Preview — Springfield Science Hero
Springfield
by Naik Group
Est. 1939 Do Something Green Today
Data Science Research  ·  3-Month Study  ·  28 Plant Species  ·  MSc Published
The science
behind the green
85 years of knowledge — now backed by data.

We didn’t just tell customers that plants are good for them. We set up sensors, measured the air for three months straight, and let the numbers speak. Here is everything we found — and what it means for your home.

~70
ppm
CO2 reduced with green walls
~7
µg/m³
PM2.5 reduction measured
85%
accuracy
ML model validation
28
species
scientifically ranked
Why we did this

We wanted proof, not just tradition.

At Springfield by Naik Group, we have been transforming green spaces and urban landscapes since 1939. For over 85 years, we have told customers that plants are good for them — that they purify air, reduce stress and make homes and workplaces healthier.

But we had never measured it. We had never put a sensor in a room and watched what plants actually do to the air you breathe. So in 2018, we did exactly that.

“What we found was more striking than we expected.”
Study duration
November 2018 – February 2019 (3 months)
What was measured
PM1, PM2.5 and CO2 — via continuous sensors
Study setup
2 identical spaces — one with green wall, one without
Analysis method
K-means clustering · Linear regression · ML classifiers
Conducted as part of
MSc Data Science — published research
CO2 finding PM2.5 + health Correlation Plant clusters Machine learning Air calculator Plant database Shop the kit
Preview — Section 2 Key Findings
What we discovered

Five findings that changed how we think about plants

Click any finding to jump straight to the full data, charts and explanation — or scroll through each one below.

Finding 01
~70 ppm
CO2 reduced — every single week for 3 months
Green walls consistently kept CO2 near fresh outdoor levels throughout the entire study period.
Finding 02
~7 µg/m³
PM2.5 lower — linked to 13,310 avoidable hospitalisations
Every week, air in the green wall space had measurably less harmful fine particulate matter.
Finding 03
0.99
All pollutants move together — fix one, fix all
PM1 and PM2.5 correlate at 0.99. A green wall doesn’t just help one pollutant — it helps everything.
Finding 04
3 tiers
Not all plants are equal — science ranked all 28 species
K-means clustering revealed 3 clear performance groups across CO2 absorption, leaf area and pollution tolerance.
Finding 05
85%
Machine learning confirmed our plant rankings with 85% accuracy
Three ML classifiers — Random Forest, XGBoost and Extra Trees — validated that our clusters are real, not random.
3 months. 2 spaces. Continuous sensors. — This is what real evidence looks like.
Preview — Section 3 CO2 Finding
Finding 01 CO2 Concentration — 3-month weekly measurement study
CO2 concentration

Green walls reduced CO2 by ~70 ppm —
every week, for 3 months straight

This was the most dramatic finding of our study. Spaces with a green wall consistently maintained lower CO2 levels than identical spaces without — not just occasionally, but every single week of the 12-week study period.

~490
ppm CO2
Average without green wall
~420
ppm CO2
Average with green wall
~70
ppm
Consistent weekly reduction
CO2 weekly trend — with vs without green wall Live chart
Weekly CO2: With GW ~420ppm, Without GW ~490ppm throughout Nov 2018 – Feb 2019.
With green wall (~420 ppm avg)
Without green wall (~490 ppm avg)
Study period: November 2018 – February 2019. Weekly averages of continuous sensor readings.
Add your original research chart here
Upload the CO2 weekly trend image from your study to show alongside the interactive chart. In WordPress: click the + button → Image block → Upload.
What the data shows

The green wall space averaged ~420 ppm CO2 across the entire study period. The identical space without a green wall averaged ~490 ppm — a difference of approximately 70 ppm that held consistent week after week.

Why 70 ppm matters: Normal outdoor air contains approximately 415 ppm CO2. Indoor spaces without plants regularly reach 600–1,000 ppm, causing drowsiness, difficulty concentrating and reduced cognitive performance. Our green wall space stayed close to outdoor levels — the difference between a stuffy office and a breath of fresh air.

