From satellite images to agricultural systems’ maps:

(1)

From satellite images to agricultural systems’ maps:

A remote sensing multi-level object-based approach

ISRSE 37 – May 10

^th

, 2017

CIRAD - TETIS Research unit

Beatriz BELLON, Agnès BEGUE, Danny LO SEEN, Valentine LEBOURGEOIS, Raffaele GAETANO, Louise LEROUX, Margareth SIMOES

(2)

INTRODUCTION INTRODUCTION

Why characterizing agricultural systems?

– To assess the productivity of the system (intensified vs low- input agriculture…);

– To assess environmental risks (over-exploitation of ground water resources, water quality degradation, …).

To understand the overall sustainability

(3)

INTRODUCTION INTRODUCTION

Describing and locating

agricultural systems in space

meets a public-decision-making need

(4)

INTRODUCTION INTRODUCTION

How to describe and locate an agricultural system in space?

Agroforestry (TZ)

Rainfed cereals (SN)

Highland Rice (MG)

Oasis (TN)

Irrigated crop

(5)

TOSCA AGRIZONE

PROGRAM in support of GEOABC PROJECT in support of GEOGLAM

INTRODUCTION

(6)

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

INTRODUCTION

(7)

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

INTRODUCTION INTRODUCTION

Develop methodologies to improve the monitoring

of agricultural systems at a large-scale

(8)

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

GLOBAL CHALLENGE

INTRODUCTION INTRODUCTION

Develop methodologies to improve the monitoring

of agricultural systems at a large-scale

(9)

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

Develop methodologies to improve the monitoring of agricultural systems at a large-scale

GLOBAL CHALLENGE

Increase production in a sustainable way

INTRODUCTION

(10)

Explore the potential of RS techniques

INTRODUCTION INTRODUCTION

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

Develop methodologies to improve the monitoring

of agricultural systems at a large-scale

(11)

Explore the potential of RS techniques Localization and characterization

INTRODUCTION INTRODUCTION

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

Develop methodologies to improve the monitoring

of agricultural systems at a large-scale

(12)

Explore the potential of RS techniques Localization and characterization

INTRODUCTION INTRODUCTION

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE

Production

Agricultural land expansion

Develop methodologies to improve the monitoring

of agricultural systems at a large-scale

(13)

AGRICULTURAL LAND-USE SYSTEMS’ MAPPING

Explore the potential of RS techniques Localization and characterization

INTRODUCTION INTRODUCTION

TOSCA AGRIZONE

COMMON SCIENTIFIC OBJECTIVE COMMON SCIENTIFIC OBJECTIVE

Develop methodologies to improve the monitoring

of agricultural systems at a large-scale

(14)

INTRODUCTION INTRODUCTION

Land-use systems’ mapping involves:

(15)

INTRODUCTION INTRODUCTION

Land-use systems’ mapping involves:

The delineation of relatively homogeneous areas of land, referred to as land units, that are directly linked to a specific type of land use

(Driessen & Konijn, 1992; FAO, 1993)

(16)

INTRODUCTION INTRODUCTION

Land-use systems’ mapping involves:

The delineation of relatively homogeneous areas of land, referred to as land units, that are directly linked to a specific type of land use

(Driessen & Konijn, 1992; FAO, 1993)

(17)

OBJECTIVE

(18)

OBJECTIVE OBJECTIVE

Develop a multi-level approach based on

GEOBIA and vegetation index time series analysis

for large-scale mapping of agricultural land-use systems

(19)

STUDY SITE STUDY SITE

TOCANTINS, Brazil

(20)

STUDY SITE STUDY SITE

TOCANTINS, Brazil

(21)

STUDY SITE

(22)

STUDY SITE STUDY SITE

Area : 277,621 km²

(23)

STUDY SITE STUDY SITE

Field size : mostly large (~ 100 ha.) Area : 277,621 km²

(24)

STUDY SITE STUDY SITE

Field size : mostly large (~ 100 ha.) Area : 277,621 km²

Main cropping systems :

(25)

STUDY SITE STUDY SITE

Field size : mostly large (~ 100 ha.)

