Trader water use in supply chains

Introduction

This document presents analysis for a Trase Spotlight on water use in soy and beef traders’ supply chains and their link to drought (soy) and water scarcity (soy, beef) in Brazil. The analysis builds on previous analysis done per importing country, now with a focus on top Brazilian soy and beef traders.

The analysis focuses exclusively on 2017, but other years are also provided in this file. Pre-2018 trade data is the best we have, so the information here should have uncertainties minimized and therefore are more valuable for the Spotlight.

The analyses are as follows:

• Calculation of volume of freshwater use assigned to each top trader (separating soy and beef traders) in absolute terms (i.e. total m³ in a given year)
• Soy trader activity and the link to drought probability
• Soy and beef trader activity in river basins and their link to water scarcity as defined by the Brazilian National Water and Sanitation Agency (ANA) (especially “high” and “critical” water scarcity)
• Water scarcity footprint assessment (per municipality) and overlap with deforestation footprint in trader supply chains

Trader reliance on water use in soy and beef supply chains

Here we look at soy and beef traders separately and calculate the total water use associated with their supply chains in 2015-2017. Note that we use Exporter group as the designation for the time being.

Water use summary

Brazil-level summary

We first look at the overall water use linked to both soy and beef traders. Let’s start by calculating the total and blue water use linked to trade in 2015-2017 to give a summary of the big numbers.

First soy, blue water summary:

## # A tibble: 3 × 3
##   year  Mt_volumes km3_bw_rock
##   <chr>      <dbl>       <dbl>
## 1 2015        20.6       0.667
## 2 2016        23.1       1.17 
## 3 2017        19.9       0.736

then beef, blue water:

## # A tibble: 3 × 3
##   year  Mt_volumes km3_bw_tot
##   <chr>      <dbl>      <dbl>
## 1 2015        1.88       7.27
## 2 2016        1.89       7.17
## 3 2017        2.07       7.72

All of beef production requires blue water compared to soy which only uses some blue water for irrigation in a small set of municipalities.

Trader-level summary

We then break down water use per trader:

We note that beef traders require more water than soy traders in the 2015-2017 period. With >99% of water use being soil moisture regenerated by precipitation (or green water) we can do a deep dive into blue water use specifically.

Beef traders associated with > 0.5 km³ of blue water in 2015-2017 were JBS, Minerva, Marfrig and Mataboi.

Soy traders associated with blue water (as soy irrigation) in 2015-2017 were ADM, LDC, Bunge, Cargill, Cofco.

Special Note: The beef supply chain contains an “aggregated” municipality which is used to sum up small fractions of carcass weights that are traded. These volumes cannot be linked to a specific municipality (and by extension river basin) so we need to remove them from the beef trade data when trying to be specific about source locations (municipality or basin).

We provide here the sum of these beef aggregates volumes per trader (and keeping in mind that the “unknown” flows have already been removed).

The largest volume of “aggregated” flows associated to a trader was 2275 tonnes beef (as carcass weight, CW), so remaining a small fraction of the total, especially the larger traders like JBS (< 1000 tonnes), Minerva (< 2000 tonnes) and Mataboi (< 2000 tonnes).

We now plot the trader rankings separating the soy and beef traders

Soy traders and total water use (green + blue)

Soy total water use without river basins

First, we want to highlight the total water use that is linked to soy trader’s supply chains (both green and blue).

From the above graph, we extract the information for the top soy traders, namely: LOUIS DREYFUS, CARGILL, BUNGE, ADM, and COFCO

Now repeat the graphs by only showing top traders

Then we look at the corresponding table summary for top traders only:

## # A tibble: 3 × 7
##   year  volume_tot km3_tot km3_tot_rock Mt_volumes pct_km3 pct_vol
##   <chr>      <dbl>   <dbl>        <dbl>      <dbl>   <dbl>   <dbl>
## 1 2015   36525143.    70.2         125.       64.8    56.2    56.3
## 2 2016   33732809.    66.6         121.       60.5    55.2    55.7
## 3 2017   42154858.    75.7         136.       75.6    55.5    55.7

Soy total water use linked to river basin source

We repeat the above graphs by linking the water sources to river basins, and aggregating river basins that supply < 1 km³ of water into “Other”.

