G4INDO project Satellite Monitoring Component

Transcription

G4INDO project Satellite Monitoring Component
G4INDO project
Satellite Monitoring Component
Niels Wielaard
Dirk Hoekman
Eric van Valkengoed
G4AW all-partner Meeting
Lor Interntional Hotel Bogor
09.10.2014
Introduction
PusAir
Radar and optical satellite-based monitoring of
crops, core datasets:
- Sentinel 1A,1B = SAR – 20-30 m resolution
- Sentinel 2A, 2B = optical – 10 m resolution
@ LAPAN
@ Protata
(NDI)
Crop insurance system
separate database
(incl. information of clients)
Monitoring of current conditions (TRMM,
station data, status water levels rivers and
reservoirs)
+
Seasonal climate forecasts
(ECMWF, station-based)
Hydrological models
+
Crop growth models
DEWS
@ BMKG
Dynamic cropping calendar (KATAM)
@ Balitbangtan
Advice to farmers
Introduction
Cooperation LAPAN - SarVision - Wageningen University – TerraSphere
Background NL partners:
• Large crop monitoring projects e.g. Indonesia and Bangladesh since 1998
• Extending successful radar cooperation with LAPAN on forest and land
cover monitoring to agriculture with new European satellites (Sentinel-1/2)
Objective:
• build new semi-automated satellite image processing system together,
• building on and complementing ongoing programs of LAPAN and others,
• fully based on requirements of Indonesia (insurance system, hardware)
• operated in Indonesia at LAPAN
Enabling cost-effective insurance and support to agricultural crop statistics,
early warning and crop advice by improving dynamic crop calendar
New era in satellite monitoring
New era in agricultural crop monitoring using (radar) satellites:
• Better information : new satellites with improved information content
• More frequent information : new radar satellites that can see through
clouds and haze enabling reliable very frequent ‘time-series’
observations at high detail every 2-6 days
• Larger area coverage at higher detail : new semi-automated
processing techniques area available for analysis and efficient large
datasets
• More affordable : 10-30m data available for free, 5m at low cost
10 - 30m spatial resolution Radar : ASAR, Sentinel-1A/B
Optical: Landsat 8, Sentinel-2A/B, SPOT-4/5/6/7
1 - 5m spatial resolution Radar : TerraSAR-X/PAZ/CosmoSkymed
New era in satellite monitoring
What can be monitored using (radar) satellites:
• Which land use / crop types, crop varieties are planted where, how
many hectares?
• When and where is soil prepared, planted when is the start of season?
• What are different growth stages and biomass increase or decrease at
different times? Support estimation of expected and actual yield.
• Which areas are flooded, at what times and for how long? What risk
past 10 years? What is the timing and extent of crop damages and
losses such as caused by floods, droughts, pests and disease?
Satellite monitoring can reduce (or replace) costly and time-consuming
ground data collection. It can improve statistics.
Provides historical data to assess risk trends and performance over time!
New possibilities: radar and time-series
Big problems with Vegetation Index normal satellites: clouds, haze and smoke
Radar ‘sees’ through it
New possibilities: radar and time-series
New possibilities: radar and time-series
Now becomes possible to cover very large area, at high detail, more frequent:
-> From 6.25ha detail to 0.1ha detail, from every 16 days to every 5 days
Powerful method for semi-automated analysis of hundreds of satellite images.
New possibilities: radar and time-series
Now becomes possible to cover very large area, at high detail, more frequent:
-> From 6.25ha detail to 0.1ha detail, from every 16 days to every 5 days
Powerful method for semi-automated analysis of hundreds of satellite images.
