There are 12 ship types in the inventory, known as vessel classifications. These are imported from a CSV provided by PLA. A sample is shown below.
# A tibble: 67 x 6
registry_class aggregated_class group code release_height fuel_type
<chr> <chr> <dbl> <chr> <dbl> <chr>
1 ferry Passenger (ferry) 1 OFY 5 Diesel
2 cutter suction dredger Dredger 2 DCS 17.5 Diesel
3 dredger Dredger 2 DDR 17.5 Diesel
4 hopper dredger Dredger 2 DHD 17.5 Diesel
5 suction hopper dredger Dredger 2 DSH 17.5 Diesel
6 sand suction dredger Dredger 2 DSS 17.5 Diesel
7 trailing suction dredger Dredger 2 DTD 17.5 Diesel
8 trailing suction hopper dredger Dredger 2 DTS 17.5 Diesel
9 fishing (general) Fishing 2 FFS 17.5 Diesel
10 trawler (All types) Fishing 2 FTR 17.5 Diesel Each type of ship has a code and group. There are four groups.
PLA 2016 emissions covering London are imported as a CSV. These contain for each pollutant, ship type and cellid (LAEI exact cut) the estimated shipping emissions in kg per annum for that area (approx. 1km x 1km). A sample is shown below.
# A tibble: 3,159 x 5
ship_type pollutant cellid sailing berth
<chr> <chr> <dbl> <dbl> <dbl>
1 Bulk carrier NOx 1584 41.8 0
2 Bulk carrier NOx 911 22.6 0
3 Bulk carrier NOx 907 0.5 0
4 Bulk carrier NOx 2211 9.53 0
5 Bulk carrier NOx 2205 49.8 0
6 Bulk carrier NOx 2246 10.5 0
7 Bulk carrier NOx 860 72.9 0
8 Bulk carrier NOx 2208 68.5 0
9 Bulk carrier NOx 2212 2.99 0
10 Bulk carrier NOx 1587 40.7 0
The vessel_classifications are joined to the inventory_export_2016.csv by ship_type. However there are three ship_type's in the emissions that do not exactly match with the vessel_classifications file. These are edited to force them to link, as below.
| Emissions ship_type | Vessel classification ship_type |
|---|---|
| 'RoRo Cargo / Vehicle' | 'RoRo Cargo/Vehicle' |
| 'Cruise ship' | 'Passenger (cruise)' |
| 'Passenger' | 'Passenger (ferry)' |
The emissions now look like this:
# A tibble: 3,159 x 6
ship_type pollutant cellid sailing berth group
<chr> <chr> <dbl> <dbl> <dbl> <dbl>
1 Bulk carrier NOx 1584 41.8 0 3
2 Bulk carrier NOx 911 22.6 0 3
3 Bulk carrier NOx 907 0.5 0 3
4 Bulk carrier NOx 2211 9.53 0 3
5 Bulk carrier NOx 2205 49.8 0 3
6 Bulk carrier NOx 2246 10.5 0 3
7 Bulk carrier NOx 860 72.9 0 3Using the group, pollutant and cellid columns they are now grouped to give the below.
# A tibble: 1,590 x 5
pollutant cellid group sailing berth
<chr> <dbl> <dbl> <dbl> <dbl>
1 NOx 231 1 54.6 0
2 NOx 231 2 6778. 2.09
3 NOx 231 3 1429. 0
4 NOx 231 4 3970. 4567.
5 NOx 232 1 3.35 0
6 NOx 232 2 521. 1.54
7 NOx 232 3 95.7 0
8 NOx 232 4 10367. 29460.
9 NOx 234 1 2.28 0
10 NOx 234 2 332. 0 This data is now joined to a geopackage of the LAEI grid exact cuts using the cellid identifier. Then the LAEI exact cuts are dissolved to create regular 1km by 1km grids n.b. this is not a spatial aggregate, actually the grid centroid coordinates are used to create the grids, and then linked. The data (grid_emissions) now looks like this:
Simple feature collection with 933 features and 5 fields
geometry type: POLYGON
dimension: XY
bbox: xmin: 523000 ymin: 174000 xmax: 558000 ymax: 187000
epsg (SRID): 27700
proj4string: +proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +towgs84=446.448,-125.157,542.06,0.15,0.247,0.842,-20.489 +units=m +no_defs
# A tibble: 933 x 6
pollutant group large_grid_id sailing berth geometry
<chr> <dbl> <dbl> <dbl> <dbl> <POLYGON [m]>
1 NOx 1 9559 0.17 0 ((548000 183000, 548000 182000, 547000 182000, 547000 183000, 548000 183000))
2 NOx 1 9717 0.26 0 ((534000 182000, 534000 181000, 533000 181000, 533000 182000, 534000 182000))
3 NOx 1 9728 20.9 0 ((545000 182000, 545000 181000, 544000 181000, 544000 182000, 545000 182000))
4 NOx 1 9729 82.8 0.01 ((546000 182000, 546000 181000, 545000 181000, 545000 182000, 546000 182000))
5 NOx 1 9730 62.9 0 ((547000 182000, 547000 181000, 546000 181000, 546000 182000, 547000 182000))
6 NOx 1 9731 59.4 0 ((548000 182000, 548000 181000, 547000 181000, 547000 182000, 548000 182000))
7 NOx 1 9732 108. 0 ((549000 182000, 549000 181000, 548000 181000, 548000 182000, 549000 182000))
8 NOx 1 9733 45.4 0 ((550000 182000, 550000 181000, 549000 181000, 549000 182000, 550000 182000))
9 NOx 1 9734 9.8 0 ((551000 182000, 551000 181000, 550000 181000, 550000 182000, 551000 182000))
10 NOx 1 9886 22179. 34.1 ((531000 181000, 531000 180000, 530000 180000, 530000 181000, 531000 181000))
# ... with 923 more rows
A 20m x 20m polygon grid (small_grid) covering the extent of the grid_emissions was now created. This was then intersected and clipped with a unique geographical representation of the emissions grid i.e. just one cell for each emission area, rather than twelve cells (3 pollutants x 4 groups). The link between the small_grid and the grid_emissions is by large_grid_id.The small_grid data looks like this:
Simple feature collection with 334932 features and 2 fields
geometry type: GEOMETRY
dimension: XY
bbox: xmin: 523000 ymin: 174000 xmax: 558000 ymax: 187000
epsg (SRID): 27700
proj4string: +proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +towgs84=446.448,-125.157,542.06,0.15,0.247,0.842,-20.489 +units=m +no_defs
First 10 features:
large_grid_id small_grid_id geometry
1 9559 1 POINT (547000 182000)
2 9559 2 LINESTRING (547020 182000, ...
3 9559 3 LINESTRING (547040 182000, ...
4 9559 4 LINESTRING (547060 182000, ...
5 9559 5 LINESTRING (547080 182000, ...
6 9559 6 LINESTRING (547100 182000, ...
7 9559 7 LINESTRING (547120 182000, ...
8 9559 8 LINESTRING (547140 182000, ...
9 9559 9 LINESTRING (547160 182000, ...
10 9559 10 LINESTRING (547180 182000, ...
366 days of AIS data for 2016 were provided by the PLA. These were processed using the Python libais library to extract the latitude, longitude and ship_type (group), and the data exported as an R Dataframe. A sample is shown below.
lon lat VESSEL_TYPE
1 0.3968550 51.44474 XTG
2 0.3413600 51.46286 GPC
3 0.3268333 51.45400 GGC
4 0.2523367 51.46938 URR
5 0.3962933 51.44480 XTG
6 0.3494783 51.45776 GPCOn 1 January 2016 (taken as an example day) there were 1,368,454 GPS points in the dataset, 336,967 (25%) of which had no category. The remaining categories and counts are as below.
BBU BCE BWC DTD DTS GGC GPC OFY OSU OYT PRR RSR TCO TEO TPD URR XFF XTG
112277 21895 97107 43833 3207 30064 17034 116524 43235 16190 26679 53785 82291 17810 4055 44135 42119 259246The AIS data was imported in turn, and spatially joined to the small_grid to create the small_grid_result. This being that each small_grid contained the count of the total number of GPS points, per group, that had been recorded in that grid square over 2016. 20m cells with less than 20 GPS points in them were treated as GPS drift error and were discarded. The map and data below show the annual count of GPS points within each grid square.
Simple feature collection with 1223680 features and 5 fields
geometry type: GEOMETRY
dimension: XY
bbox: xmin: 523000 ymin: 174000 xmax: 558000 ymax: 187000
epsg (SRID): 27700
proj4string: +proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +towgs84=446.448,-125.157,542.06,0.15,0.247,0.842,-20.489 +units=m +no_defs
First 10 features:
cellid small_grid_id group count berth_name geometry
1 231 1 1 NA <NA> POLYGON ((555000 177020, 55...
2 231 2 1 NA <NA> POLYGON ((555000 177000, 55...
3 231 3 1 NA <NA> POLYGON ((555020 177000, 55...
4 231 4 1 NA <NA> POLYGON ((555040 177000, 55...
5 231 5 1 NA <NA> POLYGON ((555060 177000, 55...
6 231 6 1 NA <NA> POLYGON ((555080 177000, 55...
7 231 7 1 NA <NA> POLYGON ((555100 177000, 55...
8 231 8 1 NA <NA> POLYGON ((555120 177000, 55...
9 231 9 1 NA <NA> POLYGON ((555140 177000, 55...
10 231 10 1 NA <NA> POLYGON ((555160 177000, 55...