Soybean/Cereal double-crop Rice/Soybean double-crop Soybean monoculture Sugarcane monoculture Area : 277,621 km²

(26)

STUDY SITE STUDY SITE

Soybean/Cereal double-crop Rice/Soybean double-crop Soybean monoculture Sugarcane monoculture Area : 277,621 km²

Main agricultural practices :

(27)

STUDY SITE STUDY SITE

Soybean/Cereal double-crop Rice/Soybean double-crop Soybean monoculture Sugarcane monoculture

Mechanical seeding, fertilization, pesticide application and harvest Dominance of zero-tillage

systems

Area : 277,621 km²

Main agricultural practices :

(28)

MULTI-LEVEL APPROACH

(29)

MULTI-LEVEL APPROACH MULTI-LEVEL APPROACH

REGIONAL LEVEL

(30)

MULTI-LEVEL APPROACH MULTI-LEVEL APPROACH

REGIONAL LEVEL

Delimit homogeneous land units

in terms of phenological patterns

(31)

MULTI-LEVEL APPROACH MULTI-LEVEL APPROACH

REGIONAL LEVEL

Delimit homogeneous land units

in terms of phenological patterns

(32)

MULTI-LEVEL APPROACH MULTI-LEVEL APPROACH

REGIONAL LEVEL

Delimit homogeneous land units

in terms of phenological patterns

(33)

MULTI-LEVEL APPROACH MULTI-LEVEL APPROACH

REGIONAL LEVEL

Delimit homogeneous land units in terms of phenological patterns

Annual cropland +

Cropping systems’ classification Identification of agricultural land-use systems

through spatial analysis

(34)

REGIONAL LEVEL

Delimit homogeneous land units

in terms of phenological patterns

(35)

METHODS > Regional level Land units delineation METHODS > Regional level Land units delineation

DATA

(36)

METHODS > Regional level Land units delineation METHODS > Regional level Land units delineation

DATA

MODIS NDVI 16-days

composites annual time series

Oct 2014 – Sep 2015 23 composite images 250m spatial resolution

(37)

METHODS > Regional level Land units delineation METHODS > Regional level Land units delineation

DATA PROCESSING

(38)

METHODS > Regional level Land units delineation METHODS > Regional level Land units delineation

DATA PROCESSING

Principal Component Analysis (PCA) Radiometric features = PC2 – PC20

PC3 PC2 PC4

(39)

METHODS > Regional level Land units delineation METHODS > Regional level Land units delineation

DATA PROCESSING RESULT

PC3 PC2 PC4

(40)

METHODS > Regional level Land units delineation METHODS > Regional level Land units delineation

DATA PROCESSING RESULT

PC3 PC2 PC4

Multiresolution segmentation eCognition Developer 9.0

(41)

Annual cropland +

Cropping systems’ classification

(42)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA

(43)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA

MODIS NDVI

annual time series

(44)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA

MODIS NDVI

Landsat 8 mosaic 30m spatial res.

(45)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA PROCESSING

MODIS NDVI

OBIA + Unsupervised Classification

(46)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA PROCESSING

MODIS NDVI

1

HSR SEGMENTATION (187741 objects)

(47)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA PROCESSING

MODIS NDVI

1

2

MEDIAN TEMPORAL NDVI PROFILE

PER OBJECT (23 composite

images)

(48)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA PROCESSING

MODIS NDVI

1

2

images)

0,2 0,4 0,6 0,8 1

1 3 5 7 9 11 13 15 17 19 21 23

NDVI

CDOY

(49)

METHODS > Field Level Classification METHODS > Field Level Classification

DATA PROCESSING

MODIS NDVI

1

2

images)

3

UNSUPERVISED

0,2 0,4 0,6 0,8 1

1 3 5 7 9 11 13 15 17 19 21 23

NDVI

CDOY

(50)

METHODS > Field Level classification METHODS > Field Level classification

3

UNSUPERVISED CLASSIFICATION 120 land units

(51)

METHODS > Field Level classification METHODS > Field Level classification

3

(52)

METHODS > Field Level classification METHODS > Field Level classification

3

(53)

METHODS > Field Level classification METHODS > Field Level classification

K-means clustering (10 classes per land unit)

3

(54)

METHODS > Field Level classification METHODS > Field Level classification

Mean temporal NDVI profile per class

3

(55)

METHODS > Field Level classification METHODS > Field Level classification

Mean temporal NDVI profile per class

K-means Clustering 1200 mean temporal NDVI

profiles (10 classes)

3

(56)

METHODS > Field Level classification METHODS > Field Level classification

RESULTS

NDVI temporal profile analysis of final classes Mean Mean + SD Mean - SD

(57)

METHODS > Field Level classification METHODS > Field Level classification

RESULTS

NDVI temporal profile analysis of final classes Mean Mean + SD Mean - SD

(58)

METHODS > Field Level classification

(59)

METHODS > Field Level classification METHODS > Field Level classification

Soybean single cropping system

(60)

METHODS > Field Level classification METHODS > Field Level classification

FEB

(61)

METHODS > Field Level classification METHODS > Field Level classification

Soybean-cereal double cropping system

FEB

(62)

METHODS > Field Level classification METHODS > Field Level classification

FEB

JAN MAY

(63)

METHODS > Field Level classification METHODS > Field Level classification

FEB

JAN

Soybean harvest MAR

MAY

(64)

