Novel Approaches to use RS-Products for Mapping and Studying Agricultural Land Use Systems Presented are novel methods that support production of agricultural.

Novel Approaches to use RS-Products for Mapping and Studying Agricultural

Land Use Systems

Presented are novel methods that

support production of agricultural land

use information as required to provide

timely spatial information to generate

food security policies and that support

land use planning studies.

Dr. C.A.J.M. de BieITC, Enschede, The Netherlands

 Commission VII, Working Group VII/2.1 on Sustainable Agriculture

Title

In many developing countries there is a general paucity of land

use information.

At national level, many countries now seek to monitor land use

change as a basis for policy guidelines and action.

Agricultural land use surveys often rely on a “Multiple area frame”

sampling technique.

This technique is costly, laborious, and mostly based on outdated

Arial Photos (APs).

Use of new high resolution RS-images (e.g. Aster of 15m) and of

multi-temporal NDVI images (e.g. Spot of 1km) make better and

more efficient approaches feasible.

Opening Statements

Statements

Options are discussed to improve the quality and efficiency of geo-

information production with emphasis on agricultural land uses.

Attention is drawn to the dynamic aspects of land use systems, with

crop calendar information as focal point.

Emphasis is put on recognizing plots as primary sample units to

survey for collecting agricultural land use data.

Defining Benchmarks

Topics Presented

1. Elementary concepts to carry out agricultural sustainability studies,

2. De-aggregation of tabular crop statistics to 1km pixel crop maps,

3. Merging image analysis results,

4. Classifying images using NDVI profiles and known crop calendars,

5. Surveying using mobile GIS techniques, and

6. Image Segmentation based on object-oriented analysis.

Benchmarks/Topics

Bio-Physical Conditions

Socio-Economical Conditions

Land Use System

Other Land Use Systems

Livestock Systems

Context Goals

Inputs / Implements

Outputs /

Benefits

Soil / Terrain Climate / Weather Vegetation (Crops / Flora)

Wildlife (Fauna)

Infrastructure Operation Sequence

Land

Land Use Purpose(s)

Land Use

Land User(s)

Impact on land ( + or - )

Decision making / planningRequirements &

Suitability

Productivity

Impact on/from the environment

Interaction with secondary production

systems

The Concepts

The “Land Use System” (LUS) with ‘study entries’.

1.Concepts

Operation Sequences

Grazing Fallowing

1989198819751969 1979

Rainfed Cropping

J F M A M J J A S O N D

1988 1989

Observations

Operations

… many aim to control growth limiting, and yield reducing land aspects.

… many relate to growth limiting, and yield reducing land aspects.

Ploughing Harvesting Fallow

Pest AttackGermination

Trampling Hail Storm

Rill Erosion

WeedingSeeding

NPK Applic.Illustrating land use

operations

and land use obser-

vations

The “Operation Sequence” impacts on ‘sustainability’ aspects.

Land Use

Land

Land Use System

Oper.Seq.

Yie

ld

they address: growth limiting yield reducing land modifying aspects of LUSs.

Feasible

Problems

Management

Plot-to-plot variability

ProblemsProblemsProblemsProblemsProblems

What do sustainability studies do ?

They relate differences in land and management aspects to differences in system performances.

They use survey data from many plots.

we study this gap.

Sust.Studies

De-aggregation of Tabular Crop Statistics to 1km Pixel Crop Maps

The objective is to map where crops

are grown using a “mix” of existing

GIS-information and crop statistics.

2.Deaggregation

District Map

Table of number of pixels by district

Maize Crop

Statistics (5 yrs) by

district

Mask of: parks, reserves, urban, water, and trees

Masked and Classified

Masked

FAO Maize Suitability

Map(values from

0 to 100)

30 NOAA NDVI

Classes(1km pixels)

% of area to maize = 1.9 if Mod.Suit. + 2.7 if Suit. + 6.9 if Class-11 + 3.0 if Class-15 + 32.6 if Class-25 + 17.8 if Class-26 + 12.3 if Class-27 + 34.1 if Class-29 + 15.5 if Class-30

(N=110; Adj.R-Sq=74%)

Regression

Apply to masked maps

GIS flowchart

Merging Image Analysis Results

The objective is to optimize use of high

resolution satellite imagery to delineate

‘hard’ and ‘soft’ map units.

3.Merging Images

TM 453

NDVI

Classified pine trees and shade

Often specific vegetation types can be clearly

distinguished, while others can not.

‘Soft’ map units

Represents: bush, pasture,

fields, deciduous trees, etc.

‘Hard’ map units

Merged product

Distinguishing them is ‘season dependant’

TM: hard-soft

Village boundaryStreamsVillagesRoadPathsRidgesContours

Very Bare to 50% Bare

Poorly vegetatedSomewhat vegetatedWell vegetatedPine Trees

1 km grid

Results can be presented with relevant digitized lines at large scale.

and used, e.g. for local level land use planning.

TM+GIS

Classifying images using NDVI profiles and known crop calendars

The objective is to identify areas

having different crop calendars.

The relation and interpretation quality

of classified 1km NDVI time series at

country and at local-level is explored to

ascertain their link with crop calendar

information.

