Analysis and prediction of daily climate variables in central Punjab using machine learning and geospatial techniques
Loading...
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Punjab Agricultural University, Ludhiana
Abstract
The climate science community faces the significant challenge of dealing with a continuously
changing observing system. Huge datasets offer a high degree of statistical power coupled
with untapped opportunities for data miners. The advent of machine learning has made data
mining facile by providing algorithms that autonomously identify patterns with minimal
human interference. For the trend analysis, the Sen’s slope for daily maximum air
temperature, daily minimum air temperature, daily maximum soil temperature, daily
minimum soil temperature, daily morning relative humidity, daily evening relative humidity,
daily open pan evaporation and daily sunshine hours of all the months for 50 years, from 1970
to 2019 were calculated and observed highest slope magnitude on 20th March (0.093 oC/year),
18th May (0.127 oC/year), 15th April (0.167 oC/year), 2nd January (0.061 oC/year), 14th May
(0.345 percent/year), 18th January (0.789 percent/year), 2nd May (-0.104 mm/year) and 2nd
January (-0.159 hr/year) respectively. It concluded that the variability in maximum and
minimum air temperature will cause decrease in yield by 7.17% (405 Kg/ha) and 4.86% (262
Kg/ha) respectively by 2050. To calculate LST the archive satellite data of Landsat-5 TM for
the years 1994, 1996, 1998, 2008, 2010, Landsat -7 ETM+ for the year 2000, 2002 and
Landsat -8 for the year 2014, 2016, 2018 were used. The air and soil temperature can be
predicted from LST by using equation Ta = 0.775 Ts and TS = 0.884 Ts respectively (where Ta
is air temperature, TS is soil temperature and Ts is land surface temperature). A good
correlation was observed between LST and AET for maize (R² = 0.856) and wheat (R² =
0.897). The big climate data was mined using hierarchical clustering. Decadal comparison
from 1970 – 2019 was carried out between mean maximum and minimum, soil and air
temperature and pan evapotranspiration with soil and air temperature using Regression
analysis technique wherein, a linear and an exponential relationship was observed
respectively.
Description
Keywords
Citation
Bora, Sony (2020). Analysis and prediction of daily climate variables in central Punjab using machine learning and geospatial techniques (Unpublished M.Sc. thesis). Punjab Agricultural University, Ludhiana, Punjab, India.