METHODS > Field Level classification METHODS > Field Level classification

FEB

JAN

MAY

(65)

METHODS > Field Level classification METHODS > Field Level classification

Rice-Soybean double cropping system

FEB

JAN

MAY

JAN JUL

(66)

METHODS > Field Level classification METHODS > Field Level classification

RESULTS

Annual cropland classification Annual cropland

Other LCC

(67)

METHODS > Field Level classification METHODS > Field Level classification

RESULTS

Annual cropland classification

Annual cropland Other LCC

Overall Accuracy = 95.8%

Kappa Index = 0.88

(68)

METHODS > Field Level classification METHODS > Field Level classification

Cropping systems classification

RESULTS

Soybean - Cereal double cropping system Rice - Soybean double cropping system Soybean single cropping system

Other LUS

(69)

METHODS > Field Level classification METHODS > Field Level classification

RESULTS

Kappa Index = 0.86 Overall Accuracy = 95.8%

Soybean - Cereal double cropping system Rice - Soybean double cropping system Soybean single cropping system

Other LUS

(70)

Identification of agricultural land use systems through spatial analysis

REGIONAL LEVEL

(71)

Identification of Agricultural LUS Identification of Agricultural LUS

FIELD LEVEL

REGIONAL LEVEL

(72)

Identification of Agricultural LUS Identification of Agricultural LUS

FIELD LEVEL

REGIONAL LEVEL

(73)

FINAL RESULTS

(74)

FINAL RESULTS FINAL RESULTS

Unsupervised evaluation with agricultural statistics (Municípios scale)

(75)

FINAL RESULTS

(76)

ONGOING WORK ONGOING WORK

Reproducibility tests:

(77)

ONGOING WORK ONGOING WORK

Reproducibility tests:

2016 Tocantins

(78)

ONGOING WORK ONGOING WORK

Reproducibility tests:

2016 Tocantins

Burkina Faso

(79)

Tocantins Burkina Faso

ONGOING WORK ONGOING WORK

Reproducibility tests:

2016 Tocantins

Burkina Faso

(80)

PUBLICATION (Under review)

(81)

THANK YOU

ACKNOWLEDGEMENTS

(82)

EXTRAS

(83)

PUBLICATION (Under review) PUBLICATION (Under review)

MODIS NDVI 16-day composites time series (October 2013 to October 2014)

PCA Transformation

Phenological variables

Segmentation

EXTRACTION OF LAND UNITS

(84)

PUBLICATION (Under review) PUBLICATION (Under review)

Rule-based classification

CLASSIFICATION OF LAND UNITS

Land cover map (field level)

(85)

PUBLICATION (Under review) PUBLICATION (Under review)

Rule-based classification Phenological pattern analysis

CLASSIFICATION OF LAND UNITS

Land cover map (field level)

(86)

DB LULC – October 2015

IN SITU DATA COLLECTION IN SITU DATA COLLECTION

LULC Class No. of points TO

Annual cropland 193

Soybean single cropping 38

Soybean – Cereal double cropping 133 Rice- Soybean double cropping 22

Other LCC 653

Grassland and meadows 242

Fallows 28

Perennial crops 67

Shrubland 128

Forest 65

Build-up Surface 30

Bare soil 12

Water bodies 15

TOTAL 900

(87)

CONFUSION MATRIX CONFUSION MATRIX

Reference

TOTAL Producer’s accuracy (%)

User’s accuracy (%) Annual

cropland Other LCC

Classification

Annual

cropland 181 26 207 93.78 87.44

Other LCC 12 681 693 96.32 98.27

TOTAL 193 707 900 Global accuracy = 95.78%

Kappa index = 0.88

(88)

CONFUSION MATRIX CONFUSION MATRIX

Reference

TOTAL Producer’s accuracy (%)

User’s accuracy Soybean (%)

single

Soybean - Cereal

Rice - Soybean

Other LUS

Classification

Soybean

single 35 ⁸ ⁰ ¹⁴ ⁵⁷ ^92.11 ^61.40

Soybean -

Cereal 0 116 ⁰ ¹² ¹²⁸ ^87.22 ^90.63

Rice -

Soybean 0 0 22 ⁰ ²² ¹⁰⁰ ¹⁰⁰

(89)

SEGMENTATION PARAMETERS SEGMENTATION PARAMETERS

Data Spatial

resolution

Bands (all same weight)

Scale

parameter Color Shape

Land units

Principal components from

NDVI TS

250m PC2 - PC20 1000 1 0

Fields

Landsat 8 mosaic July 2015 (19 Landsat

scenes)

30m

B (b2) G (b3) R (b4) NIR (b5) SWIR1 (b6)

110 0,2

0,8

Compactness= 1 Smoothness = 0