Followed by the corresponding summary table for soy (blue + green)

We can then calculate the total water sourced from the basins that are linked to top exporters

Soy traders and blue water use (irrigation)

We first look at top soy traders and the amount of irrigation needed for the export of soy, highlighting the basin of origin.

We first look at soy traders and their links to river basins:

and the summary of exports linked to irrigation:

and link to total irrigation

## # A tibble: 15 × 6
##    year  exporter_group product km3_gw km3_bw Mt_volume
##    <chr> <chr>          <chr>    <dbl>  <dbl>     <dbl>
##  1 2015  ADM            Soy       6.41 0.113       3.30
##  2 2015  BUNGE          Soy       5.32 0.0837      2.76
##  3 2015  CARGILL        Soy       4.14 0.0624      2.14
##  4 2015  COFCO          Soy       1.87 0.0293      1.01
##  5 2015  LOUIS DREYFUS  Soy       3.73 0.114       1.99
##  6 2016  ADM            Soy       4.53 0.217       2.40
##  7 2016  BUNGE          Soy       7.03 0.186       3.48
##  8 2016  CARGILL        Soy       5.74 0.149       2.82
##  9 2016  COFCO          Soy       2.87 0.0765      1.44
## 10 2016  LOUIS DREYFUS  Soy       2.81 0.204       1.65
## 11 2017  ADM            Soy       6.27 0.131       3.57
## 12 2017  BUNGE          Soy       5.09 0.159       2.78
## 13 2017  CARGILL        Soy       3.42 0.113       1.97
## 14 2017  COFCO          Soy       1.93 0.0350      1.10
## 15 2017  LOUIS DREYFUS  Soy       3.98 0.0765      2.36

It is interesting to note the concentration of companies beyond 2015 (seems quite spread out). The largest volumes look like are associated to 2016. The main companies that are linked to soy irrigation exported < 12 Mtonnes of soy in 2015-2017 that could potentially have irrigation, so we are looking at a smaller subset:

• Bunge: 2.8-3.5 Mtonnes
• ADM: 2.4-3.6 Mtonnes
• LDC: 1.7-2.4 Mtonnes
• Cargill: 2.0-2.8 Mtonnes

It looks like in 2017 there was NO sourcing from the Atlantico Nordeste Ocidental compared to the other years. This is why the above legends are different across years (no need to fix at this stage).

The above companies source mostly from:

• Paraná basin which has average water scarcity
• São Francisco which has critical water scarcity

Beef traders and blue water use

We then look at beef traders, their virtual water exports and link to river basins in the 2015-2017 period:

and the summary table:

## # A tibble: 3 × 7
##   year    volume km3_bw Mt_volumes km3_bw_tot pct_km3 pct_vol
##   <chr>    <dbl>  <dbl>      <dbl>      <dbl>   <dbl>   <dbl>
## 1 2015  1483404.   5.70       1.87       7.25    78.6    79.2
## 2 2016  1423194.   5.29       1.89       7.16    74      75.4
## 3 2017  1528998.   5.60       2.07       7.70    72.7    73.9

We can then calculate the total water sourced from the basins that are linked to top exporters

The profile of exporters follows the amount of beef the traders export since all beef production requires water:

• JBS is the largest exporter (0.70-0.78 Mtonnes CW) and associated with 2.4-2.8 km³ y^-1 of blue water
• Marfrig (0.26-0.35 Mtonnes CW) and associated with 1.12-1.33 km³ y^-1
• Minerva (0.32-0.37 Mtonnes CW) and linked to 1.26-1.49 km³ y^-1
  
• Mataboi (0.07-0.13 Mtonnes CW) and linked to 0.33-0.54 km³ y^-1

Here the basins are more spread out throughout the country and are typically:

• Amazon which has low water scarcity
• Paraná which has average water scarcity
• Paraguai which has average water scarcity
• Tocantins-Araguaia which has average water scarcity

Interestingly Marfrig stands out with its sourcing from Uruguai and Atlantico Sul which both experience high water scarcity

Links to water scarcity and drought

Trader links to water scarcity in river basins

Now we can link volumes of blue water to river basin water scarcity for each of the main players and provide a summary of links to water scarcity.