Radar time-series monitoring
Different crop growth stages
detected by change of satellite
signal over time, example
Rice time-series monitoring at 10-30m
Courtesy European Space Agency
Colour picture made from 3 different dates of Envisat ASAR images, West Java
Temporal signatures rice growing areas: green-red, urban areas: white
Rice time-series monitoring at 10-30m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice time-series monitoring at 10-30m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice time-series monitoring at 10-30m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice time-series monitoring at 10-30m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice time-series monitoring at 10-30m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice time-series monitoring at 10-30m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice time-series monitoring at 5m
1 Nov 13
15 Nov 13
29 Nov 13
13 Dec 13 27 Dec 13
10 Jan 14
24 Jan 14
7 Feb 14
21 Feb 14
7 Mar 14 21 Mar 14
4 Apr 14
18 Apr 14 1 May 14
Rice at 5m resolution every 5-11 days
Sharp results: Multi-temporal filtering
5m can solve marginal size (0.3 hectares) and scattered
Rice time-series monitoring at 5m
Floods time-series monitoring at 10-100m
Time series of PALSAR Wide
images animation
Good visual impression of
the flooding dynamics in the
areas of focus
Black: open water
Bright white: flooded areas
covered by tree canopy)
Example from East Java
Courtesy JAXA
Example of multi-temporal ALOS radar, Nganjuk, Jombang, Kediri, East Java
Colours: e.g. different times of planting and harvesting, number of crops
Example from East Java
Courtesy JAXA
Example of multi-temporal ALOS PALSAR, Nganjuk, Jombang, Kediri, East Java
Blue colour: where is the water and when?? Information for irrigation and forecast
Toward near realtime flood monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: flooding
detected, Jan 2007
Toward near realtime flood monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Apr 2007
Toward near realtime flood monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Jun 2007
Toward near realtime flood monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Jul 2007
Toward near realtime flood monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Sep 2007
Toward near realtime flood risk monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Oct 2007
Toward near realtime flood risk monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Dec 2007
Toward near realtime flood risk monitoring
How much urban and crop
area and how many people
affected?
Use PALSAR L-band: flooding
under vegetation
Use Sentinel C-band: flooding
open areas
Every 5 days 25m resolution
PALSAR less frequent
BLUE: Flooding
detected, Dec 2007
To conclude
Cooperation LAPAN - SarVision - Wageningen University – TerraSphere
• Build new semi-automated satellite image processing system together, fully
based on requirements of Indonesia and operated in Indonesia at LAPAN
• Use new radar satellites building on and complement existing programmes
•
•
•
•
Results can help enable safe farming: access to insurance and finance
Results can help improve crop yield forecast using actual data
Results can help improve flood risk history and forecast using actual data
Results can help improve accuracy of crop statistics
Next steps November – March:
1.
Select area and discuss technical specifications needed (for insurance)
2.
Start first satellite observations and collect field photos of crop growth
Suggestions?
New era in satellite monitoring
New era in agricultural crop monitoring using (radar) satellites:
• Better information : new satellites with improved information content
• More frequent information : new radar satellites that can see through
clouds and haze enabling reliable very frequent ‘time-series’
observations at high detail (every 2-6 days)
• Larger area coverage at high detail : new semi-automated processing
techniques area available for analysis and efficient large datasets
• More affordable : 20-30m data available for free
Reliable radar observations can be made every 2-6 days:
• 1- 5m spatial resolution radar : TerraSAR-X/PAZ/CosmoSkymed
• 20-30m spatial resolution radar : ASAR, Sentinel-1
Optical: Landsat 8, Sentinel-2
New era in satellite monitoring
Type
Sensor
Resolution
Update frequency
Available
Good for
Radar
ALOS-1, ALOS-2
PALSAR Fine/Scan
20-100m
10-60m
Every 45 days
Every 14-40 days
2006-2011
As of November 2014
Land cover, biomass
wetlands, crop growth
Radar
ERS, ASAR, Sentinel1A
20-30m
Every 35 days
Every 12 days
2002-2012
As of July 2014
Cover change, floods,
crop growth
Radar
Sentinel-1A+1B
20-30m
Every 5 days
As of Dec 2015
Cover change, floods
Optical
Landsat+Sentinel-2
20-30m
Every 5 days
2015
Land cover, cover change,
floods, crop growth
Radar
TerraSAR-X+PAZ
3-5m
Every 11 (5/6 ) days
Daily
Since 2008
As of January 2015
Cover change, crop
growth
Optical
MODIS
250m
Daily
2000-now
Land cover, cover change,
floods, crop growth
Optical
RapidEye
5m
Frequent
2009-now
Land cover, cover change
Optical
Planet Labs
3m
Daily
2014-2015
Cover change, floods
Optical
UAV (drone)
5-50cm
now
Land cover, biomass,
yield etc. small areas
Rice at 5m resolution every 5-11 days
Sharp results: Multi-temporal filtering
5m detail suitable for monitoring of small farms
Rice at 5m resolution every 5-11 days
26 June 2012
12 July 2012
13 Aug 2012
21 Sep 2012
28 July 2012
30 Sep 2012
Results courtesy of: