We want to model the cocoa production data per department, for CIV and 2019 (2020-2021 for ETH). The available data sources are:

The three cocoa land use maps were prepared using different methodologies and resulted in a large difference in total area under cocoa production between the three:

The ETH estimates the CIV’s area under cocoa sums up to

## [1] 4313317

hectares.

The JRC estimates the CIV’s area under cocoa sums up to

## [1] 3684919

hectares. It is important to note that the JRC did not analyse the northern part of the CIV, where they assume cocoa production is very limited due to climatic unsuitability.

Vivid Economics, on the other hand, analysed the whole CIV and estimates the total area under cocoa is

## [1] 2778997

hectares.

NOTE: the first version Of JRC map had an cocoa area of 4 807 712 ha, which approached the FAO’s estimate (FAOSTAT 2018: 4 015 394 ha) on area under cocoa production better than VIVID, and since they took into account both full-sun and shaded cocoa, their data were used as the baseline data for the analyses. However, ETH’s map received in September 2021 is closer to the FAO estimate, and the new published version of JRC map shows lower accuracy.

To get an idea of the surface under cocoa we may be missing using JRC map - because the JRC didn’t analyse the northern departments - we can calculate the proportion of the total area under cocoa in Vivid Economics’ map that is located in these northern departments. In Vivid Economics’ map, this proportion sums up to

## [1] 0.892123

percent of the total area under cocoa.

In ETH’s map, this proportion sums up to

## [1] 6.181741

Theoretically, by using the JRC’s data, this is the percentage of area under cocoa production excluded in our analyses.

Plot cocoa area per department

**

Area under cocoa per department to cocoa production per department

It is clear from the cocoa suitability map that not all areas in the CIV area equally suitable to grow cocoa efficiently. Because of this, we cannot simply use area under cocoa per department to divide the country’s total production over the departments weighted by area under cocoa.

However, we can calculate the relative suitable area per department, representing the area under fully productive cocoa, using the suitability map.

To do this, we combined the cocoa land use map and the cocoa suitability map in Google Earth Engine by multiplying each pixel of cocoa by its respective suitability (coded as 0 to 1, not suitable to best suitable), multiplying this value by pixel size and summing these last values per department.

As each unit of this relative suitable area is scaled to represent an equal part of the potential production, we can calculate the cocoa yield per ha by dividing the total production estimate from ICCO by the total relative suitable area for the whole of CIV. Doing so we obtain the following:

Yield per relative suitable area (kg/ha):

ETH (We use 2019 ICCO data even if ETH is a map of the 2020 cocoa land use, because Trase mapped the 2019 trade data - It is assumed that the change in cocoa land use from 2019 to 2020 is not significant):

[1] 945.157

JRC:

[1] 915.3305

VIVID:

[1] 1419.238

The yields are high because the relative suitable cocoa area is lower than the total cocoa area.

Based on this yield value, we can calculate the total production per department; by multiplying it with the relative suitable area per department.

These production values per department can be presented visually as follows:

A summary of the total production values (in kg) per department is:

ETH:

##      Min.   1st Qu.    Median      Mean   3rd Qu.      Max. 
##         0    351001   8549307  19944444  33887569 123301485

JRC:

##     Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 
##        0        0 12009400 19944444 32166331 86985602

VIVID:

##      Min.   1st Qu.    Median      Mean   3rd Qu.      Max. 
##         0      8573   2391545  19944444  22777461 165402238

**

The KIT study provides 420 values of farm yields, how do these compare to our yield calculation?

A summary of the yield (kg/ha) data from the KIT study is:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     0.0   189.2   311.0   349.0   484.3  1075.0

While our RELATIVE yield value is:

ETH:

[1] 945.157

JRC:

[1] 915.3305

VIVID:

[1] 1419.238

It is logical that our yield value is on the higher end of the range of the KIT study’s yield values: the KIT study’s yield represents the yield of the complete area, not just of the suitable area ; ours is calculated by dividing the total ICCO cocoa volume not by the total cocoa area, but by the cocoa area weighed according to their suitability for cocoa production - therefore a much smaller area, which gives a higher yield - but as the yield is multiplied by the relative suitable area in each department and not by the total cocoa area in each department, it is not a problem. Direct comparison between the two results is therefore not possible.

**

DEAD END BUT INTERESTING VISUALISATION OF COOP LOCATION RELATIVE TO JRC MAP (Not updated with ETH)

The values of cocoa production per geocode can be used to assign an estimated size to each of the coops included in each geocode. To calculate this estimated size (i.e. its production level), we can divide the production per geocode by the number of coops per geocode and assign that production value to each coop in that geocode.

it makes no sense to do this because presumably not all farmers are included in cooperatives

A summary of the total production per coop is:

In this way, some coops are assigned a total production of 0 kg of cocoa. These coops may be located in departments that weren’t considered in the JRC’s analyses, or the coops are assigned to the wrong department This last option is possible when coops are located on the border of two departments.

How many cooperatives present this issue?

Where are these cooperatives located?

i.e. they are located in departments for which the JRC did not analyse cocoa cover.