# Cattle and sheep, department by department **Answer.** Only **Salto and Artigas** still have more sheep than cattle by heads (53.5% and 51.9% sheep). The national flock fell **11.5%** in one year, from 5,420,661 to 4,799,066: 621,595 fewer sheep. Notebook: https://cow-tools.sophia-labs.com/notebooks/cattle-and-sheep/cattle-and-sheep.ipynb · Page: https://cow-tools.sophia-labs.com/notebooks/cattle-and-sheep/index.html · Data: Uruguay MGAP/SNIG DICOSE 2024–2025 (graph `uruguay-cows` at https://cow-tools.sophia-labs.com/mcp). ## Findings - Only **Salto and Artigas** still have more sheep than cattle by heads (53.5% and 51.9% sheep). The national flock fell **11.5%** in one year, from 5,420,661 to 4,799,066: 621,595 fewer sheep. - Every one of the 19 departments declared fewer sheep in 2025. The steepest fall was in **Río Negro** (-18.3%); in absolute heads, Salto lost the most (118,943). ## The notebook Uruguay is a cattle country, but the north is still sheep country: on the basalt of Salto and Artigas there are more sheep than cattle. Wool built the old estancias; the national flock has been shrinking for decades. This notebook asks how the two herds split across the 19 departments, and what one year (2024 → 2025) did to the sheep. A note on comparing a sheep with a cow: we compare **heads**, the unit DICOSE counts. In livestock units a sheep weighs far less (DICOSE's own codelist weights a breeding ewe at 0.17 of a unit, a breeding cow at 1.03), so a department that is "half sheep" by heads is still mostly cattle by grass eaten. ```python # Setup. import altair as alt import pandas as pd import cowtools_client as cw import cowtools_charts as cc ``` ```python # The query: cattle (beef + dairy) and sheep per department, both years. q = cw.sparql(""" SELECT ?department ?year ?kind (SUM(?n) AS ?animals) WHERE { GRAPH ?g { ?obs a cow:AnimalFlowObservation ; cow:year ?year ; cow:department ?department ; cow:species ?species ; cow:animalsCount ?n . VALUES (?species ?kind) { ( "cattle") ( "cattle") ( "sheep") } } } GROUP BY ?department ?year ?kind """) q["code"] = q["department"].map(cw.dept_code) t = q.pivot_table(index="code", columns=["kind", "year"], values="animals", aggfunc="sum") t.columns = [f"{k}_{y}" for k, y in t.columns] t = t.reset_index().merge(cw.departments()[["code", "name"]], on="code") for y in (2024, 2025): t[f"sheep_share_{y}"] = t[f"sheep_{y}"] / (t[f"sheep_{y}"] + t[f"cattle_{y}"]) t["sheep_change"] = t["sheep_2025"] / t["sheep_2024"] - 1 t["cattle_change"] = t["cattle_2025"] / t["cattle_2024"] - 1 t = t.sort_values("sheep_share_2025", ascending=False) cc.show_table(t[["name", "cattle_2025", "sheep_2025", "sheep_share_2025", "sheep_change"]] .rename(columns={"sheep_share_2025": "sheep share of heads", "sheep_change": "sheep 2024→25"}), floatfmt=".3f", name="by-department") ``` | name | cattle_2025 | sheep_2025 | sheep share of heads | sheep 2024→25 | |:---------------|:--------------|:-------------|-----------------------:|----------------:| | Salto | 898,899 | 1,032,181 | 0.535 | -0.103 | | Artigas | 742,198 | 801,403 | 0.519 | -0.098 | | Montevideo | 2,736 | 2,369 | 0.464 | -0.004 | | Paysandú | 801,512 | 555,079 | 0.409 | -0.149 | | Maldonado | 234,191 | 101,340 | 0.302 | -0.061 | | Lavalleja | 654,451 | 261,086 | 0.285 | -0.099 | | Tacuarembó | 1,061,136 | 398,742 | 0.273 | -0.122 | | Cerro Largo | 950,412 | 303,495 | 0.242 | -0.129 | | Durazno | 952,171 | 290,920 | 0.234 | -0.131 | | Treinta y Tres | 584,734 | 178,572 | 0.234 | -0.142 | | Rivera | 639,917 | 182,585 | 0.222 | -0.094 | | Rocha | 682,644 | 154,813 | 0.185 | -0.088 | | Flores | 409,429 | 83,001 | 0.169 | -0.13 | | Florida | 1,117,097 | 202,434 | 0.153 | -0.114 | | Canelones | 338,601 | 50,487 | 0.13 | -0.041 | | Río Negro | 549,369 | 69,144 | 0.112 | -0.183 | | Soriano | 567,452 | 55,293 | 0.089 | -0.097 | | Colonia | 537,066 | 38,341 | 0.067 | -0.158 | | San José | 541,254 | 37,781 | 0.065 | -0.152 | ### Cross-check against the source Re-derive the 2025 department totals from `datosanimales.csv` and compare every department, both species, to the head. Then compare three shares with the figures the July 2026 expedition published (Salto 53.5% sheep, Artigas 51.9%, San José 6.5%). ```python csv = cw.source_csv(2025, "datosanimales.csv") csv["n"] = cw.num(csv["AnimalesPorEspeciePD_AD"]) csv["kind"] = csv["EspecieCodigo"].map({"1": "cattle", "11": "cattle", "2": "sheep"}) src = csv.dropna(subset=["kind"]).groupby(["DepartamentoCodigo", "kind"])["n"].sum().unstack() mismatches = 0 for _, row in t.iterrows(): for kind in ("cattle", "sheep"): if int(row[f"{kind}_2025"]) != int(src.loc[row["code"], kind]): mismatches += 1 cw.check("2025 cattle and sheep, all 19 departments: cube = CSV (mismatching cells)", mismatches, 0, source="datosanimales.csv 2025, AnimalesPorEspeciePD_AD") share = t.set_index("name")["sheep_share_2025"] cw.check("Salto sheep share of heads", round(share["Salto"], 3), 0.535, tol=0.0005, source="expedition 2026-07-05") cw.check("Artigas sheep share of heads", round(share["Artigas"], 3), 0.519, tol=0.0005, source="expedition 2026-07-05") cw.check("San José sheep share of heads", round(share["San José"], 3), 0.065, tol=0.0005, source="expedition 2026-07-05") ``` ```text PASS 2025 cattle and sheep, all 19 departments: cube = CSV (mismatching cells): 0 vs 0 (exactly) — datosanimales.csv 2025, AnimalesPorEspeciePD_AD PASS Salto sheep share of heads: 0.535 vs 0.535 (within ±0.0005) — expedition 2026-07-05 PASS Artigas sheep share of heads: 0.519 vs 0.519 (within ±0.0005) — expedition 2026-07-05 PASS San José sheep share of heads: 0.065 vs 0.065 (within ±0.0005) — expedition 2026-07-05 ``` ```python sheep_n = {y: int(t[f"sheep_{y}"].sum()) for y in (2024, 2025)} fall = sheep_n[2025] / sheep_n[2024] - 1 cw.check("national sheep 2024 → 2025", round(fall, 4), -0.1147, tol=0.00005, source="expedition 2026-07-05: −11.47%") majority = t.query("sheep_share_2025 > 0.5")["name"].tolist() cw.finding("headline", f"Only **{' and '.join(majority)}** still have more sheep than cattle by heads " f"({share['Salto']:.1%} and {share['Artigas']:.1%} sheep). The national flock fell " f"**{-fall:.1%}** in one year, from {sheep_n[2024]:,} to {sheep_n[2025]:,}: " f"{sheep_n[2024]-sheep_n[2025]:,} fewer sheep.") ``` ```text PASS national sheep 2024 → 2025: -0.1147 vs -0.1147 (within ±5e-05) — expedition 2026-07-05: −11.47% ``` > **Finding.** Only **Salto and Artigas** still have more sheep than cattle by heads (53.5% and 51.9% sheep). The national flock fell **11.5%** in one year, from 5,420,661 to 4,799,066: 621,595 fewer sheep. ## The map: where the sheep are Share of heads that are sheep, 2025. The north-west basalt (Artigas, Salto, Paysandú) is the old sheep country; the dairy south-west is nearly free of them. ```python cc.show(cc.choropleth(t, "sheep_share_2025", title="Sheep as a share of cattle + sheep, 2025", subtitle="heads on the holding (PD_AD)", legend="sheep share", fmt=".0%", domain=[0, 0.6]), "sheep-share-map-2025", thumbnail=True, alt_text="Map: sheep are over half the heads in Salto and Artigas; Montevideo (46%, a few thousand animals) and Paysandú (41%) follow; lowest in San José and Colonia.") print(cc.map_credit()) ``` ![Map: sheep are over half the heads in Salto and Artigas; Montevideo (46%, a few thousand animals) and Paysandú (41%) follow; lowest in San José and Colonia.](charts/sheep-share-map-2025.svg) Chart data and encoding: [charts/sheep-share-map-2025.vl.json](charts/sheep-share-map-2025.vl.json) (Vega-Lite). ```text Map: Natural Earth 1:10m admin-1 (public domain, v5.1.2), joined on the DICOSE department code. Made with Natural Earth. ``` ## The decline, department by department Every department lost sheep between the two declarations. The map is coloured on a diverging scale centred on zero, so a department that had gained would show blue; none did. ```python