Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic nowcast of the New Zealand greenhouse gas inventory

Malcolm Jones, Hannah Chorley, Flynn Owen, Tamsyn Hilder, Holly Trowland, Paul Bracewell

TL;DR

This work tackles the lag between NZ Greenhouse Gas Inventory releases and current emissions by introducing a daily, indicator-based nowcasting framework. It compares three non-parametric models and selects Extra Trees (ET) for its superior accuracy and stability in predicting sector emissions (Agriculture ~1.3% MAPE, Energy ~3.2% MAPE) with under two months latency. The approach yields interpretable relationships between indicators and emissions, provides uncertainty bounds relative to inter-annual variation, and offers actionable, high-frequency estimates that could inform policy. While ET demonstrates strong performance and repeatability across NZGGI editions, the study notes limitations in data coverage and explanatory depth, outlining a clear roadmap for extending to additional sectors and gases to strengthen near-term emissions accountability.

Abstract

As efforts to mitigate the effects of climate change grow, reliable and thorough reporting of greenhouse gas emissions are essential for measuring progress towards international and domestic emissions reductions targets. New Zealand's national emissions inventories are currently reported between 15 to 27 months out-of-date. We present a machine learning approach to nowcast (dynamically estimate) national greenhouse gas emissions in New Zealand in advance of the national emissions inventory's release, with just a two month latency due to current data availability. Key findings include an estimated 0.2% decrease in national gross emissions since 2020 (as at July 2022). Our study highlights the predictive power of a dynamic view of emissions intensive activities. This methodology is a proof of concept that a machine learning approach can make sub-annual estimates of national greenhouse gas emissions by sector with a relatively low error that could be of value for policy makers.

Dynamic nowcast of the New Zealand greenhouse gas inventory

TL;DR

This work tackles the lag between NZ Greenhouse Gas Inventory releases and current emissions by introducing a daily, indicator-based nowcasting framework. It compares three non-parametric models and selects Extra Trees (ET) for its superior accuracy and stability in predicting sector emissions (Agriculture ~1.3% MAPE, Energy ~3.2% MAPE) with under two months latency. The approach yields interpretable relationships between indicators and emissions, provides uncertainty bounds relative to inter-annual variation, and offers actionable, high-frequency estimates that could inform policy. While ET demonstrates strong performance and repeatability across NZGGI editions, the study notes limitations in data coverage and explanatory depth, outlining a clear roadmap for extending to additional sectors and gases to strengthen near-term emissions accountability.

Abstract

As efforts to mitigate the effects of climate change grow, reliable and thorough reporting of greenhouse gas emissions are essential for measuring progress towards international and domestic emissions reductions targets. New Zealand's national emissions inventories are currently reported between 15 to 27 months out-of-date. We present a machine learning approach to nowcast (dynamically estimate) national greenhouse gas emissions in New Zealand in advance of the national emissions inventory's release, with just a two month latency due to current data availability. Key findings include an estimated 0.2% decrease in national gross emissions since 2020 (as at July 2022). Our study highlights the predictive power of a dynamic view of emissions intensive activities. This methodology is a proof of concept that a machine learning approach can make sub-annual estimates of national greenhouse gas emissions by sector with a relatively low error that could be of value for policy makers.
Paper Structure (41 sections, 6 figures, 3 tables)

This paper contains 41 sections, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Overview
  4. Indicators
  5. Model selection
  6. Models
  7. Cross validation
  8. Model outcomes
  9. Results
  10. Framework to assess models to predict sector-level greenhouse gas emissions in national inventories
  11. Prediction accuracy
  12. Model Stability
  13. Over-fitting
  14. Selection
  15. Uncertainty in model compared to inter-annual variation
  16. ...and 26 more sections

Figures (6)

  • Figure 1: Heatmap of correlation matrix showing high collinearity between indicators in each sector.
  • Figure 2: Box plots picturing distribution of fitted coefficients (for LS, bounded-variable least squares) and indicator importance (for RF and ET, random forest and extra trees, respectively) values as a measure of stability of each model.
  • Figure 3: Inferred relationships between indicators and predicted emissions. Outliers are identified and labelled with their year.
  • Figure 4: Emissions estimates for each sector modelled in our analysis according to NZGGI (the actual data) and extra trees--ET (the predicted data). This demonstrates the greater timeliness of our estimates both by being more frequent and having a lower latency (nowcasting).
  • Figure 5: Overlay of COVID-19 alert levels in Auckland and elsewhere in New Zealand with daily national traffic volume at selected sites across the country. The traffic volume is expressed as a percentage of the 1990 average. It is involved in the following indicators: LEE (Liquid and electric energy) and ETV (Emissions associated to national traffic volume).
  • ...and 1 more figures