We can then sum up the risks to each of the companies based on the ANA designation of water scarcity.

We can then show the above results as bar graphs, but separating the soy and beef sectors into 2 analyses.

Beef traders and water scarcity

We look at beef traders and their link to water scarcity:

The results are really interesting and show that, for the largest exporting companies, there is a different profile of water scarcity risks, especially for Minerva, Marfrig and Mataboi. Marfrig is exposed to a larger portion of “high” water scarcity through Atlantico Sul and Uruguai river basins in Southern Brazil. “Critical” water risk was more prominent in Minerva and Mataboi Alimentos due to activities in the São Francisco basin.

Soy traders and water scarcity (irrigation)

We looked at the largest traders and focused on Louis Dreyfus, Cargill, Bunge, ADM, and COFCO and their link to critical water scarcity in 2015-2017.

• LOUIS DREYFUS has the lowest volume of blue water linked to critical scarcity with at most 0.25 Mtonnes in 2016 (3-32%, 0.002-0.065 km³ y^-1), the majority of which is located in the Paraná basin which has average water scarcity (3.3-4.5 Mtonnes). Louis Dreyfus’ link to critical water scarcity is through the São Francisco basin.
• CARGILL shows 0.49-1.31 Mt of soy linked to critical water scarcity (55-73%, 0.043-0.082 km³ y^-1) which is mostly concentrated in the São Francisco which is where most of the soy with irrigation in being sourced from (as well as some Paraná)
• BUNGE: 0.81-1.09 Mt of soy (52-67%, 0.051-0.12 km³ y^-1) which is mostly linked to the São Francisco basin (and some Paraná)
• ADM: 0.66-1.61 Mt of soy linked to river basins with critical water scarcity (61-66% blue water volume, 0.0069-1.4 km³ y^-1). ADM is the company that source a lot of volume from the São Francisco, but there is also variability among years, with 2016 being a little different.
• COFCO: 0.28-0.40 Mt of soy (65-87%, 0.02-0.07 km³ y^-1) with volumes sourced in São Francisco and Parana

Mapping water scarcity risks

We now have a look at the locations of these sources for all companies.

First we map total water use and sources

Then we focus on blue water

The maps above put the total volumes of blue water and their link to critical water scarcity into perspective:

• LOUIS DREYFUS was mostly linked to the Paraná basin in 2017 (“average” water scarcity) but showed some volumes in the São Francisco which has “critical” water scarcity. This suggests some year-on-year variability in sourcing. About 3.4-4.6 Mtonnes of soy are sources from basins with “average” water scarcity.
• CARGILL has a strong link to the São Francisco as well in 2017 and then other basins with average scarcity such as the Paraná and the Tocantins-Araguaia
• BUNGE has even more presence in the São Francisco in 2017 (with Matopiba). Blue water volumes were lowest in 2015 and some presence in Tocantins-Araguaia and Paraná which both experience average water scarcity (3.7-4.6 Mtonnes of soy sourced from average water scarcity basins) compared to 0.8-1.1 Mtonnes from São Francisco.
• ADM is linked to both the Paraná and São Francisco, but more upstream of the São Francisco (instead of the Matopiba region). The company seems to be sourcing the most soy from the São Francisco and therefore might be at most risk of water scarcity.
• COFCO is also largely sourcing it’s irrigated soy from the São Francisco

General conclusions for soy

The combined volume of soy with irrigation sourced by the soy traders in 2015-2017 is: 11.2-11.8 Mtonnes, representing 0.4-0.83 km³ y^-1

All companies source a small portion of their soy from areas that potentially use irrigation. A relatively large portion of this sourced soy comes from basins that experience critical water scarcity. While soy is almost entirely rainfed, we expect the use of irrigation to serve as additional insurance for production and so this type of sourcing is important to understand.

