Estimate the destinies of crop residues.

Splits crop residue dry matter into four destinies that sum to the total residue: fed to livestock, used as livestock bedding, burned / removed for fuel, and left on the field for soil incorporation.

Usage

calculate_residue_destinies(
  x,
  method = c("recovery_regional", "shares"),
  bedding_fraction = 0,
  unmatched_recovery = c("report", "abort"),
  recovery = c("wirsenius", "legacy"),
  feed_share = c("wirsenius", "legacy"),
  bedding = c("wirsenius", "fraction")
)

Arguments

x: A tibble with item_prod_code and residue_dm_t. The recovery_regional method also needs region_krausmann (for the recovery rate) and region_un_sub (for the feed-use fraction, the UN M49 sub-regions of regions_full$region_UN_sub). region_krausmann can use the recovery-table labels or the matching regions_full labels. The shares method needs year.
method: Destiny method: "recovery_regional" (default, the regional recovery rate times the UN-sub-regional feed-use fraction) or "shares" (the Spain-specific per-crop-year use/burn shares, flagged to_be_revised).
bedding_fraction: Fraction of the recovered non-feed residue used as livestock bedding under bedding = "fraction", one number in [0, 1]. Default 0. Must stay 0 under bedding = "wirsenius", which sets its own share.
unmatched_recovery: What the recovery_regional method does with a row that reaches no recovery rate at all, because its crop carries no Krausmann category or its region label reaches no recovery region: "report" (default) keeps the historical all-to-soil treatment and warns with the row count and tonnage, "abort" refuses to continue. Ignored by the "shares" method.
recovery: Which recovery-rate variant the recovery_regional method reads: "wirsenius" (default since whep#1330, every rate Wirsenius 2000 states, at the value it states) or "legacy" (the table as shipped before whep#1163). See the Two recovery variants section. Ignored by the "shares" method.
feed_share: Which feed-use fraction the recovery_regional method reads: "wirsenius" (default since whep#1398, Wirsenius 2000 Table 3.20, per crop by-product and Wirsenius region, the feed half of the recovery rates) or "legacy" (residue_feed_fraction.csv, one fraction per UN M49 sub-region for every crop). See the Two feed-share variants section. Ignored by the "shares" method.
bedding: Bedding rule: "wirsenius" (default, 12.4% of the recovered cereal straw, Wirsenius 2000 Table 3.21) or "fraction" (bedding_fraction of the recovered non-feed residue of every crop). See the Where the bedding share comes from section.

Value

The input tibble with residue_feed_dm_t, residue_bedding_dm_t, residue_burn_dm_t, residue_soil_dm_t, residue_bedding_fraction (the share applied: of the recovered residue under "wirsenius", zero for a non-cereal; of the recovered non-feed residue under "fraction"), method_residue_bedding, method_residue_destiny, and method_residue_recovery (the recovery variant, NA for the "shares" method) and method_residue_feed (per row, the table that priced its feed share: "wirsenius" or "legacy", NA for the "shares" method). The "recovery_regional" method also returns residue_recovery_matched, FALSE where no recovery rate was found, which is what separates a rate the table gives as zero from a zero standing in for a failed lookup.

Bedding

Bedding straw leaves the field with the rest of the recovered residue and comes back to the soil later, through the yard, as part of the managed manure. It is therefore carved out of the recovered non-feed residue – the mass the commodity balance books as other_uses – and never out of residue_soil_dm_t, which is the residue that stays on the field. That is the split IPCC 2019 Refinement Vol. 4 Ch. 10 p. 10.95 asks for when it tells inventory compilers to cross-check bedding nitrogen "relative to the amount of agricultural residues that is removed for other purposes (i.e. bedding) other than the amount of agricultural residues returned to soils or burnt", so as "to eliminate the possibility of double counting".

bedding = "wirsenius" (default) reads the one bedding figure that comes from the same accounting as the recovery rates: Wirsenius (2000), Human Use of Land and Organic Materials, PhD thesis, Chalmers University of Technology, Table 3.21 (p. 126). Of 2700 Tg DM/yr of cereals straw & stover generated, 530 Tg is "not recovered", so 2170 Tg is recovered, and 270 Tg is "used as litter for animal bedding" – 14% of the 1940 Tg distributed. This function's denominator is the recovered residue, so the share is taken against the table's own recovered total, 270 / 2170 = 0.124. That is the same distributed -> recovered conversion whep#1431 makes for the Table 3.20 feed shares; here the world table carries both totals, so no loss factor has to be assumed (the 10% loss route, 0.14 * 0.90, gives 0.126).