What makes this finding particularly compelling is its consistency. The 70 ppm gap was not a one-week result — it was sustained across all 12 weeks of monitoring, through November, December, January and February. This tells us the effect is reliable and repeatable, not a coincidence.

The correlation analysis later confirmed this: CO2 levels with and without the green wall correlated at 0.91 — they tracked the same environmental trends, but the green wall consistently kept levels lower throughout.

415 ppm
Normal outdoor CO2 level our green wall matched
12 weeks
Continuous monitoring — the result was consistent every single week
0.91
Correlation between CO2 levels — with and without green wall
Next finding: PM2.5 was measurably lower too — and here’s why that matters for your family’s health.
Preview — Section 4 PM2.5 Finding
Finding 02 PM2.5 Concentration — fine particulate matter + health impact data
PM2.5 concentration

PM2.5 was measurably lower — every week —
and here’s why that matters for your family

Fine particulate matter (PM2.5) is the most medically significant air pollutant in Indian homes. Our study found it was consistently and significantly lower in the green wall space — not just once, but across the entire 12-week monitoring period.

~75
µg/m³
PM2.5 without green wall
~68
µg/m³
PM2.5 with green wall
~7
µg/m³
Consistent weekly reduction
35
µg/m³
Safe limit — Indian NAAQS standard
PM2.5 weekly trend — with vs without green wall Live chart
PM2.5 weekly: With GW ~68 µg/m³, Without GW ~75 µg/m³ throughout study period.
With green wall (~68 µg/m³)
Without green wall (~75 µg/m³)
Weekly averages of continuous sensor readings. Nov 2018 – Feb 2019.
PM1 concentration — median comparison
With GW
Without GW
The green wall space had a lower median PM1 level and tighter spread — more stable, predictable air quality.
What the data shows

Across all 12 weeks of measurement, the green wall space averaged ~68 µg/m³ PM2.5. The identical space without a green wall averaged ~75 µg/m³ — a difference of approximately 7 µg/m³ that appeared without exception every single week.

Why 7 µg/m³ matters: The WHO guideline for annual PM2.5 exposure is just 5 µg/m³. India’s national standard is 40 µg/m³ (24-hour average). A sustained 7 µg/m³ reduction across an entire year of exposure significantly reduces cumulative health risk — moving readings meaningfully closer to safe thresholds.

What makes this finding powerful is its consistency. The reduction was not a seasonal effect or a one-week anomaly — it appeared in November, December, January and February. This makes it a reliable, repeatable result that customers can trust.

The PM1 boxplot (ultra-fine particles) tells the same story. The median PM1 level in the green wall space was lower, and the interquartile range was tighter — meaning not only lower pollution but more stable and predictable air quality overall.

Our correlation analysis confirmed the relationship: PM1 and PM2.5 correlate at 0.99 — nearly perfectly. What removes one removes both simultaneously.

Why this matters — the health data

59,400 respiratory hospitalisations annually.
13,310 are avoidable.

India records over 59,400 respiratory disease hospitalisations per year linked to PM2.5 pollution. Our cross-referenced analysis of public health data reveals that 13,310 of these cases are avoidable if indoor PM2.5 levels stay below 35 µg/m³. Green walls — using the plants in our research — measurably reduce PM2.5 levels. This is not just gardening. It is a health investment.

59,400
Annual RD hospitalisations
13,310
Avoidable if PM2.5 <35
2,360
Pneumonia cases avoidable
2,970
Avoidable in 65+ age group
Condition Annual hospitalisations Avoidable cases
All respiratory diseases59,40013,310
Pneumonia13,3102,360
COPD13,8102,730
Ages 15–6410,9501,970
Ages 65+26,3902,970
Source: PM2.5-caused RD hospitalisation data 2015–2020. Avoidable cases calculated at PM2.5 threshold of 35 µg/m³.
WHO guideline
5 µg/m³
Annual PM2.5 limit recommended by the World Health Organisation
India NAAQS standard
40 µg/m³
India’s national ambient air quality standard (24-hour average)
Springfield green wall
~7 µg/m³
Reduction achieved — moving indoor air measurably closer to safe levels
Next finding: All three pollutants move together — which means one green wall improves everything at once.
Preview — Section 5 Correlation
Finding 03 Correlation analysis — PM1, PM2.5 and CO2 move together
Correlation analysis

All three pollutants move together —
fix one, you fix all of them

This was perhaps our most surprising finding. PM1, PM2.5 and CO2 are so strongly correlated that a single green wall intervention simultaneously improves all three — you don’t need separate solutions for different pollutants.