4 Year Data

ND

VI

4.NDVI profiles

May-Jun-Jul 2001Aug-Sep-Oct 2001Nov-Dec-Jan 2002Feb-Mar-Apr 2002

1 km res. Spot Vegetation image (RGB Feb-Mar-Apr’02 )

W-Nizamabad

NDVI-profiles of 4 pixels in Nizamabad

Apr’98 May’02 By Decade

1. General Spot NDVI profile analysis for

Nizamabad area

NDVI India

W-Nizamabad

Unsupervised-classified Spot Vegetation image(30 classes; 1998-2002; 147 decadal images)

NDVI-profiles of 8 classes found in Nizamabad

Apr’98 May’02

By Decade

W-Nizamabad

Unsupervised Classification

NDVI Classes

First the NDVI-profiles were classified

unsupervised into 30 vegetation classes

2. Detailed Spot NDVI profile analysis for

Nizamabad area

15

20

25

27

29

1,2,23

18,19,24

28,30

21,22,26

14,16,17

3,4

8,10,12

6,7,9

5,11,13

Original classes

Then the profiles were visually grouped into 14

more general classes

Gets out of the image

series “what is in them”.

The expert now classifies

“supervised” the

intermediate product.

Nizam.Classes

Apr’99 Apr’00 Apr’01 Apr’02

250

200

150

100

50

0 Apr’98

ND

VI

Rice during Rabi Forest

Dryland Crops

Water

NDVI data from decadal Spot-Vegetation Images; 1 km pixels

Clouds

14 NDVI profiles across 4 years

50

100

150

200

Jun

e

Au

g

Oct

Dec

Feb

Ap

r

Forest

Water

Rice during Rabi

Dryland Crops light soils

Clo

ud

s

Clo

ud

s

Cotton Dryland Crops heavy soils

Rice during Kharif

Final avg. NDVI-profiles of the 14 vegetation classes

The NDVI profiles

Conclusion 2: Mixed pixels (1 km) generate ‘intermediate’ NDVI-profiles.

Conclusion 1: Profiles can be used for monitoring purposes.

Nizam.Profiles

Initial Map

comparison of the 2 Maps

Mandal Boundaries (10 km grid)

Kotagiri

Birkur

Rice in Rabi

Final Nizamabad map with 14 classes

Nizamabad

3. Comparing the two Spot NDVI profile maps

Conclusion 3: Post-classification process provided ‘more refined’ results.

Compare Maps

Conclusion 5: Patterns identified agree well with a 23 m IRS image.

4. Spatial validation of NDVI map units

Conclusion 4: Patterns identified agree well with terrain features.

IRS Image (18 Jan’00)

IrrigatedRainfed

Heavy soils

RainfedLight soils

Rice in Rabi & Kharif

NDVI-profiles on a DEM

DEMs, IRS

Rice in Rabi & Kharif

IrrigatedRainfedHeavy soils

RainfedLight soils

jun jul aug sep oct nov dec jan feb mar apr may

Rice

Sugarcane ( + 1 ratoon)

Sunflower

or Groundnut

or Black Gram

or Green Gram

or Red Gram

+ Rice (dominant)

+ Wheat, Sunflower, or Groundnut

or Safflower

Cotton

Sorghum

or Sunflower

or Groundnut

or Bengal Gram

RabiKharif (monsoon)

Summer Irrig

ate

dR

ain

fed

; he

av

y s

oil

Ra

infe

d; lig

ht s

oil

5. Linking the Spot NDVI profiles to crop calendars

Conclusion 6: Crop calendar groups can easily be linked to profiles.

Conclusion 7: Having crop calendar information at plot level is a must.

Crop Calendar

Surveying using Mobile GIS Techniques

The objective is to test use of mobile

GIS equipment for detailed fieldwork.

5.Mobile GIS

Jan. 2000

IRS-Image (23m Multi-spectral fused with 6m Pan)

Mar. 2002

Mar. 2002

IRS-Image (23m)Example 1: Mapping Plots

Digitized “in the field” in

Sep 2002

Plot Polygons

Often, roads are poorly mapped on topo-sheets, while (15m resolution) images of e.g. mountainous areas hardly show roads.

Roads digitized in Ghazi on a topo-sheet and on an Aster image (Febr.2001; scale 1:25,000).

Example 2: Mapping roads in hills

Digitizing roads by GPS in hills proved very useful, and accurate enough to fill the short-comings.

Road Lines

GPS-iPaq experiences

• Not for GIS amateurs

• Requires user to know facts on projection systems ‘properly’

• Requires proper preparation:

• of geo-referencing images

• of compressing images

• of iPaq and GPS settings

Once all is done well….experience shows too many advantages and

even a dependancy of

using the equipment during

fieldwork !!!

1881m2

1038m2

2 Fields

Experiences

Image Segmentation based on Object-Oriented Analysis

The objective is to identify primary

sample units (plots) for agricultural

surveys.

6.Segmentation

Plot boundaries are seen on images but not used during classification on a pixel-by-pixel basis.

Plot boundaries are the primary sample units during agricultural surveys.

Image segmentation, before classification (using eCognition) recovers this ‘loss’.

Aster image (15m) of Garmsar, Iran.

Area frame sampling techniques can greatly benefit from Image segmentation.

AFS benefits