Nigeria Gas Flaring Analysis Model
Author: Hussein Adeiza (mabera)
Role: Licensed Environmental Health Officer, Abuja Nigeria
Base Model: Gemma 3 1B (it)
Fine-tuned with: AutoScientist by Adaption Labs
Model Description
This is a LoRA adapter fine-tuned to interpret raw gas flaring statistics
from NOSDRA's official reports and produce structured environmental
analytical reasoning, in the style of an environmental compliance analyst.
Unlike question-and-answer formats, this model takes raw regulatory
statistics as input and produces trend analysis, regulatory effectiveness
assessment and infrastructure context as output.
Training Data
- Source: NOSDRA gas flare reports, cited via Business Post Nigeria,
Businessday NG, EnviroNews Nigeria and Naturenews Africa
- Dataset: 5 prompt-completion pairs expanded via Adaptive Data
- Languages: English, Hausa, Yoruba
- Quality improvement: 32.9% (Grade B → A)
- Kaggle: https://www.kaggle.com/datasets/yunusahusseinadeiza/nigeria-gas-flaring-environmental-interpreter
Training Metrics
- Win rate: 52% adapted vs 48% base model
- Base model: google/gemma-3-1b-it
- Method: LoRA — House Special + Reasoning Traces + Hallucination mitigation
- Dataset quality: 7.0 → 9.3 (+32.9% improvement, Grade A)
Key Finding From Source Data
After falling from 349.3 BSCF (2020) to a five-year low of 230.1 BSCF
(2022), Nigeria's gas flaring rose for three consecutive years, reaching
323.0 BSCF in 2025, a five-year high that directly contradicts progress
toward the government's own Decade of Gas flare-reduction target at its
halfway mark.
Why This Matters
Nigeria's gas flaring sits at the intersection of environmental degradation,
climate emissions and forgone domestic energy supply, 32,300 GWh of
electricity-generation potential lost to flaring in 2025 alone, even as
domestic gas supply meets under 43% of daily demand. This model is trained
to interpret raw regulatory statistics the way an environmental compliance
analyst would, identifying genuine trends, flagging when penalty
structures track rather than deter violations, and contextualizing
short-period snapshots against the fuller multi-year picture.
Credits
Powered by Adaptive Data — Adaption Labs
AutoScientist Challenge 2026
Data: NOSDRA Gas Flare Reports, Business Post Nigeria, Businessday NG, EnviroNews Nigeria
