Application of FengYun Meteorological Satellite in Global Wildfires - - PowerPoint PPT Presentation

Application of FengYun Meteorological Satellite in Global Wildfires Monitoring

ZHENG Wei

National Satellite Meteorological Center ( NSMC ) China Meteorological Administration ( CMA )

zhengw@cma.gov.cn

1 Developing process 2 Method and validation 3 Global application 4 New research and future plan

Outline



Wildfires in global forests, grassland and farmland are a major source

• f natural disturbance.



Satellite remote sensing systems can monitor the regional and global wildfires in near-real time, and provide the timely fire information for emergency and resource management.

Humans

火

Wildfires

impact

Active fires Burned area Fire spreading Emissions Wildfire risk …

Satellite

Developing process of wildfire monitoring

Daxinganling forest fire on 6 May, 1987 Mongolia grassland fire on 6 May, 2000

AVHRR FY-1C CHINA RUSSIA

In middle of 1980’s Data source: foreign data Steady development of FY Operational service of FY-1C In late 1999 Main data source: FY-1C

08:00-14:00

• In the wake of
• perational service of FY-2, continuous wildfire

detection based on FY-2 was used;

• Comprehensive application of geostationary satellite and polar orbit

satellite were gradually developed;

• In

early 2000, wildfires monitoring serviced mainly in China and adjacent area.

FY-1D AVHRR FY-2C FY-1D FY-2 FY-1

Nenjiang forest fires in May of 2006

Developing process of wildfire monitoring

Current wildfire monitoring capability

FY-4A FY-3B,C,D

 FY-3 and FY-4 as the second generation of Chinese meteorological satellite , the spatial, temporal and spectral resolution improved largely.  Fire monitoring capability has been enhanced greatly. More accurate and timely fire products can be generated.  Especially in global application, FY become the most important data in NSMC.  High response time  High positioning accuracy  High monitoring frequency

1 Developing process 2 Method and validation 3 Global application 4 New research and future plan

Outline

The method of wildfire detection

Automatic wildfire detection - Contextual method NSMC developed the automatic wildfire detection method with higher accuracy, considering complex earth surface, different cloud conditions, and solar radiation disturbances. Find the fire position in real time!

1) Core algorithm Mid-infrared channel is sensitive to fire temperature. The temperature difference between target pixel and background in mid-infrared and far-infrared are used. 2) Cloud contaminate Different cloud conditions （cloud, thin cloud, tiny cloud, cloud edge）. 3) Sun glint When the sun glint angle is less than 10 degrees, no fire detection. 4) Water body，desert Water body and desert can be masked by land cover data. 5) Suspected fire

The method of sub-pixel wildfire information evaluation



In daily wildfire monitoring ，tens or even hundreds of fire pixels

• ften are detected. If using pixel size as the area of active fire, it will be

much larger than the actual size.



The method to evaluate the sub-pixel size of active fire is developed, which can provide more accurate information and also be used to calculate FRP.

1) Dual channel evaluation for P,T. Using mid-infrared and far-infrared channels to evaluate sub-pixel size and temperature of fire. 2) Single channel evaluation(T is set 750K) Using single channel to evaluate sub-pixel size when the temperature(750k) of active fire is set. FRP (Fire Radiation Power) evaluation

( ) ( )

, * 1 *

N P T P N P N = + −

( ) ( )

, * 1 *

N P T P N P N = + −

) /( ) (

N N N N P − − =

* T S FRP

σ =

Burned area estimation

NDVI and near infrared channel data are utilized to discern burned

• area. Furthermore, the fusion method of FY and high spatial resolution

satellite data is developed fully using the temporal and spatial resolution.

1) Two images before and after fires 2) Single image after fires 3) Multisource satellite data

NDVI

• NDVI

C = NDVI - NDVI

Reflectance

• f

burned trees and grass strongly decreased in visible and infrared channels. Support for wildfire loss assessment and emission estimation

The spectrum features of burned area

Burned pine tree Burned birch

Expert interpretation

Bright red: active fire Dark red: burned area Green: vegetation Dark blue: water body Gray: smoke or cloud

FY-3 multiple channels composite image

The method of wildfire detection

 NSMC build fire detection team;  Expert interpretation method was developed for major fire monitoring ;  According to expert experience, the influence of cloud, water, urban heat island and other factors can be eliminated.