As the number of municipalities with irrigation are small, this focused analysis can help companies review their sourcing and pay more attention to irrigation and irrigation practices, challenges in basins experiencing critical water scarcity.

Given that the majority of water used for soy producion is green, we also look at drought probability in these trader supply chains (see Section below).

Beef animal water use and reservoir evaporation

Only very small amount of beef are sourced from river basins with high/critical water scarcity:

• JBS: <2% of volume linked to basins with critical or high water scarcity, most volume in low/average water scarcity
• MARFRIG: is mostly linked to high water scarcity (39-47% of blue water, 0.52-0.58 km³ y^-1) which is linked to Atlantico Sul and Uruguai river basins in the South of the country
• MINERVA: 5-7.8% of volume linked to high/critical water scarcity (0.07-0.11 km^{3 y-1})
• MATABOI: is also mostly linked to average water scarcity but has 9-15% of volume linked to critical water scarcity (0.03-0.08 ^km3 y^-1)

Unlike soy there are many municipalities that can source beef for exports and we can expect large discrepancies in the “footprint” across the country.

We then look at 2017 beef:

2016 beef

and 2015 beef

• JBS sourcing for exports is quite spread out across the country with larger pockets in the Tocantins-Araguaia, Paraná, Paraguai basins. The company also seems to be sourcing very little from the South (i.e nothing in 2017)
• MARFRIG and its greater presence in the South makes them more linked to basins with high water scarcity like the Uruguai. They also do not source at all from the São Francisco basin (at least for exports).
• MINERVA follows more of the JBS sourcing
• MATABOI is unique in its sourcing pattern for exports with a concentration in the Central East part of the country, Paraná, Tocantins-Araguaia and São Francisco river basin (upstream).

Soy trader links to drought probability

Given that soy is almost entirely rainfed we take a look at how traders are sourcing soy according to the probability of drought in its supply chain. Drought probability was derived following the method of de Petrillo et al (2023) in which the self-calibrated Palmer Index was used to identify dry months (scPI < -2) between Jan 1958 and Dec 2018. The metric uses the self-calibrated Palmer Drought Severity Index which ranges from -4 (extremely dry) to +4 (extremely wet). The probability is calculated for each municipality for all months between Jan 1958 and Dec 2018 and counts all months where the index was < -2. (de Petrillo et al 2023).

The total water footprint is determine by yield and so it is worth also checking the link between the soy water footprint and the drought probability. We can check that with the trade data at first and dig deeper later on if needed.

After carrying out our own code to extract data from CRU, we do not find the same probabilities as de Petrillo et al. (2023). Below is the summary of the drought probability across Brazil.

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00   10.30   19.20   19.53   26.80   62.70

We can create 5 categories of drought probability:

• 0
• 0-10%
• 10-20%
• 20-40%
• larger than 40%

Let’s first look at what the drought probability looks like across the country

## # A tibble: 27 × 4
##    nm_uf            sigla_uf mean_drought_prob sd_mean_drought_prob
##    <chr>            <chr>                <dbl>                <dbl>
##  1 ACRE             AC                      17                  4.5
##  2 ALAGOAS          AL                      27                 13.5
##  3 AMAPÁ            AP                      36                  6.2
##  4 AMAZONAS         AM                       8                  5.2
##  5 BAHIA            BA                      31                 11.1
##  6 CEARÁ            CE                      25                  4.2
##  7 DISTRITO FEDERAL DF                      21                 NA  
##  8 ESPÍRITO SANTO   ES                       7                  2  
##  9 GOIÁS            GO                      16                  5.8
## 10 MARANHÃO         MA                      30                  9.5
## # ℹ 17 more rows

Then let’s put this on a map.