Wirsenius models cereals straw and stover as the only bedding material (p. 85, after Section 2.3.1), so the share applies to the five cereal categories (Wheat, other cereals, Rice, Paddy, Maize, Sorghum, Millet) and every other crop beds nothing. It is one world share applied in every region: the thesis prints no regional litter total, and its regional litter-use rates (Table 3.13, p. 86, kg straw per kg housed manure) are "very rough figures" (p. 85), with South & Central Asia cattle set to zero because "no reasonable values on litter use could be identified" – not an estimate to inherit. The bedding is carved out of the recovered non-feed residue and capped at it; where the cap binds a warning says how much.

bedding = "fraction" keeps the caller-set rule: bedding_fraction of the recovered non-feed residue of every crop, default 0 (no bedding). Two other anchors are on that denominator family, neither of them global:

Statistics Denmark HALM/HALM1/HALM2 – the only official statistic with a bedding-only column: 16-21% of straw production, about 30% of removed straw.
Bentsen, Felby & Thorsen (2014), Prog. Energy Combust. Sci. 40:59-73, Table 5 – Denmark, barley 16% and wheat 11% of production.

FAO GLEAM's FracRemove and IPCC 2019 Eq. 11.6's FracRemove both merge bedding with feed and construction into one term, so neither is usable. The rule used is recorded in method_residue_bedding.

Where the recovery rates come from

The recovery_regional recovery rates live in inst/extdata/coefs/residue_recovery.csv. Its numeric columns are sourced separately and carry a provenance column each (source_ratio, source_recovery and source_recovery_wirsenius), because they agree with the source to different degrees. All are Wirsenius (2000), Human Use of Land and Organic Materials, PhD thesis, Chalmers University of Technology – Table 3.17 (recovery rates, p. 94) and Table 3.16 (harvest index, p. 92, from which residue_dm_product_dm is the residue:product ratio rounded to one decimal). The file was named after Krausmann and its two key columns still are, but no coefficient in it is Krausmann's. Only the crop category (items_prod_full$Cat_Krausmann) really is his: region_krausmann here holds regions_full$region_HANPP labels, whose eight values are Wirsenius's eight region names (whep#1132), though not his country membership (see the Two feed-share variants section).

Table 3.17 states its rates per crop category, not per crop: one "Cereals straw & stover" row governs every cereal, and one "Sugar crops tops & leaves" row governs both crops Table 3.16 files under sugar crops, cane and beet. Read that way it governs twelve of the twenty categories here, and source_recovery splits them (whep#1150):

nine match it cell for cell;
three sit below it – Groundnuts in Shell, Sugar Beets and Sugar Crops nes, each against 0.90 in every region;
three – roots and tubers, cassava and oil palm – are residues the thesis does model (Table 2.6, pp. 48-49: cassava leaves and tops, potato tops, oil palm leaves and trunks) but Table 3.17 does not list, and for those Wirsenius states that recovery rates "were assumed to be close to 100 percent" (p. 94);
five – dry beans, pulses, castor beans, permanent crops and fodder crops – have no residue flow in the thesis at all: Table 2.6 gives pulses, fruits, tree nuts, vegetables and stimulants "no representation of by-products" (p. 47), models forage crops whole, and has no castor. The p. 94 default does not reach them, so the source is silent on them (whep#1163).

Two recovery variants

recovery = selects the rate column, and method_residue_recovery records which one was used:

"wirsenius" (default) reads recovery_rates_wirsenius: every rate the thesis states, at the value it states. The three below-source categories take 0.90, the three p. 94 categories take 1.00 – "close to 100 percent" read as 1.00, which is a reading of the text and not a number it prints – and the five categories the source is silent on keep the legacy assumed rate. source_recovery_wirsenius labels each.
"legacy" reads recovery_rates, the table as shipped before whep#1163, whose departures from the source are all downward.

"wirsenius" is the only variant in which every rate is traceable to the cited source, and it is the default since whep#1330. Its switch waited on the gross residue base these rates multiply, once thought to be about 36% too high (step 2 of whep#1132). whep#1330 found that excess gone in dry matter, but it measured the crop_residues pin as it was then read, and that pin holds residue already multiplied by the "legacy" rate (whep#1195). Read as the residue produced, world cereal residue in dry matter was 16.1% above the three-method mean of Smerald, Rahimi & Scheer (2023), doi:10.1038/s41597-023-02587-0 , over 1997–2021. Since whep#1448 get_primary_residues() estimates cereal residue from yield, and the 1997–2021 mean is 0.8% above theirs, inside their band in every year (validation/residue_base_dm.R). The rates the switch moves are all non-cereal, and no published global total for the non-cereal residue base was found to check it against.

Both rate columns are keyed on Wirsenius's eight regions, so a row takes his own region membership (Table 3.1, p. 58), not the HANPP label's: from its HANPP region and UN M49 sub-region, through residue_feed_regions.csv. The HANPP label files Southeast Asia, Russia, Belarus and the Caucasus under South & Central Asia, whose rates are the highest (whep#1430). On the commodity balance that lowers residue production by 2.4% in 2010 and 2.7% in 2020.

