Area under cane and ethanol’s share of what filling stations sell, across São Paulo’s municipalities

Português · English

Plate No. 111, the first of five in the collection Bioenergy, crossed; the next, plate 112.

The collection crosses São Paulo’s bioenergy with what filling stations sell, with the countryside and with town halls, across the state’s 645 municipalities. The pairs came from a correlation screen between CP2B biomethane potential, fuel sales in the SEMIL Yearbook, plantation forestry, the cattle herd, the Agricultural Census, the vehicle fleet and MUNIC. Each pair went through raw and partial rank correlation (net of urbanisation, density and income), an effective sample corrected for spatial autocorrelation, Benjamini–Hochberg q values and bivariate LISA. The last plate is descriptive, with no test, and the text says so.

The map’s callouts point to Ibaté, where ethanol is 69% of the energy sold in the Otto cycle (the highest among municipalities of 10,000 or more) and cane covers 51% of the area; São Luiz do Paraitinga, with no cane and 16% (the lowest); and São Paulo, at 39%, just below the state.

What the files contain

Layer (GPKG)	Geometry	Content
municipios	polygon	645 municipalities with cane and ethanol, biomethane and diesel, the fleet, native cover and pulpwood, cattle and organic fertiliser, distances to the pipeline, piped gas and the MUNIC answers on waste; classes b111, b113, b114, arranjo115 and clusters lisa111, lisa113, lisa114
gasodutos	line	transmission pipelines and distribution network (tipo)
regioes_intermediarias, limite_estadual	polygon	reference

The GPKG is the same for all five plates and carries a QGIS style for each. The CSV has the municipal table, without geometry. Empty fields are missing data, not zero. LISA clusters come as AA (high-high), BB (low-low), AB, BA and ns (not significant); the first letter is the municipality’s variable, the second its neighbours’.

Method

1. pct_area_cana: area under sugar cane (MapBiomas, collection 10, 2024) over municipal area, in %.
2. etanol_l and gasolina_l: litres of hydrous ethanol and petrol C sold in the municipality in 2024 (SEMIL/SP, Municipal Energy Yearbook 2025). pct_etanol_otto = ethanol × 21.3 MJ/L ÷ (ethanol × 21.3 + petrol C × 32.2 MJ/L), in %; petrol C already carries blended anhydrous ethanol, which stays in the petrol share.
3. Thirds of cane: up to 3.4%, 3.4% to 35.6% and over 35.6%; of ethanol: up to 39.8%, 39.8% to 52.5% and over 52.5%. b111: bivariate class, tens = third of cane (1 to 3), units = third of ethanol (1 to 3). From bottom to top and left to right, the nine cells hold 169, 39, 6; 38, 102, 74; 7, 73, 135 municipalities. Anhumas and Redenção da Serra have no fuel sales in the Yearbook and are left out (643 municipalities).
4. Cane × ethanol in the Otto cycle: Spearman rank correlation 0.70 (643 municipalities), partial 0.59, effective sample 310, p < 0.001, q < 0.001. Without the capital, 0.59; outside the São Paulo intermediate region, partial 0.56.
5. lisa111: bivariate LISA clusters (cane in the municipality × ethanol in its neighbours): 179 high-high, 176 low-low, 2 high-low and 8 low-high; bivariate Moran’s I 0.68.
6. Partial correlation: the correlation between the residuals of each variable’s ranks regressed on the ranks of the controls, which are urbanisation (residents in urban areas), log of population density and log of income per head (2022 Census). Effective sample under spatial autocorrelation: n·(1 − Iₓ·Iᵧ)/(1 + Iₓ·Iᵧ), where Iₓ and Iᵧ are the Moran’s I of the residuals over a 6-nearest-neighbour neighbourhood; Benjamini–Hochberg q values across all pairs in the screen. Bivariate LISA: the municipality’s value against the mean of its 6 neighbours on the other variable, 999 permutations, seed 20261019, p < 0.05.

Limitations

• Sales are not consumption: the SEMIL Yearbook counts what filling stations and distributors sell in the municipality, including to people just driving through or to fleets based elsewhere. Towns on busy highways may sell far more diesel than they use.
• Relative pump prices, the flex-fuel fleet and distance to the nearest mill may lie behind the link; the map does not separate these effects. Ethanol’s share is measured in energy, not litres: in litres it would be larger.
• Ecological reading: the map compares municipalities, not people or farms. A correlation between municipalities does not say that the same farms or drivers have both traits, nor that one causes the other.