Then, let’s check the correlations between drought probability and the water footprint.

## 
## Call:
## lm(formula = soy_wf_drought$prob_drought_mean_pct ~ soy_wf_drought$soy_wf_rock)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -35.458  -5.949  -1.731   5.230  37.947 
## 
## Coefficients:
##                             Estimate Std. Error t value            Pr(>|t|)    
## (Intercept)                5.0563505  0.2258479   22.39 <0.0000000000000002 ***
## soy_wf_drought$soy_wf_rock 0.0041140  0.0001149   35.79 <0.0000000000000002 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.341 on 44563 degrees of freedom
## Multiple R-squared:  0.02794,    Adjusted R-squared:  0.02792 
## F-statistic:  1281 on 1 and 44563 DF,  p-value: < 0.00000000000000022

There isn’t really any relationship seen here which is expected because the drought probability is over a long period of time and the soy water footprint and yield are one specific year.

We then look at each trader’s municipal soy supply as it relates to drought probability as a means to highlight a climatic risk. We take the soybean that is purely rainfed, so municipalities with irrigation are removed from the analysis.

We then plot link drought probability in municipalities of production with soy trader supply chains.

This first cut of exposure shows the following:

• There are different exposures for each trader, some traders are linked to municipalities that have greater drought probability (e.g. BUNGE, CARGILL, ADM) compared to others (COFCO, LOUIS DREYFUS)
• Traders that exposure smaller volumes of soy seem to be less linked to higher drought probability (it could be a volume thing)

Then we can plot the locations where these drought probabilities are taking place for each of the traders.

Note that we remove municipalities with irrigation here.

We look at 2017 soy, note that the municipalities with drought probability > 40% have a red outline (some are hard to see, so you can see the total number of municipalities in the title of each graph with a table underneath):

## [1] year           exporter_group name           state          macro_basin   
## [6] volume_t      
## <0 rows> (or 0-length row.names)

##   year exporter_group                 name    state
## 1 2017        CARGILL           DOM ELISEU     PARÁ
## 2 2017        CARGILL          PARAGOMINAS     PARÁ
## 3 2017        CARGILL          ULIANÓPOLIS     PARÁ
## 4 2017        CARGILL           BOM JARDIM MARANHÃO
## 5 2017        CARGILL BOM JESUS DAS SELVAS MARANHÃO
##                    macro_basin  volume_t
## 1           Tocantins-Araguaia 165282.53
## 2           Tocantins-Araguaia  77126.18
## 3 Atlântico Nordeste Ocidental  14402.81
## 4 Atlântico Nordeste Ocidental  15780.00
## 5 Atlântico Nordeste Ocidental   5400.00

##   year exporter_group                    name    state
## 1 2017          BUNGE       GOIANÉSIA DO PARÁ     PARÁ
## 2 2017          BUNGE         IPIXUNA DO PARÁ     PARÁ
## 3 2017          BUNGE             PARAGOMINAS     PARÁ
## 4 2017          BUNGE          RONDON DO PARÁ     PARÁ
## 5 2017          BUNGE              TRACUATEUA     PARÁ
## 6 2017          BUNGE             ULIANÓPOLIS     PARÁ
## 7 2017          BUNGE      ITINGA DO MARANHÃO MARANHÃO
## 8 2017          BUNGE VILA NOVA DOS MARTÍRIOS MARANHÃO
##                    macro_basin     volume_t
## 1           Tocantins-Araguaia     99.43612
## 2           Tocantins-Araguaia   2947.19282
## 3           Tocantins-Araguaia 225114.08127
## 4           Tocantins-Araguaia  39074.08392
## 5 Atlântico Nordeste Ocidental    570.58183
## 6 Atlântico Nordeste Ocidental 109800.92551
## 7 Atlântico Nordeste Ocidental  35574.59182
## 8           Tocantins-Araguaia   8960.33471