Night multiple channels composite image

Provide targeted decision-making service products

Validation based on man-made fire experiment

 NSMC made a man-made fire field experiment coincident with satellite overpasses in

• 2005. The man-made fire field was in circular shape , and was laid over firewood, tree

branches and trunks.  Thermal imaging system instrument was used to measure the radiance and the temperature distribution in the field.  The experiment indicated the methods of fire detection and sub-pixel size evaluation are effective and satisfied.

Validate the accuracy of fire monitoring

In recent years, NSMC have hold many experiments for validation of wildfire product accuracy.

Method validation based on field investigation and experiment of wildfire

2007 2013 2014 2015 2018 2019

 In May 2007, fire intensity evaluating investigation in Heilongjiang using the helicopter.  In September 2007, active fire area evaluating investigation in Heilongjiang using the UAV.  In May 2013, grassland burned area field spectral measurement in Inner Mongolia.  In June 2014, farmland burned area investigation in Henan Province.  In July 2015, fuel load measurement and investigation in the northeast forest area.  In May 2018, farmland straw active fire field monitoring experiment.  In July 2019, background temperature field measuring experiment in Heilongjiang.

Comparison between expert interpretation and automatic wildfire monitoring results

Typical regions were selected and compared the automatic fire points with the expert interpretation ones which are thought as the truth value. The comparing results show the accuracy of the FY-3 automatic fire monitoring algorithm is acceptable.

Validate the accuracy of FY-3 automatic detection at global scale

1 Developing process 2 Method and validation 3 Global application 4 New research and future plan

Outline

Operational flowchart of wildfire monitoring in NSMC

... ... Meteorological Satellite Data Real Time Receiving ( FY-3, FY-4, … ) Data processing And Products Generating

Internet, Fax, Hard copy ...

Daily global wildfire product of FY-3D (On 21 August,2019)

 Daily FY-3 global fire products with 1 km spatial resolution;  Product contents: fire location, sub-pixel size , intensity and FRP

Monthly global wildfire accumulation density product using FY-3D (In August,2019)

Density

Wildfire in Arctic circle monitoring

Fire distribution map of the Arctic circle monitored by FY-3 meteorological satellites (July 2018 VS July 2017)

Arctic Arctic Circle Circle

Russia Finland

 In the summer of 2018, continuous extreme high temperature weather hit the northern hemisphere, wildfires burned into the Arctic circle.  FY-3 fire distribution map showed that in July 2018, wildfires in the Arctic Circle

• f

Eurasia increased significantly compared with the same period in 2017.

FY-3D Wildfire dynamic monitoring map of California, USA

Camp Wildfire Camp Wildfire Woolsey Wildfire

9 Nov.，2018 10 Nov.，2018

 Based on the long-time FY fire information dataset, wildfire frequency maps showed that wildfires in California are widely distributed.  In the five years from 2014 to 2018, the number of fire pixels in 2016 was the largest.

FY-3D Wildfire statistic analysis of California, USA

Wildfire statistic analysis of California

2014 2015 2016 2017 2018（to 11.5）

FY-3A monitored burned area of forest fire in the northeast of China 5 May , 2009

FY-3A monitored burned area of forest fire continuously from 28 April to 5 May, 2009.

Burned area estimation based on FY data

Burned area

Wildfire evaluation by combining burned area and NDVI

China Russia Mongolia Russia Mongolia China Grassland fires 20 March , 2015 to 20 April , 2015 Burned area of 2015 overlaying the vegetation index map of 2014 based on FY-3C

2019-11-16

9 Nov.,2019

FY-3D Wildfire dynamic monitoring map of Australia

15 Nov.,2019

Wildfire product tools

http://rsapp.nsmc.org.cn/geofy/en

SWAP online

http://satellite.nsmc.org.cn/PortalSite/Default.aspx

Wildfire product tools

FENGYUN satellite data center FY-3 daily fire products can be downloaded from FY satellite data center .

