Forecasting Surface Temperatures to help UNEP Implement Special Sustainability Programs

Project Details




Aneesh Chandran, Vaibhav Mathur, Pradeep Kumar Grandhi, Nishikant Mishra, Divya Dewan, Mahesh Panse





Increasing global land temperature has become a major cause of concern for countries around the world.
Remarkably, this is the third consecutive year a new global annual temperature record has been set. The
average global temperature across land and ocean surface areas for 2016 was 0.94°C (1.69°F) above the
20th century average of 13.9°C (57.0°F) (1). Increasing global temperature is an indicator of global climate
change. This phenomenon affects energy consumption, precipitation cycles and crop production apart
from a rise in sea levels which can displace people living near the coastal areas. Despite alarming
evidences and consistent warnings from scientific communities to curb the impact of rising temperatures,
the governmental organizations have not responded at the desired pace.
United Nations Environmental Programme (UNEP), a UN body which coordinates its environmental
activities by assisting countries in implementing environmentally sound policies and practices, wants to
introduce its Special Sustainability Programs for countries which are at most risk due to the rising
surface temperature.

Problem Description:
Business Problem: Successful implementation of the special sustainability program is an important
factor in help UNEP reduce the impact of increasing global temperatures. To help implement these
measures, UNEP wants to study their impact and use the results to convince member countries to
implement them in the future.
Forecasting Problem: UNEP has approached us to create a forecasting model to predict land
temperatures for the selected countries. The goal is to:
- Forecast yearly average temperature for next 5 years (2013-2017) to sensitize the countries about the
risk of climate change and convince them to deploy the existing sustainability measures in a more
aggressive manner
- Forecast monthly average temperatures for next 24 months (Jan’13-Dec’14) as benchmarks to study
the effect of new sustainability programs (using test and control groups)
Success criteria:
Actual temperatures recorded in the future should be as close to the forecasts as possible. Any positive
deviation would signify the country is enforcing the sustainability measures properly.

Data Description:
UNEP has provided us with data for monthly average temperatures from Jan' 1750 to Dec’ 2012 for 10
countries. These countries were directly selected by UNEP based on several factors such as
industrialization in the country, developing/developed status of the country, CO2 emissions etc. Here we
have used the data for a single country, China, to forecast its land temperatures. In addition, we analyzed
the data for the remaining 9 countries including India, US, UK, South Korea, France, Germany, Saudi
Arabia, Australia and Canada.

Key Characteristics: Through initial EDA, we find that the data has the following characteristics:
1. Trend: The yearly data (average over 12 months) follows a 3rd order polynomial trend
2. Seasonality: The monthly average land temperature data follows a monthly seasonality with a
slightly increasing upwards linear trend.

Application Area: