Coffee Pests and Diseases in Costa Rica

SPATIAL DECISION SUPPORT FOR COFFEE PESTS AND DISEASES RISK MANAGEMENT

IN COSTA RICAN AGROFORESTRY SYSTEMS

AFS – August 23 - 29 - Nairobi - Kenya

AVELINO, Jacques (CIRAD/IICA-PROMECAFE/CATIE)LADERACH, Peter (CIAT)COLLET, Laure (CIAT)BARQUERO, Miguel (ICAFE)CILAS, Christian (CIRAD)

1/13

Justifications and objective


Objective

To show how better decisions and disease risk-adapted agroforestry practices, for coffee growing regions, can be derived, based on spatial decision support tools and ground data.

Justifications

Patchiness in the distribution of plant pests and diseases due to spatial heterogeneity of the environment and the agronomic management.

Environmental information can be combined with spatial analyses to determine potential pests and diseases distribution, make better decisions and improve the risk management.

2/13

Study area, sampling design, shade and disease descriptors


Costa Rican coffee growing regions

Central Valley

Disease Descriptors

ALSD: Attack intensity index calculated on an attack and defoliation severity scale

CLR and CB: Maximal annual % of infected leaves

Main Diseases

Mycena citricolor : American Leaf Spot Disease (ALSD)

Hemileia vastatrix: Coffee Leaf Rust (CLR)

Phoma costarricensis: Coffee Blight (CB)

Sampling Design

Data from a two-year survey on coffee diseases in Costa Rica (Avelino et al., 2007)

27 geo-referenced plots sampled in Central Valley

Plot size surveyed: 100 coffee trees

Shade Assessment by using a spherical densiometer

Shade cover range: 0 - 65 %

3/13

American Leaf Spot Disease (ALSD) (Mycena citricolor)

Severe infection

Left: asexual fructifications (gemmae) Right: sexual fructifications (carpophore)

Lesions on leaf and fruits


Coffee Blight (CB) (Phoma costarricensis)

Lesion on leaf

4/13

Coffee Leaf Rust (CLR) (Hemileia vastatrix)

Rust lesion with uredosporesInfected leaf with

coalescent lesions

A coffee plantation before and after a severe Coffee Leaf Rust attack

AFS – August 23 - 29 - Nairobi - Kenya5/13

Spatial analyses and statistics

Bayesian statistics and spatial analyses

To delimitate areas with distinct disease risks as a function of environmental factors under two conditions of shade (below 15 % and above 15 %)

4 main steps in the model building stage

1. Identification of disease driving environmental factors (predictors) from literature

2. Disease driving environmental factors generated for the study region

by using the WorldClim climate database (Hijmans et al., 2005) and the Shuttle Radar Topography Mission (SRTM) data (Jarvis, 2004)

3. Probablity prediction for each condition of shade through Bayesian statistics: Prediction per attack intensity class (Pi for the class i ; n classes) as a function

of categorized environmental predictors

Calculation of a synthetic weighted variable, the score S

S = WiPi where Wi = i-1/n-1The higher the score, the higher the probability of high attack intensity.

4. Calculation of certainty, a measure of confidence of the prediction depending on the number of observations. The higher the certainty, the higher the confidence of prediction.


Raw environmental data fromWorldClim and SRTM

Disease driving environmental factors generated for the study region: rainfall (1 km resolution); slope % and aspect, elevation (90 m resolution)

Spatial analyses and statistics (cont.)

Observed geo-referenced disease attack intensities under low shade and high shade conditions

Bayesian Statistics(CaNaSTA algorithm, O’Brien 2004)

Predicted probability map of disease risk for two shade conditions

Low Shade %= 0-15

High Shade %= 15-65

Comparing score predictions with high certainty


Score prediction for Mycena citricolor attack intensity index

with high shade model (15 - 65%) and low shade model (0 -15 %)

15 - 65 % shade cover 0 - 15 % shade cover

Pi : Probablity prediction for class i of attack intensityWi = i-1/n-1

n: number of classes

Score = WiPi


Score value for Mycena citricolor attack intensity index

with high shade model (15 - 65%) and low shade model (0 -15 %)Certainty > 0.7

Certainty: a measure of confidence of the prediction depending on the number of observations

15 - 65 % shade cover 0 - 15 % shade cover


Comparison of score predictions for Mycena citricolor attack intensity index

with high shade (15 - 65%) and low shade (0 -15 %) cover

1. Low scores with high and low shade cover: environment unfavourable for disease development

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8

Prediction made with shade model

Pre

dic

tio

n m

ad

e w

ith

su

n m

od

el

4 behaviours :

2. Similar scores with high and low shade cover: no effect of shade

1

23. Higher scores with low shade cover : sun exposure is favourable to disease development

3

4

4. Higher scores with high shade cover : shade is favourable for disease development


Interactions shade-environment for Mycena citricolor development

10/13

3

4

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8


Pre

dic

tio

n m

ad

e w

ith

su

n m

od

el

0

3. Higher scores with low shade cover : sun exposure is favourable for M. citricolor development

4. Higher scores with high shade cover : shade is favourable for M. citricolor development

Comparison of driving environmental factors for groups 3 and 4

Group 3 Group 4

Rainfall June to August (mm)

1034 986

Rainfall August to December (mm)

1209 1155

Elevation (m) 1155 1109

Slope inclination (%) 9.4 9.5

Slope aspect (% of points with East or South

orientation)3 63

Significant differences, P < 0.05

In Central Valley, shade could be used for ALSD control on West and North oriented slopes, inadequately exposed to sun (decreased dew on coffee leaves ?), and should be avoided on East and South oriented slopes, well exposed to sun (decreased radiation ?)


Comparison of score predictions for Coffee Leaf Rust and Coffee Blight

with high shade (15 - 65%) and low shade (0 -15 %) cover

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0


Pre

dic

tio

n m

ad

e w

ith

su

n m

od

el

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0Prediction made with shade model

Pre

dic

tio

n m

ad

e w

ith

su

n m

od

el

Coffee Leaf Rust Coffee Blight

No clear interaction: in Central Valley, shade decreases CLR attacks (due to

probable fruit load reduction and decreased leaf susceptibility)


Interaction shade-altitude in Central Valley: increased CB attacks at high altitudes

(reduction of maximum temperatures by shade ?) and decreased CB attacks at lower

altitudes (increase of minimum temperatures by shade ?)

Altitude= 1399 m

Altitude= 1107 m

12/13


Conclusions

1. A method to delimitate areas with distinct disease risks based on spatial decision support tools and ground data

2. A method to analyze cropping practices effects, and especially shade effects

3. Evidence of interactions between shade and environment for coffee diseases

4. Need of site specific shade practices according to coffee disease risks and environment characteristics

13/13

Thank you

Coffee Pests and Diseases in Costa Rica

Travel

shade conditions low

low shade cover

low shade model

shade cover afs august23

conditions of shade

effect of shade

condition of shade

shade model prediction