1 Developing process 2 Method and validation 3 Global application 4 New research and future plan

Outline

Combining FY-3D/MERSI-II far-infrared and mid- infrared data

New method research for wildfire detection

Time sequence of FY-4

Field experiment

FY-3D/MERSI-II have 250m resolution in far-infrared channels, which can provide more accurate position and intensity information. Time sequence method detect fire based

• n

the temperature difference in the adjacent observation time of the pixel. The method can improve the sensitivity of fire detection .

Wildfire spreading estimation

 Fire spreading estimation has an important reference significance for fire fighting work .  NSMC is developing the method based on satellite, fire behavior and GIS.  Satellite information is used as initial value of factor, including position, length of fire line, the direction and speed of spread. The length of fire line is calculated by sub-pixel size of active fire.

10:15 on 29 Apr. , 2009 FY-3A 12:47 on 29 Apr. , 2009 NOAA-18

Estimate the fire spread range within 2 hours and 32

• min. from 10:15 to 12:47.

V1 =V0·Ks·Kw·Kf·Kt·Km

V0：initial speed, Ks：the adjustment of fuel type based on land cover. Kw：the adjustment of wind speed, Kf：the adjustment of landform based on DEM. Kt：different type of vegetation cover where fire spread may encountered, like non forest or grass land area, burned area, bare soil and water body. Km ：the time condition when fire spread

V0= Fint *（0.0299T+0.047W+0.009(100-h)-0.304）

Fint：fire intensity factor， Fint = e fint*0.2 ，fint fireintensity calculated from sub-pixel size of active fire T：temperature of fire field, derived from far infrared temp. in background. W：Wind speed in fire field. M： Moisture content of fuel.

Wildfire risk prediction

For wildfire prevention to avoid property losses,wildfire risk prediction method is studied based on the long-time series fire information, meteorological observation and forecast information. The prediction method applied in South China has a good results. The fire point is superimposed on the fire risk prediction area, with high consistency. The method can be extended to other regions.

Fire risk prediction and fire point map Fire risk prediction and fire point map

t：dailymaximum temperature f：daily minimum relative humidity minimum relative humidity v：dailymaximum wind speed m：dailymaximum rainfall and consecutive days without rain R：fire point statistic from FY satellite data

= ( )

M R

F U I R +

( ) ( ) ( ) ( )

M t f v m

U I t I f I v I m = + + +

Wildfire emission estimation

 The wildfire releases lots of trace gases and particulate matter, which change the composition of atmosphere and have an important impact on the biogeochemical , climate , air quality and human health .  NSMC is developing the methods to estimate wildfire emission based on FY and others data. 1）Using burned area information

BA is the burning area based on satellite identification, F is the fuel load (kg dry matter m2), CF is the combustion efficiency, EF is the emission coefficient (g kg-1).

2）Using radiant power

FRE is the radiant energy at the fire point (MJ), β is the conversion coefficient of radiant energy and fuel quantity (kg dry matter MJ-1), EF is the emission coefficient (g kg-1), FRP is the radiant power at the fire point (MW), t1 and t2 are the start and end time of biomass combustion.

Emissions BA F CF EF

= × × ×

∑

EF EF

Emissions FRE FRPdt β β

= × × = × ×

∑∫

Receiving fire information in real time

Satellite real-time fire monitoring intelligent platform

Global fire monitoring APP will be researched and developed.

Desktop terminal

Fire list (time, position, size, type, frequency, image, etc.) Multi-source satellite Auto Warning

Mobile terminal

 Improve FY-3 automatic algorithm of wild fire detection at global scale;  Develop FY-3 automatic algorithm of burned area and smoke detection;  Develop forest and grassland fire spread estimation model;  Develop wildfire weather risk prediction method;  Enhance the cooperation with international experts.

Future plan

Thanks for your attention!