Modelling Bluetongue vectors occurence using GIS and Remote Sensing techniques

Martins H., Nunes T., Boinas F.

Introduction

� Objectives

� Spatial modelling of Culicoides imicola occurence

� Compare modelling techniques (discriminant analysis vs logistic regression)

Materials and Methods

� Base data for modelling

Temperature

- Entomological data

EggsLarvae

Animal density

Temperature

Humidity

- Entomological data

- Climatic and

- environmental data

- Animal population data

Materials and Methods

� Entomological data

- Ouput variable

- Entomological surveillance program� May 2005May 2005

� 3670 valid catches

� 216 sampling locations

- Data processing criteria� Maximum count of individual

� 2 valid catches (September – November)

� Treshold value = 10 specimens

� Boolean classification (73 neg, 47 pos)Absence Presence

Material and Methods

� Climatic and Environmental data

BIO1 = Annual Mean Temperature

BIO2 = Mean Diurnal Range (Mean of monthly (max temp - min temp))

BIO3 = Isothermality (P2/P7) (* 100)

BIO4 = Temperature Seasonality (standard deviation *100)

BIO5 = Max Temperature of Warmest Month

WorldClim package

BIO5 = Max Temperature of Warmest Month

BIO6 = Min Temperature of Coldest Month

BIO7 = Temperature Annual Range (P5-P6)

BIO8 = Mean Temperature of Wettest Quarter

BIO9 = Mean Temperature of Driest Quarter

BIO10 = Mean Temperature of Warmest Quarter

BIO11 = Mean Temperature of Coldest Quarter

BIO12 = Annual Precipitation

BIO13 = Precipitation of Wettest Month

BIO14 = Precipitation of Driest Month

BIO15 = Precipitation Seasonality (Coefficient of Variation)

BIO16 = Precipitation of Wettest Quarter

BIO17 = Precipitation of Driest Quarter

BIO18 = Precipitation of Warmest Quarter

BIO19 = Precipitation of Coldest Quarter

Material and Methods

� Climatic and Environmental data

- 30 remotely sensed variables

- NOAA satellite, AVHRR sensor� NDVI

LST Temporal encapsulation through Fourier trasnformation� LST

� MIR

Temporal encapsulation through Fourier trasnformation

MaximumMaximumMinimumMinimumAverageAverageAmplitudeAmplitudePhasePhase

MaximumMaximumMinimumMinimumAverageAverageAmplitudeAmplitudePhasePhase

Anual, bi-anual,

tri-anual cycles

Material and Methods

� Climatic and Environmental data

- SRTM sensor

Elevation Slope Slope AspectElevation Slope Slope extension

Aspect

Material and Methods

� Animal population data

- Bovines (SNIRB, DGV)

- Small ruminants (SNIRA, DGV)

- Kernel density

Bovines Small Ruminants Ruminants

Material and Methods

� Conceptual Framework

Results

� Discriminant analysis

- Significant variables� Mean Temperature of warmest quarter (-)

� Minimum LST (+)� Minimum LST (+)

� Maximum MIR (+)

� Tri-annual LST amplitude (+)

- Accuracy assessment� Se = 76,6%

� Sp = 75,3%

� Global Accuracy = 75,8%

Absence Presence

Results

� Logistic Regression- Significant variables

� Mean Temperature of warmest quarter (-)� Precipitation of wettest quarter (-)� Minimum NDVI (+)Slope (-)� Slope (-)

� Bi-annual MIR phase (-)� Tri-annual LST amplitude (+)� Minimum LST (+)� Annual MIR amplitude (+)� Mean temperature of driest quarter (+)

- Accuracy assessment� Se = 80,9%� Sp = 83,6%� Global Accuracy = 82,5%

Probability

Results

BTV1 and 4 outbreaks

(2006, 2007)

Absence Presence

Probability

(2006, 2007)

Methodology Class Temp. Warmest quarter

Min. LST

Slope Min. NDVI

Discriminant analysis

Absent 20,1 2778,4 7,3 1246,6

Present 23,5 2837,8 3,4 1345,7

Logistic regression

Absent 20,5 1946,9 4,8 919,1

Present 23,6 2839,0 2,9 1264,8

Conclusions

� Logistic regression models were more robust and accurate

� Temperature variables were significant in both modelsboth models

� Future developments:� Include other remote sensing variables

� Use higher resolution images (spatial, temporal and spectral)

� Repeat methodologies with time-dependent dataset

� Other methodologies (variable selection, modelling)

Work in progress

Acknowledgements

� Entomological Laboratory Staff

� National Veterinary Authority

Thank you!