MAPA PSR — Rural Insurance¶

License: CC-BY (Brazilian federal government public data). Classification: livre

Open data from SISSER/MAPA — the Rural Insurance Premium Subsidy System (Sistema de Subvencao Economica ao Premio do Seguro Rural). Policies and claims (indemnities) of Brazilian rural insurance with federal subsidy, published by the Ministry of Agriculture.

Production revision proxy: high soybean claims in Q1 precede cuts in the CONAB Q2 estimate.

Installation¶

Does not require optional dependencies. Uses only httpx + pandas (core).

API¶

from agrobr.alt import mapa_psr

# Rural insurance claims (indemnities paid)
df = await mapa_psr.sinistros()

# Filter by crop and state
df = await mapa_psr.sinistros(cultura="SOJA", uf="MT")

# Filter by year or range
df = await mapa_psr.sinistros(ano=2023)
df = await mapa_psr.sinistros(ano_inicio=2020, ano_fim=2024)

# Filter by predominant event
df = await mapa_psr.sinistros(evento="seca")

# Filter by municipality
df = await mapa_psr.sinistros(municipio="SORRISO")

# All policies (including those without a claim)
df = await mapa_psr.apolices()

# Filtered policies
df = await mapa_psr.apolices(cultura="MILHO", uf="PR", ano=2023)

# Synchronous API
from agrobr.sync import alt
df = alt.mapa_psr.sinistros(cultura="SOJA")
df = alt.mapa_psr.apolices(uf="MT")

Parameters — `sinistros`¶

Parameter	Type	Default	Description
`cultura`	str \| None	None	Filter by crop (partial match, accent-insensitive, e.g. "cafe" matches "CAFE ARABICA")
`uf`	str \| None	None	Filter by state (abbreviation, e.g. "MT")
`ano`	int \| None	None	Single-year filter (e.g. 2023)
`ano_inicio`	int \| None	None	Start year of the range (inclusive)
`ano_fim`	int \| None	None	End year of the range (inclusive)
`municipio`	str \| None	None	Filter by municipality (partial match)
`evento`	str \| None	None	Filter by predominant event (e.g. "seca")
`return_meta`	bool	False	Returns a (DataFrame, MetaInfo) tuple

Columns — `sinistros`¶

Column	Type	Nullable	Description
`nr_apolice`	str	No	Policy number
`ano_apolice`	int	No	Policy year
`uf`	str	No	State abbreviation of the property
`municipio`	str	Yes	Municipality name
`cd_ibge`	str	Yes	IBGE code of the municipality
`cultura`	str	No	Insured crop (uppercase)
`classificacao`	str	Yes	Product classification (AGRICOLA, PECUARIO, etc.)
`evento`	str	No	Predominant event (lowercase)
`area_total`	float	Yes	Total insured area (ha)
`valor_indenizacao`	float	No	Indemnity amount (R$) — always > 0
`valor_premio`	float	Yes	Net premium (R$)
`valor_subvencao`	float	Yes	Federal subsidy (R$)
`valor_limite_garantia`	float	Yes	Coverage limit (R$)
`produtividade_estimada`	float	Yes	Estimated yield
`produtividade_segurada`	float	Yes	Insured yield
`nivel_cobertura`	float	Yes	Coverage level (%)
`seguradora`	str	Yes	Insurer legal name

Parameters — `apolices`¶

Parameter	Type	Default	Description
`cultura`	str \| None	None	Filter by crop (partial match, accent-insensitive)
`uf`	str \| None	None	Filter by state
`ano`	int \| None	None	Single-year filter
`ano_inicio`	int \| None	None	Start year of the range
`ano_fim`	int \| None	None	End year of the range
`municipio`	str \| None	None	Filter by municipality
`return_meta`	bool	False	Returns a (DataFrame, MetaInfo) tuple

Columns — `apolices`¶

Same columns as sinistros, plus:

Column	Type	Nullable	Description
`taxa`	float	Yes	Premium rate (%)

Data pipeline¶

Resolves the required periods (2006-2015, 2016-2024, 2025) based on the year filters
Bulk CSV download from the dados.agricultura.gov.br portal (3 files, variable encoding)
Detects encoding (UTF-8 → UTF-8-sig → Windows-1252 → ISO-8859-1 → chardet) and separator (; or ,)
Removes PII columns (NM_SEGURADO, NR_DOCUMENTO_SEGURADO) and geolocation
Normalizes column names (standardized snake_case)
Converts types (float64 for monetary values, int for year)
For claims: filters VALOR_INDENIZACAO > 0 and non-empty EVENTO_PREPONDERANTE

MetaInfo¶

df, meta = await mapa_psr.sinistros(return_meta=True)
print(meta.source)           # "mapa_psr"
print(meta.source_method)    # "httpx"
print(meta.parser_version)   # 1
print(meta.records_count)    # varies by filter

Performance note¶

The SISSER CSVs can be large (the 2006-2015 file has ~500k rows). The module downloads only the periods required based on the year filters. Read timeout: 180 seconds.

Datasets¶

seguro_rural — wraps mapa_psr.apolices() and mapa_psr.sinistros() via tipo= dispatch

Source¶

URL: https://dados.agricultura.gov.br/dataset/sisser3
Format: CSV (3 files per period)
Update frequency: annual
History: 2006+
License: livre (CC-BY, federal government public data)