7. Istanbul 2004
Al of
OSA |
eat sown area
ern part of
| plain
period
2003
| by county
sown area
ern part of
plain
raction of
998 to 2000
2001 to 2003
an value)
by county
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
3 highly productive QL 25. 50 100 150 00
mm Kilometers
productive
Classification by the pattern of
temporal changes of
winter wheat sown area
for the period from 1998 to 2003
decreasing
Ë m 4 highly decreasing
E not productive
Figure 12. Classification of counties by the pattern of temporal
changes of winter wheat sown area
(Average: <15%) ^ Not productive
No |
(Trend: «-1.0) Highly decreasing
No |
(Trend: -1.0 - -0.5) >Decreasing
No
(Average: 45%<) & (Variation: <0.2) => Highly productive
No
(Otherwise) ^"Productive
ligure 13. Decision criteria of classification of counties by the
pattern of temporal changes
The author overlaid 3 factors of the estimated values mentioned
above, and he could produce a classification map, which
showed the pattern of temporal changes of winter wheat sown
area for the period from 1998 to 2003. Figure 12 depicts the
result of classification, where the criteria of each class is
described in Figure 13. Figure 12 exhibits effectively the
distribution of counties, which showed the temporal
characteristics of changes of agricultural land use. For example,
Shunyi county categorized into the class of highly decreasing
had actually showed the drastic decreasing of winter wheat
sown area since 2000, which was verified by the detail analysis
using LANDSAT-TM/ETM+ data.
Both of the methods introduced in this article could provide the
data of winter wheat sown area, as one of the major agricultural
land use, for wide area. The advantageous points of these
methods were first their simple structures in terms of the
determination of parameters of the formula, and second the
flexible applicability to other major agricultural land use.
Actually, the author attempted to estimate the sown areas of
both winter wheat and cotton in a part of Shandong Province
using the linear unmixing method and could reproduce the
pattern of distribution in general.
The significant constraint of the both methods could be their
sensitivity to the accuracy of location at the process of
overlaying multitemporal data. When we utilize the existing
dataset of NDVI, and even if we perform the additional
geometric correction to the data, the accuracy of location would
be at best around 0.1 pixel, that might induce the critical error
of estimation. In order to reduce the effect of the deviation of
location, the performance of appropriate spatial aggregation
would be a candidate of solution.
Another point of consideration is the variation of temporal
profile of NDVI over the cultivated area of the specific crop.
The methods adopted in this study neglected the influence of
variation of NDVI at the same growth stage of wheat, which
could be varied by the difference of factors such as
morphological condition of plant, damage of growth, and so on.
Therefore, it is required to integrate the studies on the
relationship between NDVI and crop growth condition for the
purpose of verifying the robustness of estimatión method.
4. CONCLUSIONS
Monitoring of agricultural land use for wide area is expectable
application of satellite remote sensing. This study attempted to
develop methods to estimate the sown area of winter wheat in
the major crop productive area of China using temporal
characteristics of NDVI by rather primitive but easily
applicable procedures to other crops. The results were not
accurate enough to discuss the detail spatial distribution,
however, were evaluated for providing the information, which
would be used to characterize the pattern of temporal changes
by setting an appropriate spatial unit.
A number of dataset of NDVI in regional scale have been
produced from NOAA-AVHRR for the past and should be
produced in the future from not only NOAA-AVHRR but also
Terra-MODIS. Therefore, the sub-pixel classification to
estimate the acreage of the specific land use type using these
dataset is considered to be a key technique to contribute to the
production of basic information on global environmental issues.
REFERENCE
Uchida, S., 2001. Sub-pixel classification of land use using
temporal profile of NDVI. J. Japan Society of Photogrammetry
and Remote Sensing, 40(1), pp.43-54.
ACKNOWLEDGEMENTS
The author thanks to Dr.Tang Huajun, Director General of
Institute of Natural Resources and Regional Planning (INRRP),
Chinese Academy of Agricultural Sciences, for his
encouragement in pursuing the collaborative research programs.
He is also grateful to Dr.Chen Youqi, Head of Department of
Remote Sensing Application of INRRP, and the staffs for their
contribution to the programs.