The grid_emissions are split between sailing and berth. The berth emissions will be distributed to the berths that are within each large_grid_id, weighted by the number of GPS points within the small_grid cell of the berth. A berths.shp was therefore imported, and joined to the small_grid, adding an extra column to identify if that cell held a berth or not. The berths data looked like this
Simple feature collection with 180 features and 1 field
geometry type: POINT
dimension: XY
bbox: xmin: -1.797693e+308 ymin: -1.797693e+308 xmax: 589000 ymax: 183040
epsg (SRID): 27700
proj4string: +proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +towgs84=446.448,-125.157,542.06,0.15,0.247,0.842,-20.489 +units=m +no_defs
First 10 features:
berth_name geometry
1 Tower Stairs Tier POINT (533225 180490)
2 St Katharine's Dock POINT (533875 180325)
3 President Quay POINT (533970 180200)
4 Express Wharf POINT (537000 179700)
5 West India Lock POINT (538320 179920)
6 Pura Foods POINT (539060 180660)
7 Orchard Wharf POINT (539230 180710)
8 Thames Wharf POINT (539600 180500)
9 Royal Primrose Works POINT (540200 179850)
10 Silvertown Wharf POINT (540440 179720)
The small_grid_result was grouped by large_grid_id to calculate the total number of GPS points in the parent grids. Then the count of GPS points in each 20m by 20m grid, was divided by the total of GPS points within the larger grid (1km by 1km) for that area, to give a weighting for each small grid to draw down the emissions. The grid cells containing berths were calculated in a similar manner but indenpendantly i.e. only cells that contained berths were divided by the berth_count. This created the contribution column.
Simple feature collection with 126714 features and 8 fields
geometry type: POLYGON
dimension: XY
bbox: xmin: 523860 ymin: 174420 xmax: 558000 ymax: 186900
epsg (SRID): 27700
proj4string: +proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +towgs84=446.448,-125.157,542.06,0.15,0.247,0.842,-20.489 +units=m +no_defs
First 10 features:
large_grid_id small_grid_id group count berth_name sailing_count berth_count contribution geometry
1 9717 2928 1 2 <NA> 9 NA 0.222222222 POLYGON ((533280 181060, 53...
2 9717 2981 1 3 <NA> 9 NA 0.333333333 POLYGON ((533300 181080, 53...
3 9717 3085 1 2 <NA> 9 NA 0.222222222 POLYGON ((533300 181120, 53...
4 9717 3397 1 1 <NA> 9 NA 0.111111111 POLYGON ((533300 181240, 53...
5 9717 3642 1 1 <NA> 9 NA 0.111111111 POLYGON ((533000 181360, 53...
6 9728 5496 1 2 <NA> 719 NA 0.002781641 POLYGON ((544700 181000, 54...
7 9728 5497 1 5 <NA> 719 NA 0.006954103 POLYGON ((544720 181000, 54...
8 9728 5498 1 2 <NA> 719 NA 0.002781641 POLYGON ((544740 181000, 54...
9 9728 5499 1 5 <NA> 719 NA 0.006954103 POLYGON ((544760 181000, 54...
10 9728 5500 1 4 <NA> 719 NA 0.005563282 POLYGON ((544780 181000, 54...The small_grid was now duplicated 3 times, a pollutant column added to each and populated (PM2.5, NOx and PM), and then joined back together again. It was then joined to the emissions by pollutant, group and unique_geom_id. Subsequently the contribution was multiplied by the emission for that large grid square, and stored as emissions.
There are 1km by 1km emission grid areas from the PLA, with emissions for sailing/berth, where there is no GPS data.
small_grid_result %>%
as_tibble() %>%
group_by(large_grid_id, pollutant, group) %>%
summarise(gps_count = sum(count, na.rm=T)) %>%
left_join(grid_emissions, ., by = c("large_grid_id" = "large_grid_id",
"group"="group",
"pollutant"="pollutant")) %>%
filter(gps_count == 0 & (sailing > 0 | berth > 0))
# A tibble: 62 x 6
pollutant group large_grid_id sailing berth gps_count
<chr> <dbl> <dbl> <dbl> <dbl> <int>
1 NOx 1 9559 0.17 0 0
2 NOx 1 9893 0.02 0 0
3 NOx 1 10081 0.01 0 0
4 NOx 2 9383 0.16 0 0
5 NOx 2 9384 0.900 0 0
6 NOx 2 9556 1.55 0 0
7 NOx 2 9557 5.76 0.48 0
8 NOx 2 9559 5.17 12.9 0
9 NOx 2 9560 0.64 0 0
10 NOx 2 9717 0.01 0 0
# ... with 52 more rowsHowever the emissions in these large_grid_squares are a very small proportion of the total emissions (shown below) so we have decided not to investigate this issue further.
| Pollutant | Sailing | Berth |
|---|---|---|
| NOx | 0.003% | 0.006% |
| PM | 0.002% | 0.006% |
| PM2.5 | 0.002% | 0.006% |
Due to the way that the emissions were calculated on a grid, there are some artefacts when visualising the data, such as shown below. These have not been resolved (as there is no easy way to do so). It is a limitation of the underlying data/method before our work.
