PowerPoint Presentation.pdf

The 2017 ISCR in Lima, Peru, highlighted significant global climate change impacts on cocoa production. Key trends and projections include rising temperatures and variable precipitation patterns, with regional differences and high uncertainty. The IPCC 2013 report and Dai (2013) data analyzed 24 bioclimatic variables at a 20-arcmin resolution using WorldClim interpolated data for current climate and 10 Global Climate Models (RCP 6.0) for 2050s projections.

Random Forests for Classification:

Researchers employed Random Forests, decision tree ensembles, in two stages:

1. Unsupervised Learning:
- 250 forests with 4 variables to identify cocoa climate zones.
- 100 trees each in 100 forests for finer classification.

2. Supervised Learning:
- Balanced training set with 5 forests and 200 trees.
- 6 selected variables with node size 2 for precise categorization.

Climate Zones for Cocoa (Classification & Characteristics):

| CLUSTER | SYMBOL | TYPICAL REGION | TEMPERATURE | PRECIPITATION |
|---|---|---|---|---|
| 1 | WA | West Africa | High | Low, seasonal |
| 2 | SLA | Seasonal Latin America | Average | High, seasonal |
| 3 | HIH | Highlands | Low | Low |
| 4 | Ama | Amazon | Even | High |

Evaluation:

- Classification: Accurately grouped cocoa-growing regions based on climate characteristics.
- Votes for Occurrence Groups ("Suitability"): Provided insights into the suitability of different zones under future climate scenarios.
- Proximity Matrix: Used for novelty detection, helping identify new potential cocoa-growing areas or zones at risk due to climate change.