Measured on get_primary_residues()'s fresh value (the commodity balance's basis), "wirsenius" against "legacy" in 2010 raises recovered residue from 6939 to 7148 Mt fresh matter (+3.0%), the feed destiny from 1964 to 2025 Mt (+3.1%) and the burned/other-use destiny from 4976 to 5123 Mt (+3.0%), and lowers the soil destiny from 1789 to 1580 Mt (-11.7%); over 1961–1965 the same moves are +6.3%, +5.4%, +6.6% and -20.4%. Roots and tubers, cassava, sugar beet and groundnut are the only categories that move.

Three categories move no mass at all today, which is why the fodder rate of 0 is not the live problem it looks like: Sugar Crops nes and Oil Palm Fruit are in the table and in no production item, and no Fodder crops item reaches the residue pin.

feed_share = selects the fraction of the recovered residue booked as feed (the rest is the burned / other-use destiny), and method_residue_feed records, per row, which table priced it:

"wirsenius" (default) reads residue_feed_assignment.csv: Wirsenius (2000) Table 3.20 (p. 102), the "Assignm." rows for crop by-products, which the thesis coordinates directly with the Table 3.17 recovery rates – "assumptions on recovery rates were directly coordinated with those on assignment for use as feed" (p. 95). One row covers rice straw, one the other cereals (read for millet too, as Table 3.17 is), and one each cassava leaves, potato tops (read for all roots and tubers), sugar cane tops, sugar beet tops, groundnut stalks and the other oil crops. The seven categories it has no row for keep the legacy fraction, and their rows say "legacy".
"legacy" reads residue_feed_fraction.csv (Smil 1999, Lal 2005, Krausmann 2008, Erenstein 2014, McIntire 1992): one fraction per UN M49 sub-region, 0.05 to 0.45 around a 0.20 global default, the same for every crop. It is not paired with the recovery rates, and was the only feed share before whep#1398.

Table 3.20 is a share of the amount distributed, which is not the amount recovered: Wirsenius sets the distribution and storage losses of all crop by-products to 10% (p. 98), and in North America & Oceania leaves 5% of the recovered cereal and soybean straw in the field (p. 95, note 117). So the share of recovered residue fed is the printed value times 0.90 (0.855 for those North American straws): wheat straw in South & Central Asia is 0.80 of the distributed straw and 0.72 of the recovered. Read without the conversion it would book more feed than Wirsenius does.

Like the recovery rates, the table is keyed on Wirsenius's own region membership (Table 3.1, p. 58), through residue_feed_regions.csv. It matters more here: the HANPP label files Southeast Asia, Russia and Belarus, and the Caucasus under South and Central Asia, whose cereal feed share is 0.80, where Wirsenius has them at 0.30, 0.10 and 0.60.

Measured on get_primary_residues() in dry matter, the cereals' feed share of their generated residue is 0.224 (2010) and 0.225 (2020) under "legacy", 0.305 and 0.309 under "wirsenius". Wirsenius reports "some 33 percent of the amount generated" for the early 1990s (p. 177), and Smerald, Rahimi & Scheer (2023), Scientific Data 10:685, doi:10.1038/s41597-023-02587-0 , a share of 0.331 of the residue produced going to livestock. That figure is read from the authors' Zenodo notebook (whep#1132), not printed in the article, and their livestock category is "animal feed or bedding", so it is an upper bound for feed alone. The residue the commodity balance books as feed (Straw and Other crop residues) rises from 1089 to 1550 Tg of dry matter in 2010 (+42%) and from 1293 to 1856 Tg in 2020 (+44%); recovered and soil-returned residue do not move. In the 2010 national feed allocation (get_feed_intake()), residue intake rises from 1578 to 1985 Tg of dry matter and grass intake falls from 2703 to 2367 Tg (-12.4%), the residue displacing grazing. These are measured on the yield-based cereal residue base of whep#1448.

The default is "wirsenius" because it is the half of a matched pair whose other half, the recovery rates, is already the default, every value it reads is traceable to the source, and it lands within 8% of both published anchors above, where the legacy table falls about a third short of them.

Examples

calculate_residue_destinies(
  tibble::tibble(
    item_prod_code = "15", residue_dm_t = 100,
    region_krausmann = "Western Europe", region_un_sub = "Western Europe"
  )
)
#> # A tibble: 1 × 14
#>   item_prod_code residue_dm_t region_krausmann region_un_sub method_residue_feed
#>   <chr>                 <dbl> <chr>            <chr>         <chr>              
#> 1 15                      100 West Europe      Western Euro… wirsenius          
#> # ℹ 9 more variables: residue_recovery_matched <lgl>, residue_feed_dm_t <dbl>,
#> #   residue_burn_dm_t <dbl>, residue_soil_dm_t <dbl>,
#> #   residue_bedding_dm_t <dbl>, residue_bedding_fraction <dbl>,
#> #   method_residue_bedding <chr>, method_residue_destiny <chr>,
#> #   method_residue_recovery <chr>