n topo-
| Aster
ey and
e map
> trees,
; many
lor on
found
or). ]
7. CLASSIFY IMAGES USING CROP CALENDAR
SURVEY DATA
The “operation sequence” (Figure 3) is an essential component
of any crop calendar. A crop calendar is defined here as: “A
sequential summary of the dates/periods of essential operations,
including land preparation, planting, and harvesting, for a
specific land use; it may apply to a specific plot, but is
frequently generalized to characterize a specified area.” Plot
specific crop calendars form the key to map land use with the
support of (multi-temporal) RS-imagery (Figures 6 and 8).
Note that the spatial characteristics of a land use system define
its boundary. For agricultural purposes, a land use system can
be limited to a plot. A plot is defined here as "A piece of land,
considered homogeneous in terms of land resources and
assigned to one specific land use."
A cropping pattern is traditionally defined as (ASA 1976;
FAO 1996): “The yearly sequence and spatial arrangement of
crops or of crops and fallow on a given area”. In view of the
crop calendar definition, the cropping pattern definition can be
sharpened to: “The spatial and temporal arrangement of crops
(trees) on a specific plot.” Generally, a cropping pattern refers
to a period of one year, but may also contain information on
crop rotation. The definition contains spatial information
(within a plot) that is not present in a crop calendar, but lacks
actual date/period references as provided by a crop calendar.
Cropping pattern terminology is area a-specific and therefore
often used to classify land use. Legends of land use maps will
considerably improve when cropping pattern syntax is used (see
list of classifiers in the PhD thesis of the author (De Bie 2000).
Note that the development of a universal land use classification
system is not considered desirable. Instead, classifiers that
define and differentiate between existing (commonly used) land
use classes can develop into a standardized turn-key system that
is able to merge and generalize data belonging to different
classification systems. Use of “standardized” classifiers
supports standardization, but should never cause users to change
class boundaries unwillingly (practicality must come first!).
3 weeks agricultural fieldwork - Garmsar, iran
Surveyed fiolds
Cover on Jul'01
N »
Aster image, Jul'01
Figure 6. Generalization of Plot Specific Crop Calendar Data
with the Resulting Land Use Map.
IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring", Hyderabad, India,2002
A practical example regarding *mapping through surveying crop
calendars’ is provided in Figure 6. In a spreadsheet by plot
collected data are made visual and sorted on the basis of
differences in cover that is visual on the Aster image of July
2001. The problem of time lag between image date and survey
date has thus become non-relevant.
After developing 3 groups of crop calendars, using the pixels of
surveyed sites, the image was classified. The map accuracy was
determined on the basis of additional plots surveyed. Figure 7
provides the accuracy matrix on the basis of pixel counts.
Accuracy and class-differentiation can be further improved by
using multi-temporal images (Figure 8).
classification results
test set
Average Accuracy. 70.096
Average Reliability 91.696
Overall acc.: 77.2%
Figure 7. Accuracy Matrix of Map Presented in Figure 6.
& 2150000
Paddies with low cropping
Intensity (1 crop/yr)
2140000
|
Paddies with high cropping
Intensity (2-3 crops/yr)
5000
Uplands (rainfed arable
cropping and orchards)
Le
Footslopes (rainfed arable nid
cropping and orchards) 5000
Figure 8. Example of Using Multi-Temporal Images to Map
Paddies with Different Crop Intensities.
2 2500
“7500 520000 2500 5000