0.99
Correlation between PM1 and PM2.5 — near-perfect co-movement
When PM1 goes down, PM2.5 goes down almost identically. When CO2 drops, particulate matter drops too. One green wall. One intervention. All pollutants improved simultaneously.
Correlation heatmap — all 6 variables From our study
PM1 With GW
PM1 Without GW
PM2 With GW
PM2 Without GW
CO2 With GW
CO2 Without GW
PM1 With GW
1.00
1.00
0.99
1.00
0.96
0.86
PM1 Without GW
1.00
1.00
1.00
1.00
0.95
0.86
PM2 With GW
0.99
1.00
1.00
1.00
0.93
0.83
PM2 Without GW
1.00
1.00
1.00
1.00
0.93
0.84
CO2 With GW
0.96
0.95
0.93
0.93
1.00
0.91
CO2 Without GW
0.86
0.86
0.83
0.84
0.91
1.00
0.83
1.00
Red = strongest correlation (near 1.0). Blue = strong correlation (0.83+). All values indicate very strong relationships between pollutants.
What the heatmap shows

Every cell in this heatmap shows the correlation between two variables from our study. A value of 1.00 means perfectly linked — when one goes up, the other goes up identically. A value of 0.83 still represents an extremely strong relationship.

The critical insight: PM1 and PM2.5 correlate at 0.99 — near-perfect co-movement. CO2 with green wall correlates with particulate matter at 0.93–0.96. This means the green wall is not solving one problem — it is improving the entire indoor air quality ecosystem simultaneously.

Notice that CO2 Without Green Wall shows slightly lower correlations (0.83–0.86) with particulate matter — suggesting the green wall creates a more integrated, stable air quality environment where all pollutants respond together to the presence of plants.

From a practical standpoint this is powerful news: you do not need to choose which pollutant to target. A single well-chosen green wall installation improves CO2, PM1 and PM2.5 at the same time.

Key correlations — plain language summary
Measurement pair Correlation Strength What it means
PM1 With GW ↔ PM2 With GW
0.99
Near-perfect PM1 and PM2.5 move in near-perfect lockstep
PM1 Without GW ↔ PM2 Without GW
1.00
Perfect Completely identical movement without green wall
CO2 With GW ↔ PM1 With GW
0.96
Very strong CO2 and PM1 strongly linked in green wall environment
CO2 With GW ↔ CO2 Without GW
0.91
Very strong CO2 responds to same environmental trends — GW keeps it lower
CO2 Without GW ↔ PM2 Without GW
0.84
Strong Lowest link — without plants, pollutants are less integrated
One solution, total impact
Because all pollutants are correlated, a single green wall installation simultaneously reduces CO2, PM1 and PM2.5. You don’t need three separate air-quality interventions.
Consistent year-round
The high correlations held across all 12 weeks — November through February — covering different seasons and conditions. The effect is not seasonal, it is structural.
Green walls create stability
The green wall environment showed slightly higher correlations between pollutants — suggesting plants create a more stable, integrated air quality system overall, not just lower numbers.
Next finding: Not all plants are equal — science ranked all 28 species into 3 performance tiers.
Preview — Section 6 Plant Clusters
Finding 04 Plant clustering — K-means analysis of 28 species across 3 performance tiers
Plant cluster analysis

Not all plants are equal —
science ranked all 28 species into 3 performance tiers

Our K-means clustering study revealed that plant species fall into three clearly differentiated groups based on their air purification performance. Choosing the right cluster makes the difference between a decorative plant and a genuinely effective air purifier.

Algorithm
K-means (k=3, iter=10)
Features used
CO2 absorption · Leaf area · Pollution tolerance
Species studied
28 plant species — all clustered
Validated by
Extra Trees classifier — 85% accuracy
🏆
Cluster 3
Top performers
Highest air purification
CO2 absorption
Leaf surface area
Pollution tolerance
CO2 absorption range High (1.8–2.6)
Leaf surface area Large (3/3)
Pollution tolerance High
Species count 10 species
Key species
H. helix T. vulgaris E. amygdaloides H. stemii P. scolopendrium H. officinalis
✦
Cluster 2
Good performers
Solid air purification
CO2 absorption
Leaf surface area
Pollution tolerance
CO2 absorption range Medium (0.4–0.7)
Leaf surface area Small (1/3)
Pollution tolerance Medium
Species count 4 species
Key species
B. buxifolia S. media S. hybrid
Cluster 1
Moderate
Aesthetic + mild benefit
CO2 absorption
Leaf surface area
Pollution tolerance
CO2 absorption range Low (-0.8 to 0)
Leaf surface area Mixed (1–2/3)
Pollution tolerance Lower
Species count 14 species
Key species
B. sempervirens H. albicans G. odoratum P. veris
Visual cluster separation — from our research data
Each dot represents one of the 28 plant species. The clear separation between colour groups confirms the clusters are genuine and meaningful — not random.
CO2 Absorption vs Leaf Surface Area
Scatter plot of CO2 absorption vs leaf surface area by cluster.
Tolerance to Pollution vs CO2 Absorption
Scatter plot of pollution tolerance vs CO2 absorption by cluster.
Cluster 3 — top performers (10 species)
Cluster 2 — good performers (4 species)
Cluster 1 — moderate (14 species)
Why leaf surface area is the key variable

Plants absorb CO2 and particulate matter primarily through stomata — tiny pores on the surface of leaves. These pores exchange gases directly with the surrounding air, filtering pollutants as air moves across the leaf surface. A plant with a larger leaf surface area has more stomata in contact with polluted air, making it significantly more effective per plant at purifying the space around it. This is why our research specifically used leaf surface area as one of the three key clustering variables — and why Cluster 3 plants, with their large leaves, consistently outperform smaller-leaved species.

Next finding: Machine learning confirmed these clusters are real — not random groupings — at 85% accuracy.
Preview — Section 7 Machine Learning
Finding 05 Machine learning validation — Extra Trees, Random Forest & XGBoost classifiers
Machine learning validation

We trained an AI to rank plants by
air purification. It was 85% accurate.

To confirm our plant clusters were genuinely meaningful — not just a mathematical artefact — we trained three machine learning classifiers to predict which cluster any plant belonged to based solely on its measurable characteristics. The results validated everything.

85%
Extra Trees accuracy
What this means
Our plant clusters are real — confirmed by machine learning, not just data intuition.
85% accuracy means the model correctly identified a plant’s performance tier 85 times out of every 100 — using only measurable biological characteristics. This rules out the possibility that our clusters were random or coincidental.
🏆 Best performer
Extra Trees
Extremely Randomized Trees
85%
Classification
accuracy

Builds multiple decision trees using random feature subsets at each split — reducing variance and generalising better to unseen plant data. Our most accurate model.

Used in Springfield Plant Finder
Runner up
Random Forest
Ensemble Decision Trees
81%
Classification
accuracy

A robust ensemble method that builds many decision trees and averages their predictions — strong baseline with 81% accuracy on our plant dataset.

Validated cluster integrity
Runner up
XGBoost
Gradient Boosting
81%
Classification
accuracy

Gradient boosting builds sequential trees, each correcting the previous one’s errors. Matched Random Forest at 81% — further confirming our clustering results.

Confirmed cluster patterns
Classifier accuracy comparison
All three models confirm the plant clusters are genuine — accuracies of 81–85% on unseen data
Classifier accuracies: Extra Trees 85%, Random Forest 81%, XGBoost 81%.
Why we validated with ML

K-means clustering will always produce clusters — even from random data. The critical question is whether those clusters are meaningful or coincidental. Machine learning validation is the standard scientific method for answering this question.

