diff --git a/Cargo.lock b/Cargo.lock
index 2c34376..6877702 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -2656,6 +2656,7 @@ dependencies = [
  "num-bigint",
  "paste",
  "seq-macro",
+ "simdutf8",
  "snap",
  "thrift",
  "twox-hash 2.1.0",
@@ -3507,6 +3508,7 @@ dependencies = [
  "duckdb",
  "geoarrow-array 0.1.0-dev",
  "geojson",
+ "parquet",
  "pyo3",
  "pyo3-arrow",
  "pyo3-async-runtimes",
diff --git a/Cargo.toml b/Cargo.toml
index 05f7a72..82917e6 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -43,6 +43,7 @@ tokio = { version = "1.44.0", features = ["rt-multi-thread"] }
 pyo3-log = "0.12.1"
 tracing = "0.1.41"
 pyo3-object_store = "0.2.0"
+parquet = "55.1.0"
 
 [build-dependencies]
 cargo-lock = "10"
diff --git a/python/rustac/rustac.pyi b/python/rustac/rustac.pyi
index d5bec1a..65f5fb3 100644
--- a/python/rustac/rustac.pyi
+++ b/python/rustac/rustac.pyi
@@ -350,6 +350,7 @@ async def search_to(
     filter: str | dict[str, Any] | None = None,
     query: dict[str, Any] | None = None,
     format: str | None = None,
+    parquet_compression: str | None = None,
     store: AnyObjectStore | None = None,
     use_duckdb: bool | None = None,
 ) -> int:
@@ -389,6 +390,10 @@ async def search_to(
             It is recommended to use filter instead, if possible.
         format: The output format. If none, will be inferred from
             the outfile extension, and if that fails will fall back to compact JSON.
+        parquet_compression: If writing stac-geoparquet, sets the compression
+            algorithm.
+            https://docs.rs/parquet/latest/parquet/basic/enum.Compression.html
+            is a list of what's available.
         store: An optional [ObjectStore][]
         use_duckdb: Query with DuckDB. If None and the href has a
             'parquet' or 'geoparquet' extension, will be set to True. Defaults
@@ -428,6 +433,7 @@ async def write(
     value: dict[str, Any] | Sequence[dict[str, Any]],
     *,
     format: str | None = None,
+    parquet_compression: str | None = None,
     store: AnyObjectStore | None = None,
 ) -> dict[str, str] | None:
     """
@@ -439,6 +445,10 @@ async def write(
             can be a STAC dictionary or a list of items.
         format: The output format to write. If not provided, will be
             inferred from the href's extension.
+        parquet_compression: If writing stac-geoparquet, sets the compression
+            algorithm.
+            https://docs.rs/parquet/latest/parquet/basic/enum.Compression.html
+            is a list of what's available.
         store: The object store to use for writing.
 
     Returns:
diff --git a/src/error.rs b/src/error.rs
index aa44e2e..58595be 100644
--- a/src/error.rs
+++ b/src/error.rs
@@ -41,6 +41,9 @@ pub enum Error {
     #[error(transparent)]
     StacDuckdb(#[from] stac_duckdb::Error),
 
+    #[error(transparent)]
+    Parquet(#[from] parquet::errors::ParquetError),
+
     #[error(transparent)]
     TokioTaskJon(#[from] tokio::task::JoinError),
 }
diff --git a/src/search.rs b/src/search.rs
index d7dad03..a8f0723 100644
--- a/src/search.rs
+++ b/src/search.rs
@@ -58,7 +58,7 @@ pub fn search<'py>(
 }
 
