Transcript Folie 1 - ThierryBenoit

EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• ESA request D2.3 (19 Oct) to clarify discrepancies
between EOFI results and IFAD statistics
– Explain discrepancy in the absolute area covered by irrigated
crops (rice & irrigated crops classes) compared to IFAD figures
– Specifically provide reasoning for the different level of
observed irrigated crops area in 1996
• Response by GeoVille on 29 Oct
• Today’s telco important to...
– Improve mutual understanding in interpreting IFAD’s area statistics
– Put benefits & limitations of satellite based monitoring in context
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• Benefits of satellite-based monitoring
– Synoptic views over large
areas
1996
2000
2007
2010
– Time series with flexible
time intervals
– Uniform scale and content
– Guaranteed quality levels
– Low cost compared to
terrestrial measurements
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• Can we compare the area figures?
EOFI
PHBM-I area
IFAD
140
140
140
EOFI
EOFI
PHBM-I area
EOFI
IFAD
IFAD
EOFI
120
EOFI
120
What are possible
EOFIreasons for the
PHBM-I area
“delta”?
EOFI
IFAD
PHBM-I area
IFAD
120
Irrigated crops
Rice
80
80
80
60
km²
100
km²
100
km²
100
60
40
2007
40
1996
20
2007
Rice
 Definition of features
20
0
1996
1996
2007
2007
1,061 ha  5,230 ha
+ 4.169 ha
Rice
 Reference area
40
1996
2007
0
1996
Rice
60
20
0
Irrigated crops
 Methods of data collection
(terrestrial sampling vs. blanket
satelliteIrrigated
coverage)
crops
Irrigated crops
199619961996
2007
2007
8,943 ha  11,612 ha
+ 2.669 ha
2007
1996
Accuracy
/ error1996margins of land
cover maps
1996
 Others?
Sources:
http://www.capfida.mg/site/spip.php?article143
http://www.phbm.mg/appuiauxinitiatives/article_irrigation_sylvie.htm
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• #1: Are the figures representative for the same area?
– EOFI covers entire PHBM project area (8,300 km²) for 1996-2010
– IFAD coverage depends on project phase – is this correct?
•
•
Phase 1 (1996-2001): 3,200 km² (4 northern communes of Tsivory, Elonty, Mahaly and Marotsiraka)
Phase 2 (2001-2009): 8,300 km² (11 communes)
1996
140
2000
EOFI
120
2007
EOFI
100
2010
km²
80
Sources:
http://www.capfida.mg/site/spip.php?article143
http://phbm.mg/sommaire/cartes.htm
 Reduction to 4
communes of
PHBM-I leads to
a decrease from
8,943 ha to
5,290 ha
irrigated land
60
40
20
0
1996
2007
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• #2: How shall the IFAD area figures be interpreted?
– Was IFAD’s 4,169 ha increase achieved by establishing new irrigated
land or by rehabilitating existing irrigation schemes or both?
– Where does the 1996 value (1,061 ha) originate? Is our assumption
correct, that this value only represents actively managed irrigated land?
– An UNOPS mission report of PHBM-I states 3,200 ha of irrigated rice
fields for 1996 – how does this relate? See below IFAD weblink
– Has the area of 5,230 ha irrigated land for 2007 been measured with
terrestrial methods?
 Terrestrial field
measurements provide
more accurate area
figures, but are more
expensive and timeconsuming
Sources:
http://www.capfida.mg/library/docpublic/Mesure_de_resultats_et_d_impacts_du_PHBM.pdf
http://www.ifad.org/french/operations/pf/mdg/i548mg/photos/montage_photo1_0401.pdf
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• #3: What exactly is measured?
– IFAD reports 32 irrigation schemes rehabilitated from 1996-2001 and
further 39 from 2001-2008 (total: 71), leading to an area of 5,230 ha
irrigated land:
• Does this figure represent the irrigated land from 71 rehabilitation projects or
the total irrigated land in the entire PHBM area?
• Is it valid for the PHBM-I or PHBM-II area?
Selected locations of rehabilitation projects
(Perimétres irrigués)
PHBM evaluation report states that only
larger irrigation schemes are adressed:
 “Petits Perimetres irrigues” (PPI)
larger 100ha
 “Microperimetres irrigues” larger 10 ha
 Is this correct?
 Satellite-based monitoring detects much
smaller cropland patches with an object
size of > 1ha – thus increasing the total
mapped area
Sources:
http://www.capfida.mg/library/docpublic/Mesu
re_de_resultats_et_d_impacts_du_PHBM.pdf
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• #4: What about rice fields under dry conditions and
irrigated cropland in remote locations?
– Main focus on the PHBM projects were irrigated rice and crops
• Do the area statistics only cover managed irrigated land or also dry rice fields
or abandoned irrigated land?
– PHBM evaluation reports state the “inaccessibility of the project zone”
and the “insufficient coverage of villages outside the main towns”
• Are existing rice fields and other irrigated cropland in remote regions
covered?
Dry / abandoned rice field
Irrigated cropland in remote location
Sources:
http://www.ifad.org/evaluation/public_html/eksyst/doc/prj/region/pf/madagascar/mg_376.htm
http://www.ifad.org/evaluation/public_html/eksyst/doc/agreement/pf/madagascar_376.htm
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• #5: Information content of available data not optimal for
detecting small-scale irrigation structures
– Poor availability of multi-temporal / multi-polarisation radar data
– Sometimes ambiguous differentiation of crop types due to imagery from
different seasonal dates (dry vs. wet season)
– Limited availability of ground-truthing data
EO imagery used for mapping
High-res imagery for validation
 Technical infrastructure for irrigation
(incl. dams, narrow water channels) not
visible in 10-30m satellite imagery
 Irrigated cropland contains actually
irrigated cropland PLUS in many cases
also potentially irrigated cropland
predominantly along water courses
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• #6: Resulting area statistics have a defined level of
reliability with known error margins
Require
ment
Achieved
accuracy
Rice
90%
80-85%
Irrigated crops
n.a.
60-70%
Rainfed crops
70%
70-75%
Forest
80%
>85%
Grassland
80%
80%
Bare area
80%
>85%
Urban
80%
>85%
Water
80%
>85%
Error margin of area statistics
Margin 15-20%
Margin 30-40%
Overall accuracy: 80-85%
Validation analysis shows a slight tendency towards overestimation of
rice and irrigated crops
- high level of completeness
- lower level of correctness
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
EOFI Service Trial 1:
UN-IFAD – Crop Acreage
• Summary of evaluation – assumptions
– Direct comparison of area figures is hardly possible
– Assumptions need to be verified in collaboration with IFAD
1996
IFAD
Satellite-based monitoring
1,061 ha
5,290 ha
(PHBM-I area)
(PHBM-I area)
• Existing managed irrigated land
• Managed + abandoned + potentially irrigated
land + dry rice fields
• All patches > 1 ha
2007
5,230 ha
11,612 ha
(PHBM-II area)
(PHBM-II area)
• Managed irrigated land from 32 + 39
rehabilitation projects > 10 ha
• Managed + abandoned + potentially irrigated
land + dry rice fields
• All patches > 1 ha
– Could the outline of the areas surveyed by IFAD be delivered?
– Shall we eliminate all patches <10 ha and re-compare?
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com
GeoVille group, Sparkassenplatz 2, A - 6020 Innsbruck, Austria, www.geoville.com