The test: We trained classifiers on part of the plant data, then asked them to predict the cluster of plants they had never seen. If the clusters were random, accuracy would be near 33% (random chance with 3 groups). At 85%, the clusters are clearly real and learnable patterns in the data.

All three models — from different algorithmic families — agreed. Random Forest and XGBoost both reached 81%. Extra Trees reached 85%. This cross-model agreement is the strongest possible evidence that our plant performance tiers are genuine.

What 85% means for customers

When you browse our Plant Finder or see a “Cluster 3 — Top Performer” badge on a seed packet, that classification is not our opinion. It is the output of a validated machine learning model that has been tested and proven to predict plant performance correctly 85% of the time.

A random guess between 3 clusters would be correct only 33% of the time. Our model is correct 85% of the time. That is the difference between a guess and a science-backed recommendation.

In plain English: If our model says a plant is a top air purifier, it is right 85 times out of 100. You can trust the cluster badge.

85%
Correct predictions
Our best model correctly identified a plant’s performance cluster 85 out of every 100 times on data it had never seen before.
3 of 3
Models agreed
All three classifiers from different algorithmic families confirmed the same result — ruling out coincidence or model-specific bias.
vs 33%
Random baseline
A random guess between 3 clusters would be right only 33% of the time. Our model is 2.5× more accurate than random chance.
How this research powers Springfield’s tools
Every interactive feature on this page is built on the validated ML model and research data — not opinion or marketing.
Plant Finder quiz
Recommendations are based on cluster data validated at 85% accuracy — not guesswork.
Air Quality Calculator
CO2 and PM2.5 reductions modelled from 3 months of real sensor data — not estimates.
Plant Database
All 28 species ranked by their scientifically verified cluster — searchable and filterable.
Cluster badges on products
Every “Cluster 3 — Top Performer” badge is backed by ML validation, not marketing copy.
That’s the science — now put it to work
Ready to choose plants that are
proven to improve your air quality?

Use our free Air Quality Calculator to see exactly how much improvement your space can expect — or browse all 28 scientifically ranked species in our Plant Database below.

Next: Calculate exactly how many plants your room needs — powered by our real research data.
Preview — Section 8 Air Quality Calculator
Interactive tool

Calculate your room’s
air quality improvement

Enter your room details and get a personalised estimate of how many plants you need — and exactly how much CO2 and PM2.5 will improve. Powered by our real 3-month research data.

Step 1 — tell us about your space Research-backed
Room size 200 sq ft
Ceiling height 9 ft
Space type
Ventilation
Your personalised air quality report Based on 3-month research data
Plants recommended
—
Cluster 3 species
CO2 reduction
—
ppm improvement
PM2.5 reduction
—
µg/m³ improvement
Room volume
—
cubic feet
Estimated PM2.5 level after plants
▲ —
0 — Excellent 35 — Safe limit 100+ — Harmful
Before vs after green wall
CO2 level (ppm) 490  →  —
PM2.5 level (µg/m³) 75  →  —
With green wall plants
Without plants
Recommended Cluster 3 plants for your space
Ready to improve your air quality?
Get these research-backed plants from Springfield — delivered across India.
Results calculated using data from Springfield’s 3-month green wall study (Nov 2018–Feb 2019).
Baseline CO2 ~490 ppm (without GW), ~420 ppm (with GW). PM2.5 baseline ~75 µg/m³, reduced to ~68 µg/m³.
K-means clustering of 28 plant species · Extra Trees classifier · 85% accuracy · Springfield by Naik Group
Next: Explore all 28 scientifically ranked plant species — filter, search and find your perfect plant.

Select the fields to be shown. Others will be hidden. Drag and drop to rearrange the order.
  • Image
  • SKU
  • Rating
  • Price
  • Stock
  • Availability
  • Add to cart
  • Description
  • Content
  • Weight
  • Dimensions
  • Additional information
Click outside to hide the comparison bar
Compare