 #[pyfunction]
-#[pyo3(signature = (outfile, href, *, intersects=None, ids=None, collections=None, max_items=None, limit=None, bbox=None, datetime=None, include=None, exclude=None, sortby=None, filter=None, query=None, format=None, store=None, use_duckdb=None, **kwargs))]
+#[pyo3(signature = (outfile, href, *, intersects=None, ids=None, collections=None, max_items=None, limit=None, bbox=None, datetime=None, include=None, exclude=None, sortby=None, filter=None, query=None, format=None, parquet_compression=None, store=None, use_duckdb=None, **kwargs))]
 #[allow(clippy::too_many_arguments)]
 pub fn search_to<'py>(
     py: Python<'py>,
@@ -77,6 +77,7 @@ pub fn search_to<'py>(
     filter: Option<StringOrDict>,
     query: Option<Bound<'py, PyDict>>,
     format: Option<String>,
+    parquet_compression: Option<String>,
     store: Option<AnyObjectStore>,
     use_duckdb: Option<bool>,
     kwargs: Option<Bound<'_, PyDict>>,
@@ -95,12 +96,18 @@ pub fn search_to<'py>(
         query,
         kwargs,
     )?;
-    let format = format
+    let mut format = format
         .map(|s| s.parse())
         .transpose()
         .map_err(Error::from)?
         .or_else(|| Format::infer_from_href(&outfile))
         .unwrap_or_default();
+    if matches!(format, Format::Geoparquet(_)) {
+        if let Some(parquet_compression) = parquet_compression {
+            tracing::debug!("setting parquet compression: {parquet_compression}");
+            format = Format::Geoparquet(Some(parquet_compression.parse().map_err(Error::from)?));
+        }
+    }
     if use_duckdb
         .unwrap_or_else(|| matches!(Format::infer_from_href(&href), Some(Format::Geoparquet(_))))
     {
diff --git a/src/write.rs b/src/write.rs
index 2923257..9db293c 100644
--- a/src/write.rs
+++ b/src/write.rs
@@ -6,12 +6,13 @@ use stac::{Item, ItemCollection};
 use stac_io::{Format, StacStore};
 
 #[pyfunction]
-#[pyo3(signature = (href, value, *, format=None, store=None))]
+#[pyo3(signature = (href, value, *, format=None, parquet_compression=None, store=None))]
 pub fn write<'py>(
     py: Python<'py>,
     href: String,
     value: Bound<'_, PyAny>,
     format: Option<String>,
+    parquet_compression: Option<String>,
     store: Option<AnyObjectStore>,
 ) -> PyResult<Bound<'py, PyAny>> {
     let value: Value = pythonize::depythonize(&value)?;
@@ -25,10 +26,17 @@ pub fn write<'py>(
         serde_json::from_value(value).map_err(Error::from)?
     };
     pyo3_async_runtimes::tokio::future_into_py(py, async move {
-        let format = format
+        let mut format = format
             .and_then(|f| f.parse::<Format>().ok())
             .or_else(|| Format::infer_from_href(&href))
             .unwrap_or_default();
+        if matches!(format, Format::Geoparquet(_)) {
+            if let Some(parquet_compression) = parquet_compression {
+                tracing::debug!("setting parquet compression: {parquet_compression}");
+                format =
+                    Format::Geoparquet(Some(parquet_compression.parse().map_err(Error::from)?));
+            }
+        }
         let (stac_store, path) = if let Some(store) = store {
             (StacStore::from(store.into_dyn()), None)
         } else {
diff --git a/tests/test_write.py b/tests/test_write.py
index eedb0af..70c2c25 100644
--- a/tests/test_write.py
+++ b/tests/test_write.py
@@ -5,6 +5,7 @@
 import pyarrow.parquet
 import rustac
 import stac_geoparquet
+from pyarrow.parquet import ParquetFile
 from rustac.store import LocalStore
 
 
@@ -35,3 +36,14 @@ async def test_write_includes_type(tmp_path: Path, item: dict[str, Any]) -> None
     await rustac.write(str(tmp_path / "out.parquet"), [item])
     data_frame = pandas.read_parquet(str(tmp_path / "out.parquet"))
     assert "type" in data_frame.columns
+
+
+async def test_write_parquet_compression(tmp_path: Path, item: dict[str, Any]) -> None:
+    await rustac.write(
+        str(tmp_path / "out.parquet"), [item], parquet_compression="zstd(1)"
+    )
+    parquet_file = ParquetFile(tmp_path / "out.parquet")
+    metadata = parquet_file.metadata
+    for row_group in range(metadata.num_row_groups):
+        for column in range(metadata.num_columns):
+            assert metadata.row_group(row_group).column(column).compression == "ZSTD"