Investigation of Spatial Mosquito Population Trends Using EOF Analysis: Model Vs Count Data in Pasco County Florida

1
Investigation of Spatial Mosquito Population
Trends UsingEOF Analysis Model Vs Count Data in
Pasco County Florida

• Cory Morin

2
Presentation Outline

• Outline of Objectives of Study
• Background of Research Why Study Mosquitoes?
• Introduction to DyMSiM
• Model Runs Correlation and Regression
  Coefficients
• EOF Analysis
• Conclusions and Discussion

3
Objectives

• Validate Model (DyMSiM) with Mosquito Count Data
• Using 25 Locations within Pasco County Florida
  (1995-1997,2002-2004)
• Correlation Coefficients (Daily)
• Regression Coefficients (Daily, Weekly, and
  Monthly)
• EOF Analysis of Model and Trap Data
• Spring, Summer, and Fall (weekly)

4
Mosquitoes Aedes Aegypti

• Characteristics
• Urban, Container Breeding Mosquito
• Tropical Habitat
• Dengue Fever Vector
• Dengue Fever
• 100 Million Cases a Year Worldwide
• 4 Serotypes without Cross Immunity
• Dengue Hemorrhagic Fever from Multiple Infections

Picture taken from http//www.interet-general.info
/IMG/Aedes-Aegypti-2.jpg
Picture from http//www.cdc.gov/ncidod/dvbid/ deng
ue/map-distribution-2005.htm
5
Mosquitoes Culex Quinquefasciatus

• Characteristics
• Urban Mosquito
• Feeds on Humans and Animals
• West Nile Virus Vector
• West Nile Virus
• Arrived in New York 1999
• Symptoms Mild Fever-Encephalitis

Image taken from http//www.lahey.org/Medical/ Inf
ectiousDiseases/WestNileVirus.asp
Data from CDC.gov
6
Modeling Mosquitoes

• Inputs
• Temperature, Precipitation, Latitude
• Evaporation Derived (Hamons Equation)
• Irrigation/Land Cover
• Governing Rules
• Development Rates
• Death Rates
• Reproductive Rates
• Larval/Pupa Capacity
• Water Flux (sources and sinks)

7
Conceptual Model (DyMSiM) Dynamic Mosquito
Simulation Model
8
Data

• Temperature Data was Obtained from the National
  Climate Data Center
• Precipitation Data was Obtained from the National
  Climate Data Center and The Pasco County Vector
  and Mosquito Control District
• Mosquito Data was Obtained from the Pasco County
  Vector and Mosquito Control District

Image from http//pix.epodunk.com/locatorMaps/fl/F
L_8834.gif
9
Sample of Model Run
10
Regression Correlation Coefficients

• Regression Coefficient
• Best fit line in the data that minimizes the sum
  of the square of the error
• Shows how the magnitude of one variable changes
  with another
• Correlation Coefficient
• Calculated from the square root of the variance
  explained
• Describes the relationship between two variables
  (Range from -1 to 1)

11
Correlation/Pearson Coefficients
Time Span D-Value Average Correlation Significant (0.95)
1995-1997 0.4111 0.1343 Yes
2002-2004 0.3821 0.1190 Yes

Time Span Daily Regression Weekly Regression Monthly Regression
1995-1997 0.0910 0.7337 0.9354
2002-2004 0.0695 0.7068 1.0270
12
EOF Analysis

• Used to Analyze Spatial Patterns in a Dataset
• The 1st EOF Shows the Largest Fraction of
  Variance Explained in a Dataset
• Found from Eigenvalues and Eigenvectors
• Only a limited number of EOFs are Significant
  (North Test)

13
Spring North Test
- The first two EOFs in both Whisker Plots are
Significant
14
EOF 1 for Spring
1st EOF for Trap Data
1st EOF for Model Data
15
EOF 2 for Spring
2nd EOF Trap Data
2nd EOF Model Data
16
Summer North Test
Only EOF 1 is Significant for the Summer
17
EOF 1 for Summer
1st EOF for Trap Data
1st EOF for Model Data
18
Fall North Test
The 1st and 2nd EOFs are Significant
19
EOF 1 for Fall
1st EOF for Trap Data
1st EOF for Model Data
20
EOF 2 for Fall
2nd EOF for Trap Data
2nd EOF Model Data
21
Conclusions

• 1st EOF Dominates in Each Season for both Trap
  and Model Data
• One individual location sticks out in particular
  (Large Population)
• 2nd EOF Model and Trap Data share some common
  characteristics but are not identical
• Physical Mechanisms Behind the EOFs Need to be
  Analyzed (Surface Cover / Precipitation Patters)
• Overall, the EOF Analysis Supports the Utility
  and Accuracy of DyMSiM

22
Model Limitations

• All Models are Wrong, Some are Useful
  -George Box
• The model only accounts for climate and land use
  factors
• Predation, Pesticides, Food Availability, Human
  Behaviors, and Migration are not accounted for
• Trap Data is Not Truth
• Trapping mosquitoes may largely effect population
  dynamics
• Microenvironments are important for mosquitoes
  but are not caught with climate data

23

• Thank You for Your Attention
• Any Questions?