Low-cost sensors and high-end instruments for …€¦ · Low-cost sensors and high-end instruments...

Michael Müller, Morten Hundt, Lukas Emmenegger, Christoph Hüglin, Béla Tuzson,

Christoph Zellweger

and many other from the

Empa, Laboratory for Air Pollution / Environmental Technology

Low-cost sensors and high-end instruments

for mapping urban NOx pollution

GAW Reactive Gases SAG Meeting, Stanley, Tasmania, 10. November 2016

Outline

GAW Reactive Gases SAG Meeting, Stanley, Tasmania, 10. November 2016 [email protected]

Low-cost sensors and high-end instruments for the measurements of NO/NO2 will be presented.

Michael Müller

Christoph Hüglin

Morten Hundt

Lukas Emmenegger

• 2x Aeroqual O3 SM50

• 3x Alphasense NO2 B42F

• Temperature

• Relative humidity

• GSM module for data transfer

NO2/O3 sensor unit – Aircube (AC)


11 June 2015 – 03 August 2016

NO2 passive samplers (two-week mean)

GoogleEarth

Setting of the case study

• Extension of an air quality monitoring network

• Replacement of passive samplers

Main interests

• Operation of a low-cost sensor network

• Achievable accuracy

• Options for performance monitoring

• Current state

• Specific devices

Findings probably transferable to other settings

Empa-SN + AQM sites of UGZ/FOEN

No elevated sites, sites not located

at a motorway

Range of NO2 within Zurich (30 minutes averages)


AC: 18 months of continuous operation

Reliable sensor unit

Wireless data transfer

Sensor raw data stored in database

• Data analysis

• Post-processing

NO2 air quality monitoring stations

NO2 Aircubes

NO2 passive samplers

30 minutes means

30 minutes means

Two-week means

NO2 sensors


AC NO2 sensors vs NO2 passive samplers (PS) PS vs CLD

Two-week means

Two-week means

AC roadside AC background

Diurnal variations

NO2 sensors

Results sufficient ???

Several issues remain to be

resolved / improved for

sufficient usability

WD, May-July

2016

WD, May-July

2016

Summary of NO2 sensor performance


REF NO2 [ppb]

AC

NO

2 [

pp

b]

REF O3 [ppb]

AC

O3 [

pp

b]

Two examples of sensor raw data

NO2 O3

Individual sensor calibration

• Calibration of individual sensors required (field calibration at AQM stations

• Use of a statistical sensor model • Description of sensor behavior in specific conditions (w.r.t. pollutant concentration, meteorology)

• Calibrated data associated with larger uncertainties in strongly deviating conditions

• In-field calibration • Conditions given by location and time period

• Correlations between observations (e.g. T, O3, NO2)

• Limitations: Infrastructure / man-power / data set might not be appropriate for sensor model


AC

NO

2 [

pp

b]

REF NO2 [ppb] REF O3 [ppb]

AC

O3 [

pp

b]

Initial calibration: 06 Feb 2015 – 18 May 2015

Check: 05 Feb 2016 – 14 Mar 2016

Slow / sudden changes in sensor behavior possible

Need for continuous sensor performance monitoring / sensor adjustment

Redundancy in sensor network

Link sensor data to data from AQM sites

Changes in sensor behavior


GoogleEarth

Group 1: City

Group 2: Background

Spatio-temporal data selection

Performance monitoring

Remote calibration

NO2 SITE – NO2 ZUE [ppb]

Performance monitoring / calibration

Lab tests (sensitivity to different pollutants, RH)

Sensor tests in the laboratory

80

% 50

% 20

%

10 ppb NO2

𝐷𝑅𝐻(𝑡) = ∆𝑆𝑅𝐻,𝑡𝑖 ∙ 𝑒𝑡𝑖−𝑡∆𝑡0

𝑡

𝑡𝑖=𝑡𝑚𝑖𝑛

Sensor tests in the laboratory


New features

• Single box

• 2 NO + 2 NO2

sensors

• Battery-powered

• Improved inlet

NO2 B42F

NO2 B43F

NO2

NO

Second generation Aircubes


Performance of NO2 sensors good but not yet sufficient Data post-processing not trivial

Challenging to obtain accuracy of passive samplers

Optimized mathematical description of the sensor

Laboratory / Factory calibration of the sensors

o Further improved sensors needed

Comprehensive sensor testing necessary, depending on application Required accuracy

Expected operating time

Environmental conditions

Continuous performance monitoring Link between low-cost sensor data and data from AQM sites feasible

Low-cost sensors are currently by no means a suitable replacement of either traditional AQ instruments or passive samplers. However, future developmens might improve capabilities of such sensors.

