The proposed step-wise approach will initially address
relatively more land use de-aggregation GIS-issues (a top-down
approach), and ultimately more land use data capturing GIS-
issues (a bottom-up approach”).
Topics covered are:
* Mapping and de-aggregating tabular land use statistics
(top-down).
* Options to improve land use surveys (bottom-up):
o Merging image analysis results.
o Classify images using crop calendar survey data.
o Classify images using NDVI profiles and known crop
calendars. :
o Surveying using mobile GIS techniques.
o Segmentation of images based on object-oriented
analysis.
4. MAPPING AND DE-AGGREGATING TABULAR
LAND USE STATISTICS
The presentation and use of available countrywide tabular
statistical data on cropped area and crop production can be
vastly improved when presented as crop maps and made
available as GIS layers (Web based).
This requires preparation of geo-referenced crop maps at sub-
national level that are compatible with current GIS systems. It
builds on readily available (basic) agricultural statistical data.
Products provide basic spatial information on cropped
agricultural land. They do not provide full details on land use
purposes or cover aspects of crop calendars, multiple cropping
and carried out operations (inputs, dates, etc.). Products are
immediately of use for integration in early warning crop
monitoring activities.
The activity basically builds on readily available statistics and
maps to generate through statistical inference a new GIS
product.
Input data comprise of crop statistics at a sub-national level, e.g
published agricultural census data and/or annual bulletins on
cropped areas by administrative areas for the whole country.
Annual statistics must be properly scrutinized through
evaluating time series. A 5-year period update must be aimed at.
The FAO (AGL-department) is presently compiling the required
statistics for many developing countries using 10-yearly
Agricultural Census reports and when un-available, by
compiling series of annual crop statistical publications.
Spatial GIS data comprise of RS-images, expert rules on agro-
ecology and of thematic maps.
Very useful images, which are freely available through the
Internet, are the SPOT-4 Vegetation 1-km NDVI images. At
present a 4-years decadal global dataset exists. The data are
superbly geo-referenced and allow the user full control on (de)-
selecting pixels on the basis of the provided radiometric quality
and cover type seen (land, water, ice, snow, cloud, shadow).
Time series of images can be subjected to an unsupervised
classification routine to stratify and differentiate relevant
vegetation profiles (Figure 1). The number of required classes
to prepare is evaluated through an iteration process.
Expert rules relate to evaluating Agro-Ecological Zone maps
(weather and land) regarding the suitability (possibility) of each
IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India, 2002
zone to grow the crop under scrutiny. Noted must be that
suitable areas might be cropped by more profitable crops while
areas evaluated as less suitable in the eyes of the researcher are
often deemed suitable by marginal subsistence farmers. The
shown example notably showed little explanatory power of the
shown crop suitability map (modified FAO-product) to explain
reported crop statistics by administrative area.
DistrMap Ha .
Abura/Asebu/Kwamankese 5 Maize Crop
Ae | Statistics
Adansi West
Adidome
Afigya Sekyere
Afram Plains
Agona
Ahafo- Ano North
Ahafo Ano South
Ahanta West
Ajumako/Enyan/Esunafo
Akatsi
Akwapim North
Akwapim South
Amannsie East
Amansie West
Aowin-Suaman
Asante Akim South
Asante-Akyem North
Asikuma/Odoben/Brakwa
Assin
Asunafo
Asuogyaman
Asulifi
Atebubu
Alwima
Awutu/Efutu/Senya
Bawku East
Bawku West
Maize
Suitability
: 0.10
Maize Crop Map 11 Reservas
Ghana (%-area) m e 0.00
n 250 km
Figure 1. De-Aggregating and Mapping Crop Statistics.
e The ‘Area Mask’ comprises of the location of :..:::. and
urban areas, water, and a 100 m pixel radar image, classified by
comparing it with Aster images (15 m) to identify pixels
representing trees.
,
* The NDVI image represents a classified image (30 classes) of 4 -
years, decadal, 1 km Spot Vegetation NDVI Images.
% 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%); preliminary result.
in
pr:
lat
20
ele
of
to
(de
be
op
pu
To
fol