Talk: Making Open Weather Data More Accessible: Extracting Seasonal Insights from Singapore Weather Station Data

[ talks  ]

Quick speaker notes on making open weather data more accessible - using Singapore APIs as a case study + talk reflection

Where: Open UP Global Summit

When: 1 December 2019

Location: Syntrend Creative Park, Taipei, Taiwan

Resources used:


A recap of what I went through during my talk:

  1. An intro about myself and why this talk
    • A data engineer from Singapore who loves to travel
    • How many seasons does Singapore have?
  2. Extracting Weather Data from Open Government Data Portal
    • Realtime Weather Readings from
    • Issues faced in using API
      • Nested JSON
      • Inconsistent schema order
    • Open Data - Available, Readable, Accessible
      • Is it enough for open data to be available?
  3. Weather Data API Scraping
    • What is it + Scope of API Scraping
    • Challenges faced in developing scraping tool
    • Demo
  4. Time Series Analysis Demo
    • Extracting Seasonal Insights from Scraped Weather Data
  5. Key Takeaways


openupsummit - 1 December 2019 - Making Open Weather Data More Accessible: Extracting Seasonal Trends from Singapore Weather Station Data

GitHub Repository for Demo



Open UP Summit 2019 Ong Chin Hwee - Extracting Seasonal Insights from Singapore Weather Station Data

Quick reflection

This was my first international conference, and my first time delivering a talk with a demo segment.

Why did I choose to conduct a demo? While I was preparing my talk, I was thinking of what would be the best way to bring out the core idea of making open realtime weather data more accessible to anyone - developers and non-developers included. While preparing time series visualizations from the weather readings extracted from the APIs, I thought: “Why not showcase both the usage of the scraping tool + seasonal insights from time series visualizations to keep the suspense? That’d also demonstrate how the data is made more accessible to the audience.”

I tested out both demo segments until 3am to ensure that I could showcase the demo with the latest weather readings. On the day of the talk itself, Murphy’s Law of Demos struck with a Wifi connection that kept dropping every 10 minutes and issues with my Jupyter server. I made a snap decision to truncate my demo, explain a bit about the scraping code and showcase an offline notebook of the time series visualization with all the codes executed beforehand. Not too sure if the audience picked up that I was kinda “panicking” with my demo, though at least two of the speakers couldn’t tell that it was only my first year of speaking or conducting a demo on stage (I’ll take that as a compliment!).

Would I give a talk with a demo next time? Why not, but I’d prepare way beforehand for all sorts of technical scenarios next time.

Things that could have been improved

  1. Relative smaller audience - probably because there was an ongoing workshop and a demo on Generative Adversarial Networks (GAN) at the same time.
  2. Audience was a mix of data and design people. I did try to explain the technical concepts and Python libraries during my talk so that the non-technical folks could also follow along, but I suspect at least half of the audience might have still been lost.
  3. Even some speakers didn’t know that I’m a speaker. :( Maybe I should have a more obvious speaker vibe?

Thing that went well

  1. My API scraping demo, which kept failing before my talk, worked smoothly (though slowly) on stage when I instructed it to scrap 6 days of data.
  2. At least two people approached me to say that they liked my talk + demo, so that’s better than zero!
Written on December 3, 2019