lost_all = bool((t["sheep_change"] < 0).all()) worst = t.sort_values("sheep_change").iloc[0] cw.check("every department lost sheep", lost_all, True, source="the table above") cc.show(cc.choropleth(t, "sheep_change", title="Change in sheep, 2024 → 2025", subtitle="per department; brown is a fall, blue a rise", legend="change", fmt="+.0%", diverging=True), "sheep-change-map", alt_text="Diverging map: every department is brown, a fall in sheep everywhere; the steepest is Río Negro.") ``` ```text PASS every department lost sheep: True vs True (exactly) — the table above ``` ![Diverging map: every department is brown, a fall in sheep everywhere; the steepest is Río Negro.](charts/sheep-change-map.svg) Chart data and encoding: [charts/sheep-change-map.vl.json](charts/sheep-change-map.vl.json) (Vega-Lite). ```python cw.finding("decline", f"Every one of the 19 departments declared fewer sheep in 2025. The steepest fall was in " f"**{worst['name']}** ({worst['sheep_change']:+.1%}); in absolute heads, " f"{t.assign(d=t.sheep_2024 - t.sheep_2025).sort_values('d').iloc[-1]['name']} lost the most " f"({int(t.assign(d=t.sheep_2024 - t.sheep_2025)['d'].max()):,}).") ``` > **Finding.** Every one of the 19 departments declared fewer sheep in 2025. The steepest fall was in **Río Negro** (-18.3%); in absolute heads, Salto lost the most (118,943). ## Small multiples: both herds, both years One panel per department, cattle and sheep side by side, 2024 and 2025. The panels share a scale, so the size of each department's herd is comparable. ```python long = pd.concat([ t[["name", f"{k}_{y}"]].rename(columns={f"{k}_{y}": "animals"}).assign(kind=k, year=str(y)) for k in ("cattle", "sheep") for y in (2024, 2025) ]) order = t.sort_values("cattle_2025", ascending=False)["name"].tolist() sm = ( alt.Chart(long) .mark_bar() .encode( x=alt.X("year:N", title=None, axis=alt.Axis(labelAngle=0)), xOffset="kind:N", y=alt.Y("animals:Q", title=None, axis=alt.Axis(format="~s", tickCount=3)), color=alt.Color("kind:N", scale=alt.Scale(domain=["cattle", "sheep"], range=[cc.CATTLE, cc.SHEEP]), title=None), tooltip=["name:N", "kind:N", "year:N", alt.Tooltip("animals:Q", format=",")], ) .properties(width=95, height=80) .facet(facet=alt.Facet("name:N", sort=order, title=None), columns=5) .properties(title=alt.Title("Cattle and sheep by department, 2024 and 2025", subtitle="heads on the holding; panels ordered by cattle, largest first")) ) cc.show(sm, "small-multiples", alt_text="Nineteen small panels, one per department, each with cattle and sheep bars for 2024 and 2025; sheep bars shrink everywhere.") ``` ![Nineteen small panels, one per department, each with cattle and sheep bars for 2024 and 2025; sheep bars shrink everywhere.](charts/small-multiples.svg) Chart data and encoding: [charts/small-multiples.vl.json](charts/small-multiples.vl.json) (Vega-Lite). ## Where every number came from ```python cc.show_table(cw.checks()[["check", "got", "expected", "rule", "source"]], name="checks") cw.provenance() ``` | check | got | expected | rule | source | |:--------------------------------------------------------------------------|--------:|-----------:|:---------------|:------------------------------------------------| | 2025 cattle and sheep, all 19 departments: cube = CSV (mismatching cells) | 0 | 0 | exactly | datosanimales.csv 2025, AnimalesPorEspeciePD_AD | | Salto sheep share of heads | 0.535 | 0.535 | within ±0.0005 | expedition 2026-07-05 | | Artigas sheep share of heads | 0.519 | 0.519 | within ±0.0005 | expedition 2026-07-05 | | San José sheep share of heads | 0.065 | 0.065 | within ±0.0005 | expedition 2026-07-05 | | national sheep 2024 → 2025 | -0.1147 | -0.1147 | within ±5e-05 | expedition 2026-07-05: −11.47% | | every department lost sheep | 1 | 1 | exactly | the table above | ```text kind target rows \ 0 sparql uruguay-cows 76 first line 0 SELECT ?department ?year ?kind (SUM(?n) AS ?an... ``` ```python cw.save() ```