Summary and conclusions low cost sensors



High-end Instruments

NAREP Seminar – Morten Hundt

Motivations: Laser spectroscopy

Sensing of small molecules

(CO, CO2, N

xO

y, ...)

Environment

Air quality control

Process control

Medical diagnosis

Traffic security

Gases

Selective

Sensitive

Portable

Low power

Leak detection

Our approach:

Laser-based

Mid-IR optical sensors

Motivations: Laser spectroscopy



Motivation: Mobile NO2 monitoring

NO/NO2 are pollutants responsible for damage to human’s

respiratory systems

Monitoring of these gases in cities is difficult due to their large

spatial (on the scale of meters) and temporal variability

A small number of fixed measurement stations cannot give a complete

picture

Mobile measurements are an option to extend the data basis

Problem: CLD is bulky, slow and very shock sensitive, electrochemical

detectors lack sensitivity and need frequent recalibration

-> Laser spectroscopy!




Light source Interaction Detection Signal processing Electronics

Laser detector

Approach: Mid-infrared detection



Realization: Tram-Instrument

Compact laser spectrometer

36×40×15 cm

multipass cell

computer

detector laser data

acquisition

1599.85 1599.90 1599.95 1600.00 1600.05 1600.10 1600.1597

98

99

100

Tra

nsm

issio

n (

%)

Wavenumber (cm-1)

Multiple Peak Fit

120 ppb NO2

0.0

5.0x10-20

1.0x10-19

1.5x10-19

2.0x10-19

Lin

e Inte

nsity (

cm

/mol)

Realization: Tram-Instrument



Instrument performance

Best precision of 30 ppt

Good stability (in the laboratory)

What happens in the field?

1 10 100 1000

10-2

10-1

Zero-air

Dark-noise

Alla

n D

evia

tion (

ppb)

Integration Time (s)

0 1000 2000 3000

-3

-1

1

Data Collection Time (s)

NO

2 (

pp

b)

Instrument performance



Calibration and validation

Comparison to Nabel fixed station CLD

Accuracy better than 1 ppb

-> Ready to go!

Calibration and validation

18:00 00:00 06:00 12:00 18:00 00:00 06:00 12:000

5

10

15

20

25

30

35

NO

2 (

ppb

)

Time (HH:mm)

Nabel CLD

QCLAS

0 10 20 300

10

20

30

NO

2 Q

CL

AS

(p

pb

)

NO2 CLD (ppb)



First field measurements First field measurements



Tram data: Visualization Tram data: Visualization



Mobile application

Comparison with fixed

stations (UGZ)

Mobile application



Statistical modelling – based on a large number of

measurements and spatial information

x5,y5

x12,y12

x6,y6

x18,y18

x4,y4

x15,y15

x10,y10

Measurement

network

GoogleEarth

x7,y7

x8,y8

x9,y9

x11,y11

x14,y14

x13,y13

x16,y16

x3,y3

x2,y2

x1,y1

x17,y17

Spatial information

- Traffic intensity

- 3D city model (position,

shape, height of buildings)

- Heating systems

- Topography

- etc.

Generalized Additive Model (GAM) 𝑃 𝑥, 𝑦 = 𝑠1 𝑔𝑒𝑜1 𝑥, 𝑦 + 𝑠2 𝑔𝑒𝑜2 𝑥, 𝑦 + ⋯+ 𝜀 P: measured concentration

si: smooth non-parameterized function

geoi: explanatory variable

ε: residual

x,y: coordinates of the measurement location

Mobile measurements

Static sensors

Estimated model is used for the computation of

the air pollutant concentration at each position of

the modeling domain


Statistical modelling of NO2 in Zurich Statistical modelling of NO2 in Zurich



New Technology: Dual-color QC laser

DFB region

21 QCL periods

Optical mode overlap

23 QCL periods

Separation region





DFB region

21 QCL periods

Optical mode overlap

23 QCL periods

Separation region

Heterogeneous cascade active region design

Electrical separation region 2 Sections,

2 DFBs of different periodicity




Atmospheric measurements of NO/NO2

J. Jágerská et. al., Optics Express 23,(2015)

Atmospheric measurements of NO/NO2



Conclusion

We developed a QC laser spectrometer for measurements on a tram

in Zurich

We obtained high precision NO2 data that can be used for statistical

modelling to generate pollution maps

We are developing a compact high precision dual-laser instrument

for simultaneous measurements of NO and NO2

Next steps:

Further exploration and validation of the statistical modelling for

pollution maps

Development of multi-component gas sensors

Conclusion and Outlook


Thank you!

Low-cost sensors and high-end instruments for …€¦ · Low-cost sensors and high-end instruments...

Documents

Transcript of Low-cost sensors and high-end instruments for …€¦ · Low-cost sensors and high-end instruments...