Near Real Time Remote Characterisation of Explosive Eruptions for Mitigation of Impacts and Loss in SE Asia

Benoit Taisne Assistant Professor Assistant Chair (Graduate Studies) Asian School of the Environment

23rd October, 2019 1 Identification and Alerts Decisions made and implications

Salak 2018: False alert Anak Krakatau 2018: Late alert

Anak Krakatau, Salak, Indonesia

? X

Confirmed 10:47 UTC, 10/10/18 Eruption 13:55 UTC, 22/12/18 Unconfirmed 12:39 UTC, 10/10/18 Unconfirmed 19:10 UTC, 22/12/18 No Eruption 13:24 UTC, 10/10/18 Confirmed 14:00 UTC, 23/12/18

Uncertainty to the civil aviation sector: ~3 hours 24 hours

[email protected] 2 Volcanic Impacts Complex response through time and space

Jenkins et al., 2015

[email protected] 3 Disaster Risk Reduction Decisions made and implications

Risk reduction approach Risk reduction through proactive measures ● More accurate forecasts ● Capacity building between agencies ● Development and implementation of contingency plans

Example: Eyjafjallajökull 2010 The failed management of a crisis: ● 108,000 flights cancelled; over US$ 1.7 billion in loss to the airline industry (Eurocontrol 2010) ● Total cost of US$ 5 billion (Oxford Economics 2010)

Lessons learned Need for: ● Collaboration between agencies ● Improved forecasts: dynamic framework for data assimilation

[email protected] 4 Ash Dispersal Forecast Reduce uncertainties during crises

Operational scenarios consider one realization amongst many

As eruption and crisis develop: ● How to better characterize eruption? ● How to assimilate new information?

Wealth of data: ● Multispectral, radar, infrasound, crowdsourcing

Aim: ● Reduce uncertainties ● Increase accuracy of forecasts

[email protected] 5 Social Media Analysis Where and When? Countries Where States / cities When Fine-grained locations What

{name of the volcano, cities impacted, ...} Feasibility done for Kelud 2014 ● Indonesia fastest growing user of Twitter eruption ● Preparedness: AI to be applied on past known event to refine combination of keywords

● Crisis: To be run in real time to trigger event identification

[email protected] 6 Infrasound Where, When, How big, How long?

● Infrasounds work day/night and cloudy/clear and travel thousands of kilometers

● Currently only able to detect eruptions (where, when?)

Feasibility study 1: Kelud 2014

How long? → Eruption style

Feasibility study 2: Sangeang Api 2014

How High? → Eruption intensity

(3000 km away!)

[email protected] 7 Satellite Technology What we have… and don’t have yet

Until now: Opportunistic measurements Need: Fill the data gaps for effective monitoring from inappropriate instruments of volcanic plumes Determine data needs and algorithms for reliable volcanic plumes determination and

https://earthobservatory.nasa.gov Prototype of novel dedicated instrument to: ● Retrieve 3D structure of plumes (orbits) ● Overcome limitation of cloud cover with AI on training data ● Concentration values to initialize models

Reliable characterization of One of the unique examples of a satellite detection of a evolution of ash plumes in volcanic plume, Kelud, 13/02/2014 (Himawari, Calypso) atmosphere

[email protected] 8 Framework Overview Research potential and inter -connectivity

Prior information on potential eruption Continuous monitoring for potential from independent sources eruption

Targeted Analysis in space Global search of anomalies -> eruptive scenarios for known volcano -> no a priori eruptive scenario

Artificial Technological and Intelligence and data processing Machine Learning development

Data assimilation and analysis to refine Air Traffic Management range of eruptive scenarios Civil Aviation Met Services Airlines Communicating the results with end-users Royce Rolls (engines) Insurance Re-insurance ...

[email protected] 9 Identification and Alerts Would the outcome be different now? Salak 2018: False alert Anak Krakatau 2018: Late alert Anak Krakatau, Indonesia Salak, Indonesia

[email protected] 10 Impact on Singapore How to improve the resilience of smart cities?

Pinatubo, June 1991 Ash in Singapore

Pinatubo is 2500 km from Ash ‘dusted’ Clean Singapore… while closest volcano is in Sumatra 500 km away

● Low probability / high impact events ● Mitigation through preparedness ● Reduce impacts on Changi airport, population (health), critical infrastructures (e.g. AC, MRT) Kelud, February 2014 Ash in

[email protected] 11 Near Real Time Remote Characterisation of Explosive Eruptions for Mitigation of Impacts and Loss in SE Asia Q&A Benoit Taisne Assistant Professor Assistant Chair (Graduate Studies) Asian School of the Environment

23rd October, 2019 12 [email protected] There will be ash! But where!?!

