Name: Mr. Bhogendra Mishra ID: 109716 Title: Analysis and ...

foulchilianΤεχνίτη Νοημοσύνη και Ρομποτική

20 Οκτ 2013 (πριν από 4 χρόνια και 6 μήνες)

133 εμφανίσεις


Mr. Bhogendra Mishra




Analysis and forecasting of snow cover using ANN in Kaligandaki basin, Nepal




As Himalayas has been considered as one of the most vulnerable region in the world from the
climate change
point of view, it is very important to study about the trends of climate variability
and snow cover area in higher mountainous regions. On the other hand, timely and accurate data
are always a big problem in this kind of study. Hence, this study aims prima
rily to access the
usability of various remote sensing data, and identifying the bias and calibration based on the
observed dataset. Secondarily, trend analysis using nonparametric techniques was carried out.
Finally, snow cover area has been forecasted us
ing an artificial neural network based on

A2 scenario.

It is shown that remote sensing technology can detect the spatial
temporal pattern of temperature
and snow cover in an inaccessible terrain of Himalayas but the precipitation data obtained fro
remote sensing do not seem reliable. Similarly, the MODIS products have already been ensuring
in different region of the world in different ways, but the accuracy of MODIS snow products
have still not ensured in the Himalayas. Therefore, MOD10A1 snow pro
duct was also related
with ASTER snow cover area in this study by assuming that the ASTER has 100% accuracy.
And it was found that MOD10A1 has approximately 81% accuracy with respect to ASTER snow

methods (ie. Mann
Kendall and Sen’s)
were used to identify the trend. Increasing
trends of temperature, approximately 0.03

was obtained from the test whereas the mixed
seasonal trend of precipitation was obtained. In general, we can conclude that, there is the
increasing trend in summer

and spring and decreasing trend in winter. Therefore, the flash floods
and winter droughts are also increasing in the region. While talking about snow cover area, a
significant negative winter and spring snow cover trend was identified. Consequently the g
retreatment was also noticed significantly. From the analysis, it is concluded that there is clear
indications that the regional warming is affecting the precipitation and snow cover area in the

Similarly in another study, Artificial Neu
ral Network (ANN) models were developed to predict
monthly snow cover area in higher Himalayas based on the selected GCM as input. Two types of
models were used to forecast, one was time series and another war normal. The accuracy of both
models obviously
depends on the accuracy of input climatic variables. But, the accuracy
gradually goes down on increasing the lead time in NARX time series network. Whereas the
accuracy remains same for GRNN till the indefinite time, but it solely depends with the accuracy

of input variables. While comparing the performance of two networks NARX gives the better
result for the short lead time where as GRNN, the second network, seems better for the longer
lead time.


Climate change, Artificial neural network, Himala
yas, Trend, Snow cover area