SEI-PCS quality flags

Introduction

This study tests the proposal of data quality flags to be applied to Brazilian soy SEI-PCS results pre- and post-2017 considering the following flags:

pre-2017:

• 1a: none
• 1b: 1.1 FARM IN STATE OF ORIGIN, 1.2 FARM NOT IN STATE OF ORIGIN
• 2a: 2.1 SILO IN STATE OF ORIGIN, 2.2 SILO NOT IN STATE OF ORIGIN, 3.1 CRUSHING/RETAIL IN MUNICIPALITY WITH FARM/SILO
• 2b: 3.2 CRUSHING/RETAIL WITH SINGLE FARM/SILO IN STATE OF ORIGIN, 3.3.1 CRUSHING/RETAIL WITH SINGLE SICASQ FARM/SILO IN STATE OF ORIGIN
• 2c: 3.3.2 CRUSHING/RETAIL WITH CLOSEST SICASQ FARM/SILO IN STATE OF ORIGIN, 3.3.3 CRUSHING/RETAIL WITH CLOSEST FARM/SILO IN STATE OF ORIGIN
• 3a: 3.4 CRUSHING/RETAIL IN MUNICIPALITY WITH LARGE PRODUCTION, 4.1 MUNICIPALITY WITH LARGE PRODUCTION
• 3b: none
• Ua and Ub: as “UNKNOWN” (cannot be separated at this stage)

post-2017:

• 1a: none
• 1b: none
• 2a: 1.1 - FARM/SILO
• 2b: none
• 2c: none
• 3a: 1.2 - LARGE EXPORTER, 1.3 - LARGE PRODUCER
• 3b: 2.1 - MDIC/CNPJ STATE + STICKY HUBS (CNPJ 14), 2.2 - MDIC/CNPJ STATE + STICKY HUBS (CNPJ 8), 2.3 - MDIC/CNPJ STATE + CLOSEST HUB, 3.1 - MDIC STATE + AVERAGE PORT-COMMODITY-STATE HUBS, 3.2 - MDIC STATE + AVERAGE COMMODITY-STATE HUBS, 3.3 - MDIC STATE + AVERAGE STATE HUBS
• Ua and Ub: as “UNKNOWN” (cannot be separated at this stage)

Results with Brazilian soy

Once we assign the flags, we can look at the internal flags to compare results

## # A tibble: 20 × 5
##     year flag_int       vol   vol_tot   pct
##    <dbl> <chr>        <dbl>     <dbl> <dbl>
##  1  2016 1b       11488255. 67597523.  17  
##  2  2016 2a       30998558. 67597523.  45.9
##  3  2016 2b        6588758. 67597523.   9.7
##  4  2016 2c        9214794. 67597523.  13.6
##  5  2016 3a        2216986. 67597523.   3.3
##  6  2016 U         7090172. 67597523.  10.5
##  7  2017 1b       13660992. 83966862.  16.3
##  8  2017 2a       38589821. 83966862.  46  
##  9  2017 2b        6826070. 83966862.   8.1
## 10  2017 2c       13608749. 83966862.  16.2
## 11  2017 3a        2930414. 83966862.   3.5
## 12  2017 U         8350816. 83966862.   9.9
## 13  2019 2a       14752883. 84226235.  17.5
## 14  2019 3a       12322264. 84226235.  14.6
## 15  2019 3b       42942094. 84226235.  51  
## 16  2019 U        14208994. 84226235.  16.9
## 17  2020 2a       16869166. 97927513.  17.2
## 18  2020 3a        9644257. 97927513.   9.8
## 19  2020 3b       53655767. 97927513.  54.8
## 20  2020 U        17758323. 97927513.  18.1

And now look at the results that we would communicate to the users

## # A tibble: 14 × 6
##     year flag_ext flag_ext_desc            vol   vol_tot   pct
##    <dbl> <chr>    <chr>                  <dbl>     <dbl> <dbl>
##  1  2016 1        to production      11488255. 67597523.    17
##  2  2016 2        to facility        46802110. 67597523.    69
##  3  2016 3        to remote facility  2216986. 67597523.     3
##  4  2016 U        unknown             7090172. 67597523.    10
##  5  2017 1        to production      13660992. 83966862.    16
##  6  2017 2        to facility        59024641. 83966862.    70
##  7  2017 3        to remote facility  2930414. 83966862.     3
##  8  2017 U        unknown             8350816. 83966862.    10
##  9  2019 2        to facility        14752883. 84226235.    18
## 10  2019 3        to remote facility 55264358. 84226235.    66
## 11  2019 U        unknown            14208994. 84226235.    17
## 12  2020 2        to facility        16869166. 97927513.    17
## 13  2020 3        to remote facility 63300025. 97927513.    65
## 14  2020 U        unknown            17758323. 97927513.    18

Plots of results

We explore how to best represent these differences and communicate them.

We now look at the external flag