##   year exporter_group        name state                  macro_basin  volume_t
## 1 2017            ADM PARAGOMINAS  PARÁ           Tocantins-Araguaia 108346.04
## 2 2017            ADM ULIANÓPOLIS  PARÁ Atlântico Nordeste Ocidental  39692.13

##   year exporter_group       name state        macro_basin volume_t
## 1 2017          COFCO DOM ELISEU  PARÁ Tocantins-Araguaia 6146.774

Coincidence of deforestation and water scarcity impacts

We now look at the 80th percentile of municipalities supplying traders considering:

• municipalities in the supply chain with largest deforestation footprint
• municipalities in the supply chain with largest water scarcity footprint

We can then compare the municipalities and say something about hotspots and show where, from the supply chain focus, the largest impacts overlap.

We start by looking at soy which has more of a dispersed supply given that a small portion of soy is irrigated. Leaving out COFCO at the moment, but can add later

We see very little overlap in municipalities when looking at both deforestation and water scarcity footprints of main soy traders. Overlaps mostly happen in the Matopiba region where there was both deforestation and irrigation. Bunge is one of the most interesting cases as the Matopiba shows a region of concentration of both deforestation (green) and water scarcity impacts (blue) and the overlap (red).

There seems to be more overlap in 2015 and 2016, see BUNGE and ADM.

This is very different than beef which has a much wider spread of performance across regions in Brazil.

We can now derive the total amount of soy and beef in the supply chain that is included in these “overlap” municipalities, starting with soy:

## # A tibble: 15 × 3
##    year  exporter_group  volume
##    <chr> <chr>            <dbl>
##  1 2017  ADM            504129.
##  2 2017  CARGILL        422330.
##  3 2017  BUNGE          307750.
##  4 2017  LOUIS DREYFUS  215391.
##  5 2017  COFCO           97686.
##  6 2016  CARGILL        308813.
##  7 2016  BUNGE          290658.
##  8 2016  LOUIS DREYFUS  187777.
##  9 2016  COFCO          183650.
## 10 2016  ADM             38935.
## 11 2015  CARGILL        487002.
## 12 2015  BUNGE          361236.
## 13 2015  LOUIS DREYFUS  215534.
## 14 2015  ADM            123949.
## 15 2015  COFCO           90544.

and the beef:

## # A tibble: 12 × 3
##    year  exporter_group    volume
##    <chr> <chr>              <dbl>
##  1 2017  JBS               44241.
##  2 2017  MINERVA           35351.
##  3 2017  MATABOI ALIMENTOS 14407.
##  4 2017  MARFRIG            5221.
##  5 2016  JBS               54118.
##  6 2016  MINERVA           35639.
##  7 2016  MATABOI ALIMENTOS 12264.
##  8 2016  MARFRIG             974.
##  9 2015  JBS               57156.
## 10 2015  MINERVA           33765.
## 11 2015  MATABOI ALIMENTOS  7658.
## 12 2015  MARFRIG             566.

General conclusions

We know that most of the beef produced in Brazil is consumed domesticallys, so we can expect quite a range of performances regarding domestic supply chains. In terms of exports, each company does show a different profile, with JBS/MARFRIG being quite spread out and those in the Cerrado being more exposed to water scarcity (Mataboi).

Overall, exporters seem less exposed to water scarcity risks in their supply chain, but it is still interesting to see the differences here.

We observe some overlap of irrigated soy with largest deforestation footprints in key municipalities of Matopiba which can highlight target action. Trader presence across Brazil shows overlap of largest deforestation footprints with largest water scarcity footprints.

References

De Petrillo et al (2023) International corporations trading Brazilian soy are keystone actors for water stewardship Communications Earth and Environment 87, 4(1), doi: 10.1038/s43247-023-00742-4.