voter outreach efforts. The Longley-Rice Model works for frequencies from 20MHz to 20GHz. Community wireless efforts could apply similar modeling methods to plan the deployment of wifi (2.4GHz) nodes.