3. 1. Introduction
Previous Project
Last year a similar project was completed by another student
9 different methods in Arc and 3 different regression methods
This student concluded that one method from Arc and one
regression method were the best methods when modelling
temperature
This years project differs because only the two best methods were
examined and comparisons between 2012 & 2013 were made
4. 1. Introduction
Reasons for the Project
In 2011, the Community Business Development Corporation(CBDC)
commissioned the AGRG
Shelburne, Queens, Lunenburg, Yarmouth and Digby counties
The goal for the CBDCs is to find the best locations for agricultural potential
based on the temperature throughout each year of the project.
5. 1. Introduction
Weather Stations & Deployment
73 stations were deployed
throughout South-West Nova
Scotia
59 Onset Stations (records
every 10 seconds)
14 Campbell Scientific Stations
(records every 5 seconds)
1 Environment Canada Station
(Greenwood)
2 HOBO Stations
Deployed in 12 transects
Averaged monthly for 2013
6. 1. Introduction
Growing Degree Days (GDD)
Measure of heat accumulation and is useful when assessing the
suitability of a region for a crop production.
GDD is calculated using the following equation:
𝐺𝐷𝐷 =
𝑇𝑚𝑎𝑥 − 𝑇𝑚𝑖𝑛
2 − 𝑇𝑏𝑎𝑠𝑒
Maximum temperature minus minimum temperature divided by 2
minus a base temperature (i.e. 10˚C for grapes)
7. 1. Introduction
Methods Used
Two methods were used to create temperature maps:
Geostatistical Interpolation
The values between the point data are a prediction
Regression Techniques
Takes into account other variables and how they change
8. 2. Methodology
Input Data
Supplied Data
Locations of stations with
monthly temp averages
DEM layer
Coastline/Roads/Hillshade
Solar GOES data
Esri solar potential data
Validation data
Predictors
Easting
Northing
Aspect
Elevation
Distance from the coast
Solar radiation
9. 2. Methodology
Arc Method
Empirical Bayesian Kriging (EBK)
Geostatistical Interpolation
Method
Presets the most difficult aspects of
a normal kriging model
Estimates the underlying
semivariogram
Advantages
Little interactive modelling
More accurate for smaller datasets
Disadvantages
Slow processing time
Limits the ability to customize the
parameters of the semivariogram
What is a Semivariogram?
It depicts the spatial autocorrelation
of the measured points.
Range: The distance where the model
first flattens out
Sill: the values on the y-axis
Nugget: The y intercept
Partial Sill: The sill minus the nugget
10. 2. Methodology
Arc Method
Semivariogram Estimation:
1. A semivariogram is estimated from the data
2. Uses the semivariogram as a model with new data being simulated
at each location of the data
3. A new semivariogram is estimated from the simulated data
4. Steps 2 & 3 are repeated numerous times creating new
semivariograms
11. 2. Methodology
Regression Method
Generalized Additive Model
(GAM)
Purpose:
Maximize the quality of the
prediction of a dependent
variable (i.e. Temperature)
Estimates the smoothing
functions of the predictor
variables which are connected
to the dependent variable. (i.e.
solar radiation, elevation,
location, aspect and distance
from the coast)
Completed completely in the stats
package R
Rasters were created and brought
into Arc for manipulation
12. 2. Methodology
Scripting
In total 4 scripts were used to produce both EBK and GAM outputs
2 Python
2 R
These scripts could produce EBK and GAM outputs for each month,
extracted differences between the two methods, and measured the
error
These scripts were written by the student last year that completed
the project
Made alterations to fit my data
16. 3. Results
Averages
This map was produced
within ArcMap using the
Raster Calculator and map
algebra.
The map shows the areas
with the highest heat
accumulation
Wolfville area
interior of SWNS near
Bridgewater.
17. 3. Results
Averages
Created using Raster
Calculator and map algebra
This map shows a similar
pattern to the locations of
the GDDs. The warmer
temperatures are found near
Wolfville and in the interior
of SWNS
The coldest temperatures
are located along the coast
and the North and South
Mountain
18. 3. Results
Validation
Validation was done using
Environment Canada data and
data from the HOBO stations
Mean Squared Error (MSE) was
calculated for each month
EBK is in red and GAM is in
green. Shows the MSE for each
station for every month
GAM is the “better” method, with
the lowest MSE
19. 4. Conclusion
Limitations
Time
Completing other class work
Trying to understand R scripts
Better validation data
One Environment Canada station is not enough
Had to incorporate other stations
20. 4. Conclusion
GAM is the superior method
Having better validation data can greatly increase in the accuracy of
the models.
There are locations that can sustain agricultural growth in SWNS
2013 was generally cooler compared to 2012
21. References
Ambrose, S. (2013). Modeling Temperature Data throughout South West Nova Scotia.
Colville, D. & Regier, W. (2012). South West Nova Scotia (SWNS) Temperature and Solar
Radiation Study, 2011 Project Summary. Retrieved February 1, 2012 from http://southshore
opportunities.com/south-west-nova-scotia-temperature-and-solar-radiation-study- 2011-project-summary-2/
Esri 2014. ArcGIS Desktop: Release 10. Redlands, CA: Environmental Systems Research Institute.
Wood, S. N. (2006). Generalized Additive Models: An Introduction with R. Boca Raton, FL:
Chapman & Hall/CRC/