• Completion of Phase 2 of the CAAS Project • -> probability of ash in the region for a 10 year period • Below is the grid of weighted impact factor for what scenario has the largest impact on the Singapore FIR

Provided reports to MSS, CAAS, and ATS for episodes of volcanic unrest that included:

 Maps with dispersion probability for the zone containing the volcano

 For specified VEI, concentration threshold, and FL level

 For each requested months Which is now an online tool!

[email protected] There will be ash! But where!?! And from where!?!

Despite being the closest to Singapore, due to trade winds, Sumatran’s volcanoes are not the most likely to impact Singapore’s FIR.

Identification of the scenario that are the most likely to impact Singapore’s FIR. - For FL270 to FL330, due to trade winds, volcanoes in the Philippines and Sulawesi are the more likely to induce disturbance within Singapore’s FIR.

[email protected] Interactive tool Probabilistic maps for ongoing or potential eruptions

A web-based interactive tool to generate reports of probability maps (from compiled matlab code). https://ashdispersion.earthobservatory.sg (email to request log in)

Input the following parameters in the form:  A volcano zone, either by:  searching for a volcano name  selecting the zone in the list  or clicking on the zone on the map  One or more months  A flight level  One or more VEI (Volcanic Explosivity Index)  One or more ash concentration thresholds  One or more FIR to outline

[email protected] [email protected] Detection capability in plume height Plume height detectability for Sangeang Api 2014 Based on existing network and weather data, map of minimum plume height detectable could be generated with associated uncertainties

(Tailpied et al., in prep)

[email protected] From detection to characterization Providing actionable product Infrasound identified… source unknown. Use range of parameter to estimate plume height from acoustic power

Effect of the different parameters on the plume height results, considering a monopole source and Sparks equation. (Tailpied et al., in prep) r and Qmagma are the more influent parameters.

[email protected] From detection to characterization Providing actionable product Infrasound identified… source unknown. Use range of parameter to estimate plume height from acoustic power

time

Region identified, refine knowledge on conduit geometry

Effect of the different parameters on the plume height results, considering a monopole source and Sparks equation. (Tailpied et al., in prep) r and Qmagma are the more influent parameters.

[email protected] From detection to characterization Plume height for Sangeang Api 2014

Taking into account all uncertainties, from atmospheric propagation to mass loading at the source, we demonstrate for the first time the possibility to get reasonable estimate from remote station (up to 5300km). 1,3 and 15% mass loading highlighted with *. (Perttu et al., in prep)

[email protected] Where to add one regional array type? Optimization depend on the question!

(Tailpied et al., in prep)

[email protected] Where to add one regional array type? Optimization depend on the question! • Number of volcanoes

Optimizing the monitoring of all the volcanoes

(Tailpied et al., in prep)

[email protected] Where to add one regional array type? Optimization depend on the question! • Number of volcanoes

Optimizing the monitoring of all the volcanoes

• Weighted with probability of VEI 3+

Optimizing the monitoring of the volcanoes with higher VEI 3+

(Tailpied et al., in prep)

[email protected] Where to add one regional array type? Optimization depend on the question! • Number of volcanoes

Optimizing the monitoring of all the volcanoes

• Weighted with probability of VEI 3+

Need to refine the question, including Optimizing the field and communication challenges, monitoring of the volcanoes with prior to the installation. higher VEI 3+ Quid about this question: Minimum plume height to be detected?

(Tailpied et al., in prep)

[email protected] Pinatubo, June 1991 Ash in Singapore

Kelud, February 2014 Ash in Yogyakarta

Ash ‘dusted’ Clean

PrAsh

Red is high probability of ash exceeding pick concentration of 2mg/m3 (ICAO recommendation) in any 10 years span. Between 6 km and 10 km (FL200 to FL330). Taisne et al. Output of study funded by CAAS

[email protected] Regions most likely to impact FIRs… … effect of trade winds for 6 -10km elevation (FL200 to FL330)

FL270 Conc > BANGKOK HO- JAKARTA KOTA KUALA MANILA PHNOM SINGAPORE VIENTIANE YANGON UJUNG 2mg/m3 CHIMINH KINABALU LUMPUR PENH PANDANG

Yearly Zone 19 19 17 20 20 19 19 20 18 19 33 VEI 4 4 4 4 4 3 4 4 4 4 4 IF 0.1397 0.2683 0.1793 0.3078 0.2281 0.2206 0.2106 0.3098 0.0802 0.0465 0.1675 Proba (10yrs) 55 55 59 56 56 90 55 56 37 55 45

Identification of the zones that will have the most impact on the different FIRs, with eruptive scenario that are the most likely to happen. - For FL270 to FL330, 5 regions were identified that are the most likely to impact ASEAN’s FIRs. (Philippines, PNG, Timor) - Monthly analysis and impact at different flight level possible thanks to this study

[email protected] Assessing impact for Singapore’s FIR … effect of trade winds for 6 -10km elevation (FL200 to FL330)

Impact Factor in a Region: Average probability of ash exceeded given threshold within the region of interest.

Full region affected at 50% chance Same Impact Factor Half region affected at 100% chance

[email protected] Assessing impact for Singapore’s FIR … effect of trade winds for 6 -10km elevation (FL200 to FL330)

Impact Factor in a Region: Average probability of ash exceeded given threshold within the region of interest.

Full region affected at 50% chance Same Impact Factor Half region affected at 100% chance

Weighted Impact Factor in a Region: Impact Factor weighted by the probability of the eruptive scenario to happen.

[email protected] Assessing impact for Singapore’s FIR … effect of trade winds for 6 -10km elevation (FL200 to FL330)

Impact Factor in a Region: Average probability of ash exceeded given threshold within the region of interest.

Weighted Impact Factor in a Region: Impact Factor weighted by the probability of the eruptive scenario to happen.

[email protected] Assessing impact for Singapore’s FIR … effect of trade winds for 6 -10km elevation (FL200 to FL330)

Despite being the closest to Singapore, due to trade winds, Sumatran’s volcanoes are not the most likely to impact Singapore’s FIR.

Identification of the scenario that are the most likely to impact Singapore’s FIR. - For FL270 to FL330, due to trade winds, volcanoes in the Philippines and Sulawesi are the more likely to induce disturbance within Singapore’s FIR.

[email protected] Anak Krakatua December 2018 – January 2019

Hourly PMCC detection and VAAC/VONA reported height

[email protected] A CubeSat for Monitoring

A nanosatellite in a 15o or less inclined orbit in Low Earth Orbit (<500 km) provides coverage over the region with a frequency of 90 minutes of less. Six such evenly spaced satellites can provide real time monitoring capability.

In the Figure is a 6U (30x20x10 cm dimension) CubeSat which has the capacity to carry an imager with dimensions of 30x10x10 cm.

A conceptual instrument with dimensions of 30x10x10 cm. A 6U CubeSat in a 15 degree inclination orbit For this study we are considering an infrared Imaging Cryo- imager with a optics Cooler miniaturized cryo-cooler assembly. [email protected] A CubeSat for Volcanic Ash Monitoring

Conceptual imager inside a 6U CubeSat frame Miniature General instrument design for an infrared cryo-cooler imaging system used for infra- red imaging

[email protected] A CubeSat for Volcanic Ash Monitoring

• WP3 will perform a conceptual design of the ideal imager for volcanic ash monitoring to fit within the constraints of the CubeSat platform. Ima • A proto-type imager will be built ger with environmental tests and space qualification tests carried out. • A conceptual CubeSat concept will be developed with collaborators from the Satellite Research centre at NTU. Spacecraft Electronics & Batteries

[email protected] NRF-CRP 2018 Near Real Time Remote Characterisation of Explosive Eruptions for Mitigation of Impacts and Loss in SE Asia

Work Package 1 - Infrasound Work Package 2 - Mining Social Media Work Package 3 - Space-based remote sensing - Sensors and data available in the region - Potential for being the earliest source of - Open-access high resolution space-based data - Potential to record volcanic explosions 1000s information available (high traffic and usage in available in SE Asia of kilometers away SE Asia). - Capability to visualise volcanic ash in the - Quantitative link between acoustic signal and - Automated recognition of tweet keywords atmosphere eruption characteristics through AI used to pinpoint erupting volcano - Use AI to maximise the detection rate of - Dynamic eruptive scenarios - Optimisation and regionalisation of real-time eruptions from available space-based data mining of social media strands - New methods to overcome the limitations due to - Early information on eruption onset time and cloud cover location; continuous refinement for eruption - Designing a dedicated space-based explosive and deposit characterisation eruption monitoring tool

Fig: Himowari 8 animation of Kelud 2014 Fig: Time series of tweet frequency (10 min intervals) for Kelud Fig: Geophysical signature of the Kelud 2014 eruption, Java eruption, Java. © ASE, NTU. 2014 eruption, Java © ASE, NTU.

Work Package 4 - Actionable products for decision-making and mitigation - Combination of all work packages through data assimilation to refine eruptive scenario - Generation of input files and automated generation of ash-dispersion maps (with uncertainties)

Prototype development using existing approaches -> Short term/mid-term opening/closing of airspace Methodological development -> Mid/long-term influence on air quality and impact on aircraft maintenance and safety Project outcomes -> Short term: input for deposit extent and impact on the ground # Aircraft/Engine predictive maintenance # Air quality (e.g. filter management, public health) [email protected]# Safe/Cost-saving flight path planning Kelud Eruption 2014 Satellite; Infrasound; Social Media

[email protected]