KeeWee home

Data Science

We help industry turn environmental data into strategic assets: more accurate low-cost monitoring, less uncertainty in emissions reporting, and confidence with regulators.

Talk to usSee our apps

Projects

OpenPEMS®

Free, open-source predictive emissions monitoring software. The first PEMS approved by both Alberta Environment and the Alberta Energy Regulator.

Explore OpenPEMS

Air quality monitoring (IoT)

Low-cost particle sensors, calibrated with machine learning. A neural network cut error against a government reference monitor from an RMSE of 9.93 to 3.91.

Read the case study

Advanced data analytics

Fuel use and emission intensity across 18 in situ oil sands schemes, from monthly Petrinex operating data (2015–2019).

Read the case study

Swipe to see all 3

What clients get

  • Decision-ready insight

    Complex environmental datasets turned into clear findings, from real-time monitoring to long-term trends.

  • Lower operating cost

    Analytics that uncover inefficiencies in fuel use, emissions and sensor performance.

  • Regulatory confidence

    Accurate, auditable data processing that meets or exceeds reporting requirements.

  • Sustainability reporting support

    Transparent measurement of emissions, resource use and sustainability KPIs.

  • Historical analysis and forecasting

    Legacy data analysed for trends, policy impacts and future scenarios.

  • Custom modelling

    Tailored models for emissions forecasting, production optimisation and impact assessment.

Have environmental data you want to put to work?

Tell us about your data, your monitoring setup or the question you are trying to answer.

Talk to us

Case studies

Case study · IoT

Calibrating a low-cost particle sensor with machine learning

An internet-connected Plantower PMS 5003 sensor at Calgary Varsity was compared with a government-approved SHARP monitor. The sensor overestimated readings before calibration, likely because of changing ambient humidity, so we calibrated it with machine-learning methods, including a neural network and gradient boosting.

Correlation with the reference monitor before calibration
0.78
RMSE before calibration, and after the neural network
9.93 → 3.91
Reading interval, streamed to cloud storage
6 s
Fine particles, plus temperature and humidity
PM2.5
Read the article
Case study · Data analytics

Fuel use and emission intensity in Canadian in situ oil sands

Using Petrinex operating data, we measured actual fuel use for 18 in situ schemes and found that production ramp-up and reservoir maturity played significant roles in the decline in emission intensity. The study also evaluated Alberta’s intensity-based TIER regulation. Published in Renewable and Sustainable Energy Reviews.

Monthly fuel use reported to Petrinex
2015–2019
In situ extraction schemes
18
Fuel use per m³ of bitumen, and lower average emission intensity
SAGD < CSS
Evaluated: intensity-based systems can reward production growth
TIER
Read the article

Contact us

Questions about verification, reporting, data science or training? Send us a message and we’ll get back to you as soon as we can.

required