Title: Weather data issues for the weather derivatives market Jeff Hamlin Risk Management Solutions October
1Weather data issues for the weather derivatives
marketJeff HamlinRisk Management
SolutionsOctober 5th, 2006jeffh_at_rms.com
2Weather Data Topics
- Historical data
- Raw
- Synop vs Climate
- Cleaned
- Enhanced
- Dealing with trends in temperature data
3RAW Historical Weather Data
- Two types of raw historical weather data for any
country - SYNOP
- data that are collected in real-time at various
stations around the globe and provided through
the Global Telecommunication System (GTS) - Distributed freely around the world for use as
inputs to global forecasting models - Climate
- data that are quality controlled by the
respective national meteorological services (NMS)
where the data is collected. The 'climate data'
are the 'official' station data of the country. - WRMA has previously recommended that Climate data
is the most appropriate data to be used in
weather derivatives transactions
4SYNOP data details
- Benefits
- Available for a huge number of locations around
the world (over 5,500 stations in 183 countries) - Inexpensive and readily available
- Problems
- Data quality issues missing and erroneous
values - Some non-standard recording conventions (e.g. 12
hour periods for measuring Tmax or Tmin) - Short historical records relative to Climate data
5Climate data details
- Benefits
- Data is Quality Controlled and certified by
personnel of the relevant National Meteorological
Office - Long historical records
- Consistent recording conventions
- Problems
- Must contract with relevant Met Office to obtain
(or licensed distributor) - Data can be very expensive
- Final Edited data can be released infrequently
true-ups
6Cleaned Historical Weather Data
- Problems with raw historical weather data
- Missing values
- Erroneous values
- Missing and erroneous values generally not a
significant problem going forward, but can be a
problem historically - Solutions exist for timely, quality controlled
weather data (EarthSat CME settlement data)
7Enhanced Historical Weather Data
- Most weather stations have existed for 50 years
- Over time, station location, instrumentation, or
environment may have changed - Any change can lead to a discontinuity in
temperature record for the station - Discontinuities must be accounted for and can
greatly affect valuation calculations
8Enhanced Weather Data Example
- Station X was moved from location A to location B
on May 1, 2006. - Location A is 3 meters from a taxiway and within
15 meters of some airport buildings. Location B
is 100 meters beyond the end of a runway in the
middle of a grassy field and is located in a
shallow depression. - Weather stations were maintained at location A
and location B for the past 5 months and studies
show that Tavg as recorded at location B is
consistently 0.8 degrees cooler than Tavg as
recorded at station A. - Why should we care?
9Enhanced Weather Data Example (cont.)
- Assume we are pricing a Nov-Mar HDD contract for
a cold location. - Nov-Mar risk period contains 151 days of risk
- 151 days x 0.8 HDDs/day 120.8 HDDs
- Calculation of mean and option values based on
historical data will be erroneous, given current
conditions - Need to cool all historical temperature values by
0.8 C to make historical temperatures relevant
10Trends in Weather Data
- Significant trends exist in the historical
temperature data for many locations (global
warming/urban heat island effect) - Trends must be removed from historical data
before using historical data as input to any
model. - Many different methodologies exist for removing
trends from weather data
11Trend Example (Barcelona Nov-Mar HDDs)
What is the expected value (fair strike) for a
2006-07 swap?
12Trend Example (Barcelona Nov-Mar HDDs)
Historical average 1100 HDDs
13Trend Example (cont.)
Many different formulas that can be applied to
remove trends from data (loess linear)
14Trend Example (cont.)
15Trend Example (cont.)
Historical average 1100 HDDs Detrended average
955 HDDs
16Summary
- Climate data still the standard
- Recent advances may make SYNOP data viable for
trading - Data must be cleaned and Enhanced before being
used in pricing - For temperature data, trends must be addressed.
May be less relevant for other weather variables