🌍 TerraTiff

A lightweight, GDAL-free Python package for reading, writing, and exporting GeoTIFF raster files.

Built on tifffile + numpy + pyproj — no GDAL installation required.

---

Table of Contents

• Features
• Installation
• Tutorial
  • 1. Creating a GeoTIFF from a Numpy Array
  • 2. Reading an Existing GeoTIFF
  • 3. Choosing a Coordinate Reference System (CRS)
  • 4. Exporting with Different Data Types
  • 5. Working with Multi-Band Rasters
  • 6. Band Operations
  • 7. Binary Mask Export
  • 8. Changing Spatial Resolution (Resampling)
  • 9. Clipping by Extent
  • 10. Masking with a Polygon
  • 11. Masking with a Raster
  • 12. Converting Between Coordinate Systems
  • 13. Querying Spatial Metadata
  • 14. Working with NoData Values
  • 15. UTM Zone Utilities
• API Reference
• Supported CRS Formats
• Supported Data Types
• License

---

Features

Feature	Description
📖 Read	Load existing GeoTIFF files (single-band and multi-band)
💾 Write	Export georeferenced TIFF files with proper GeoKeys
🗺️ CRS	WGS 84, all 120 UTM zones (N/S), any EPSG code
🔢 Data Types	uint8/16/32, int8/16/32, float16/32/64, binary
📐 Resample	4 methods: nearest, bilinear, cubic, average
✂️ Clip	Crop rasters by bounding-box extent
🔷 Polygon Mask	Mask rasters using polygon geometries
🖼️ Raster Mask	Mask rasters using another raster file
🔄 CRS Convert	Reproject coordinates between WGS 84 ↔ UTM
🎚️ Multi-Band	RGB, multispectral, any number of bands
🧩 Array Export	Turn any numpy array into a georeferenced TIFF

---

Installation

pip install terratiff

Dependencies only

pip install numpy tifffile pyproj

Note: None of these dependencies require GDAL. pyproj uses the standalone PROJ library.

---

Tutorial

1. Creating a GeoTIFF from a Numpy Array

The most common use case: you have a numpy array (e.g. from a computation, a model output, or a CSV) and you want to save it as a georeferenced TIFF file.

import numpy as np
from terratiff import TerraTiff

# Create some data — for example, a 100×200 elevation grid
elevation = np.random.rand(100, 200).astype(np.float32) * 500  # 0–500m

# Wrap it in a TerraTiff with spatial metadata
raster = TerraTiff.from_array(
    elevation,
    origin_x=500000,       # easting of the top-left corner (meters)
    origin_y=4500000,      # northing of the top-left corner (meters)
    pixel_width=30,        # 30 m pixel size in X
    pixel_height=-30,      # -30 m in Y (negative = north-up)
    crs="UTM:33N",         # UTM zone 33 North
)

# Save to disk
raster.save("elevation.tif", dtype="float32")
print("Saved! ✅")

Key points:

• pixel_height should be negative for standard north-up rasters
• origin_x / origin_y are the coordinates of the top-left corner of the top-left pixel
• The array can be 2-D (single band) or 3-D (multi-band)
• By default, 3-D arrays are expected to be in (bands, rows, cols) format. If your array is (rows, cols, bands) (e.g. from an image library), pass first_band=False to automatically transpose it.

---

2. Reading an Existing GeoTIFF

from terratiff import TerraTiff

# Open and read a GeoTIFF file
# Use first_band=False if the TIFF is pixel-interleaved (rows, cols, bands)
raster = TerraTiff.open("elevation.tif", first_band=True)

# Inspect the metadata
print(f"Shape:        {raster.shape}")         # (bands, rows, cols)
print(f"Data type:    {raster.dtype}")          # e.g. float32
print(f"CRS:          {raster.crs}")            # e.g. EPSG:32633
print(f"Origin:       ({raster.origin_x}, {raster.origin_y})")
print(f"Pixel size:   ({raster.pixel_width}, {raster.pixel_height})")
print(f"Bands:        {raster.bands}")
print(f"Rows × Cols:  {raster.rows} × {raster.cols}")

# Access the raw numpy array
data = raster.data         # shape: (bands, rows, cols)
band1 = raster.data[0]    # first band as 2-D array

Output example:

Shape:        (1, 100, 200)
Data type:    float32
CRS:          EPSG:32633
Origin:       (500000.0, 4500000.0)
Pixel size:   (30.0, -30.0)
Bands:        1
Rows × Cols:  100 × 200

---

3. Choosing a Coordinate Reference System (CRS)

TerraTiff supports multiple ways to specify a CRS:

# WGS 84 (latitude / longitude) — for global geographic data
raster1 = TerraTiff.from_array(data,
    origin_x=-93.5, origin_y=42.0,         # lon, lat
    pixel_width=0.001, pixel_height=-0.001, # degrees
    crs="WGS84",
)

# UTM zone — for local projected data in meters
raster2 = TerraTiff.from_array(data,
    origin_x=500000, origin_y=4500000,     # easting, northing (meters)
    pixel_width=30, pixel_height=-30,       # meters
    crs="UTM:33N",                          # UTM zone 33, Northern hemisphere
)

# Southern hemisphere UTM
raster3 = TerraTiff.from_array(data,
    origin_x=300000, origin_y=6200000,
    pixel_width=10, pixel_height=-10,
    crs="UTM:55S",                          # UTM zone 55, Southern hemisphere
)

# Any EPSG code — as a string
raster4 = TerraTiff.from_array(data,
    origin_x=0, origin_y=0,
    pixel_width=1, pixel_height=-1,
    crs="EPSG:32635",
)

# Any EPSG code — as an integer
raster5 = TerraTiff.from_array(data,
    origin_x=0, origin_y=0,
    pixel_width=1, pixel_height=-1,
    crs=32635,
)

When to use which CRS:

CRS	Unit	Best for
"WGS84"	degrees	Global datasets, GPS coordinates
"UTM:ZoneN/S"	meters	Local/regional data, distance calculations
"EPSG:NNNNN"	varies	Any specific projection you need

---

4. Exporting with Different Data Types

Control the output data type when saving. This is useful for reducing file size or matching a required format.

import numpy as np
from terratiff import TerraTiff

# Create some floating-point data
data = np.random.rand(50, 50).astype(np.float64) * 1000

raster = TerraTiff.from_array(data,
    origin_x=0, origin_y=0,
    pixel_width=10, pixel_height=-10,
    crs="UTM:33N",
)

# === Integer types (truncates decimals) ===
raster.save("as_uint8.tif",  dtype="uint8")    # 0 – 255
raster.save("as_uint16.tif", dtype="uint16")   # 0 – 65,535
raster.save("as_uint32.tif", dtype="uint32")   # 0 – 4,294,967,295
raster.save("as_int8.tif",   dtype="int8")     # -128 – 127
raster.save("as_int16.tif",  dtype="int16")    # -32,768 – 32,767
raster.save("as_int32.tif",  dtype="int32")    # -2B – 2B

# === Floating-point types ===
raster.save("as_float16.tif", dtype="float16") # half precision
raster.save("as_float32.tif", dtype="float32") # single precision (recommended)
raster.save("as_float64.tif", dtype="float64") # double precision

# === Binary mask ===
raster.save("as_binary.tif",  dtype="binary")  # 0 or 1 only

# === Default — keeps original dtype ===
raster.save("as_default.tif")                  # float64 in this case

Choosing the right data type:

Type	Size/pixel	Use case
uint8	1 byte	RGB images, classification maps (≤ 255 classes)
int16	2 bytes	Elevation (DEM), temperature, signed integer data
float32	4 bytes	General-purpose scientific data (recommended)
float64	8 bytes	High-precision data (coordinates, large values)
binary	1 byte	Masks (land/water, cloud/clear, building/no-building)

---

5. Working with Multi-Band Rasters

Multi-band rasters store multiple layers in a single file — common for satellite imagery (RGB, multispectral).

Creating a multi-band raster

import numpy as np
from terratiff import TerraTiff

# 3-band RGB image (bands, rows, cols)
red   = np.random.randint(0, 255, (512, 512), dtype=np.uint8)
green = np.random.randint(0, 255, (512, 512), dtype=np.uint8)
blue  = np.random.randint(0, 255, (512, 512), dtype=np.uint8)

# Stack into (3, 512, 512)
rgb = np.stack([red, green, blue], axis=0)

raster = TerraTiff.from_array(
    rgb,
    origin_x=-93.5, origin_y=42.0,
    pixel_width=0.0001, pixel_height=-0.0001,
    crs="WGS84",
)

raster.save("rgb_image.tif", dtype="uint8")
print(f"Bands: {raster.bands}")  # 3

Creating a multispectral raster (7 bands)

# Simulate 7-band Landsat-like data
bands_data = np.random.rand(7, 256, 256).astype(np.float32)

raster = TerraTiff.from_array(
    bands_data,
    origin_x=500000, origin_y=4500000,
    pixel_width=30, pixel_height=-30,
    crs="UTM:33N",
)

raster.save("multispectral.tif", dtype="float32")
print(f"Shape: {raster.shape}")  # (7, 256, 256)

Reading a multi-band raster

raster = TerraTiff.open("multispectral.tif")
print(f"Number of bands: {raster.bands}")

# Access individual bands (0-indexed)
band1 = raster.get_band(0)  # first band, shape (256, 256)
band4 = raster.get_band(3)  # fourth band

# Or via the data array directly
all_bands = raster.data  # shape (7, 256, 256)

---

6. Band Operations

Add, remove, and manipulate individual bands.

import numpy as np
from terratiff import TerraTiff

# Start with a single-band raster
dem = np.random.rand(100, 100).astype(np.float32) * 500
raster = TerraTiff.from_array(dem,
    origin_x=0, origin_y=0,
    pixel_width=30, pixel_height=-30,
    crs="UTM:33N",
)
print(f"Bands: {raster.bands}")  # 1

# === Add a slope band ===
slope = np.gradient(dem)[0].astype(np.float32)
raster.add_band(slope)
print(f"Bands: {raster.bands}")  # 2

# === Add an aspect band ===
aspect = np.gradient(dem)[1].astype(np.float32)
raster.add_band(aspect)
print(f"Bands: {raster.bands}")  # 3

# === Retrieve specific bands ===
dem_band   = raster.get_band(0)  # shape (100, 100)
slope_band = raster.get_band(1)
aspect_band = raster.get_band(2)

# === Save the multi-band result ===
raster.save("dem_slope_aspect.tif", dtype="float32")

---

7. Binary Mask Export

Create and export binary (0/1) masks — useful for land-use classification, cloud masks, etc.

import numpy as np
from terratiff import TerraTiff

# Simulate an elevation array
elevation = np.random.rand(200, 300).astype(np.float32) * 2000  # 0–2000 m

# Create a binary mask: 1 where elevation > 1000m, 0 elsewhere
high_ground = (elevation > 1000).astype(np.uint8)

raster = TerraTiff.from_array(
    high_ground,
    origin_x=500000, origin_y=4500000,
    pixel_width=30, pixel_height=-30,
    crs="UTM:33N",
)

# dtype="binary" ensures output contains only 0 and 1
raster.save("high_ground_mask.tif", dtype="binary")

# Verify
loaded = TerraTiff.open("high_ground_mask.tif")
unique_values = np.unique(loaded.data)
print(f"Unique values: {unique_values}")  # [0 1]

How "binary" works: any non-zero value in the array is mapped to 1, and zeros stay 0. The output dtype is uint8.

---

8. Changing Spatial Resolution (Resampling)

Resample a raster to a coarser or finer pixel size. Choose from 4 interpolation methods — the geographic extent is preserved; only the grid dimensions change.

Resampling methods

Method	Description	Best for
"nearest"	Nearest-neighbour — fast, no blending	Categorical data, masks, classification maps
`"bilinear"$	2 \times 2 \text{weighted} \text{average} — \text{smooth} \text{transitions}	\text{Continuous} \text{surfaces} (\text{elevation}, \text{temperature})
$"cubic"$	4 \times 4 \text{Catmull}-\text{Rom} — \text{sharp}, \text{high} \text{quality}	\text{Photographic} \text{imagery}, \text{high}-\text{detail} \text{surfaces}
$"average"`	Block-mean aggregation — anti-aliased	Downsampling any data type

Basic usage

from terratiff import TerraTiff

raster = TerraTiff.open("elevation.tif")
print(f"Original: {raster.shape}, pixel: {raster.pixel_width}m")

# === Downsample to 90m with nearest (default) ===
coarse = raster.resample(pixel_width=90, pixel_height=-90)
coarse.save("elevation_90m.tif", dtype="float32")

# === Upsample to 10m with bilinear (smooth) ===
fine = raster.resample(pixel_width=10, pixel_height=-10, method="bilinear")
fine.save("elevation_10m.tif", dtype="float32")

Comparing methods

import numpy as np
from terratiff import TerraTiff

# Create a gradient surface
arr = np.linspace(0, 100, 100 * 100).reshape(100, 100).astype(np.float32)
raster = TerraTiff.from_array(
    arr, origin_x=0, origin_y=0,
    pixel_width=10, pixel_height=-10, crs="UTM:33N",
)

# Downsample with each method
for method in ["nearest", "bilinear", "cubic", "average"]:
    resampled = raster.resample(pixel_width=50, pixel_height=-50, method=method)
    resampled.save(f"gradient_{method}_50m.tif", dtype="float32")
    print(f"{method:10s}  shape={resampled.shape}  "
          f"min={resampled.data.min():.1f}  max={resampled.data.max():.1f}")

When to use each method

• nearest — Land-cover maps, classification rasters, binary masks. Zero blending means class values are never mixed.
• bilinear — DEMs, temperature grids, NDVI. Smooth interpolation avoids staircase artifacts.
• cubic — Satellite imagery, aerial photos. Sharper edges than bilinear with minimal ringing.
• average — Downsampling anything. Each output pixel is the mean of all source pixels it covers, avoiding aliasing.

---

9. Clipping by Extent

Crop a raster to a bounding box. The output is a new raster trimmed to the intersection of the requested extent and the original raster.

from terratiff import TerraTiff

raster = TerraTiff.open("elevation.tif")
print(f"Original bounds: {raster.get_bounds()}")
# (500000.0, 4497000.0, 506000.0, 4500000.0)

# Clip to a smaller area (xmin, ymin, xmax, ymax)
clipped = raster.clip(501000, 4498000, 504000, 4500000)
print(f"Clipped shape:  {clipped.shape}")
print(f"Clipped bounds: {clipped.get_bounds()}")

clipped.save("elevation_clipped.tif", dtype="float32")

Features:

• Automatically clamps to the raster extent if the box extends beyond
• Raises ValueError if there is no overlap
• Preserves pixel size and CRS

---

10. Masking with a Polygon

Mask a raster using a polygon geometry — pixels outside the polygon are set to NoData.

import numpy as np
from terratiff import TerraTiff

# Create a raster
arr = np.ones((100, 100), dtype=np.float32) * 500
raster = TerraTiff.from_array(
    arr, origin_x=500000, origin_y=4500000,
    pixel_width=30, pixel_height=-30, crs="UTM:33N",
)

# Define a polygon (list of (x, y) vertices in map coordinates)
# This rectangle covers the centre of the raster
polygon = [
    (500900, 4499100),   # bottom-left
    (502100, 4499100),   # bottom-right
    (502100, 4499700),   # top-right
    (500900, 4499700),   # top-left
]

# Mask: pixels outside the polygon → NoData
masked = raster.mask_with_polygon(polygon, nodata=-9999)
masked.save("polygon_masked.tif", dtype="float32")

# Invert: mask the INSIDE of the polygon instead
inverted = raster.mask_with_polygon(polygon, invert=True, nodata=0)
inverted.save("polygon_inverted.tif", dtype="float32")

Features:

• Polygon is automatically closed (last vertex connects to first)
• Coordinates must be in the same CRS as the raster
• Works with any number of bands (all bands are masked)
• invert=True flips the mask — useful for cutting holes

---

11. Masking with a Raster

Use another GeoTIFF file as a mask layer — for example, a land/water mask or a classification raster.

import numpy as np
from terratiff import TerraTiff

# Load your data raster
data_raster = TerraTiff.open("elevation.tif")

# Create (or load) a mask raster with the same grid dimensions
# 1 = valid, 0 = masked
mask_arr = np.ones((data_raster.rows, data_raster.cols), dtype=np.uint8)
mask_arr[0:20, 0:20] = 0    # mask the top-left corner
mask_arr[80:, 80:] = 0       # mask the bottom-right corner

mask_raster = TerraTiff.from_array(
    mask_arr,
    origin_x=data_raster.origin_x,
    origin_y=data_raster.origin_y,
    pixel_width=data_raster.pixel_width,
    pixel_height=data_raster.pixel_height,
    crs=data_raster.crs,
)

# Apply the mask
result = data_raster.mask_with_raster(mask_raster, nodata=-9999)
result.save("raster_masked.tif", dtype="float32")

print(f"Masked pixels:  {(result.data[0] == -9999).sum()}")
print(f"Valid pixels:   {(result.data[0] != -9999).sum()}")

Rules:

• Mask raster must have the same grid dimensions (rows × cols) as the data raster
• Pixels where mask = 0 → set to NoData
• Pixels where mask = mask's own NoData value → also set to NoData
• All bands in the data raster are masked

---

12. Converting Between Coordinate Systems

Transform the raster's spatial metadata from one CRS to another — for example, WGS 84 ↔ UTM.

from terratiff import TerraTiff

# A raster in UTM coordinates
raster_utm = TerraTiff.from_array(
    data,
    origin_x=500000, origin_y=4500000,
    pixel_width=30, pixel_height=-30,
    crs="UTM:33N",
)
print(f"UTM origin: ({raster_utm.origin_x}, {raster_utm.origin_y})")

# Convert metadata to WGS 84 (lat/lon)
raster_wgs = raster_utm.to_crs("WGS84")
print(f"WGS84 origin: ({raster_wgs.origin_x:.6f}, {raster_wgs.origin_y:.6f})")
print(f"WGS84 CRS:    {raster_wgs.crs}")  # EPSG:4326

# Save in WGS 84
raster_wgs.save("elevation_wgs84.tif", dtype="float32")

# Convert from WGS 84 to a specific UTM zone
raster_utm15 = raster_wgs.to_crs("UTM:15N")
print(f"UTM15N CRS: {raster_utm15.crs}")  # EPSG:32615

Important: to_crs() performs a metadata-only transformation — it reprojects the origin coordinates and pixel scale but does not warp (re-grid) the pixel data. Use this when your data is already aligned to the target grid, or when you need to update the CRS tag.

---

13. Querying Spatial Metadata

from terratiff import TerraTiff

raster = TerraTiff.open("elevation.tif")

# === Bounding box ===
xmin, ymin, xmax, ymax = raster.get_bounds()
print(f"Bounds: W={xmin}, S={ymin}, E={xmax}, N={ymax}")

# === Transform (origin + pixel size) ===
origin_x, origin_y, pixel_w, pixel_h = raster.get_transform()
print(f"Origin:     ({origin_x}, {origin_y})")
print(f"Pixel size: ({pixel_w}, {pixel_h})")

# === CRS info ===
print(f"CRS string:     {raster.crs}")            # "EPSG:32633"
print(f"CRS name:       {raster.crs_info.name}")   # "WGS 84 / UTM zone 33N"
print(f"Is projected:   {raster.crs_info.is_projected}")  # True
print(f"EPSG code:      {raster.crs_info.epsg}")   # 32633

# === Data info ===
print(f"Shape:      {raster.shape}")     # (bands, rows, cols)
print(f"Dtype:      {raster.dtype}")      # float32
print(f"Bands:      {raster.bands}")
print(f"Rows:       {raster.rows}")
print(f"Cols:       {raster.cols}")
print(f"NoData:     {raster.nodata}")

---

14. Working with NoData Values

Mark missing or invalid pixels with a NoData sentinel value.

import numpy as np
from terratiff import TerraTiff

# Create data with holes
data = np.random.rand(100, 100).astype(np.float32) * 100
data[20:40, 30:60] = -9999  # mark a region as missing

raster = TerraTiff.from_array(
    data,
    origin_x=0, origin_y=0,
    pixel_width=10, pixel_height=-10,
    crs="UTM:33N",
    nodata=-9999.0,          # ← set the NoData value
)

raster.save("with_nodata.tif", dtype="float32")

# Read it back — NoData is preserved
loaded = TerraTiff.open("with_nodata.tif")
print(f"NoData value: {loaded.nodata}")  # -9999.0

# Create a validity mask
valid_mask = loaded.data[0] != loaded.nodata
print(f"Valid pixels: {valid_mask.sum()}")

---

15. UTM Zone Utilities

Automatically determine the correct UTM zone for any location.

from terratiff import utm_zone_from_latlon, utm_epsg, parse_crs

# Find the UTM zone for a given lat/lon
zone, hemisphere = utm_zone_from_latlon(lat=42.0, lon=-93.5)
print(f"Zone: {zone}{hemisphere}")        # 15N

# Get the EPSG code
epsg = utm_epsg(zone, hemisphere)
print(f"EPSG code: {epsg}")               # 32615

# Parse it into a CRS object
crs = parse_crs(f"UTM:{zone}{hemisphere}")
print(f"CRS: {crs}")                      # CRSInfo(EPSG:32615, WGS 84 / UTM zone 15N)

# === More examples ===
print(utm_zone_from_latlon(51.5, -0.1))   # (30, 'N') — London
print(utm_zone_from_latlon(-33.9, 18.4))  # (34, 'S') — Cape Town
print(utm_zone_from_latlon(35.7, 139.7))  # (54, 'N') — Tokyo
print(utm_zone_from_latlon(-22.9, -43.2)) # (23, 'S') — Rio de Janeiro

API Reference

Class: TerraTiff

Constructors

Method	Description
TerraTiff.open(filepath)	Read a GeoTIFF file from disk. Returns a TerraTiff instance.
TerraTiff.from_array(array, origin_x, origin_y, pixel_width, pixel_height, crs, nodata=None)	Create a TerraTiff from a numpy array with user-supplied spatial metadata.

I/O

Method	Description
save(filepath, dtype=None)	Write to a GeoTIFF file. Optionally cast to a specific dtype.

Spatial Operations

Method	Description
resample(pixel_width, pixel_height, method="nearest") → TerraTiff	Resample to a new pixel size. Methods: "nearest", "bilinear", "cubic", "average".
to_crs(target_crs) → TerraTiff	Reproject origin coordinates to a new CRS. Returns a new instance.
clip(xmin, ymin, xmax, ymax) → TerraTiff	Crop the raster to a bounding-box extent.
mask_with_polygon(polygon, invert=False, nodata=None) → TerraTiff	Mask using polygon vertices. Outside → NoData.
mask_with_raster(mask, nodata=None) → TerraTiff	Mask using another raster (0 = masked).

Band Access

Method	Description
get_band(index) → np.ndarray	Return band at index as a 2-D array (0-indexed).
add_band(array)	Append a 2-D array as a new band. Modifies in place.

Queries

Method	Description
get_bounds() → (xmin, ymin, xmax, ymax)	Spatial extent in the raster's CRS units.
get_transform() → (origin_x, origin_y, pixel_width, pixel_height)	Origin and pixel scale.
copy() → TerraTiff	Deep copy of the raster.

Properties

Property	Type	Description
data	np.ndarray	Raw array, shape (bands, rows, cols).
crs	str	CRS as "EPSG:NNNNN".
crs_info	CRSInfo	Detailed CRS object with .epsg, .name, .is_projected.
shape	tuple	(bands, rows, cols).
bands	int	Number of bands.
rows	int	Number of rows.
cols	int	Number of columns.
dtype	np.dtype	Data type of the array.
origin_x	float	X coordinate of the top-left corner.
origin_y	float	Y coordinate of the top-left corner.
pixel_width	float	Pixel size in X direction.
pixel_height	float	Pixel size in Y direction (negative = north-up).
nodata	float/int/None	NoData sentinel value.

---

Utility Functions

Function	Description
parse_crs(crs_input) → CRSInfo	Parse "WGS84", "UTM:33N", "EPSG:4326", or int → CRSInfo.
utm_zone_from_latlon(lat, lon) → (zone, hemisphere)	Get UTM zone number and "N"/"S" from coordinates.
utm_epsg(zone, hemisphere) → int	Get EPSG code for a UTM zone (e.g. 33, "N" → 32633).
supported_dtypes() → list[str]	List all supported dtype strings.

---

Supported CRS Formats

Input Format	Example	Description
"WGS84"	"WGS84"	WGS 84 geographic (EPSG:4326)
"EPSG:NNNNN"	"EPSG:32633"	Any EPSG code as string
int	4326	Any EPSG code as integer
"UTM:ZoneN"	"UTM:33N"	WGS 84 / UTM zone, Northern hemisphere
"UTM:ZoneS"	"UTM:55S"	WGS 84 / UTM zone, Southern hemisphere

All 120 UTM zones (1–60, N and S) are supported.

---

Supported Data Types

Type	numpy dtype	Size	Range	Best for
"uint8"	uint8	1 byte	0 – 255	RGB images, class maps
"uint16"	uint16	2 bytes	0 – 65,535	Satellite imagery (raw DN)
"uint32"	uint32	4 bytes	0 – 4.3B	Large ID rasters
"int8"	int8	1 byte	-128 – 127	Small signed values
"int16"	int16	2 bytes	-32,768 – 32,767	DEM, temperature
"int32"	int32	4 bytes	-2.1B – 2.1B	Large signed values
"float16"	float16	2 bytes	±65,504	Compact float storage
"float32"	float32	4 bytes	±3.4×10³⁸	General scientific data
"float64"	float64	8 bytes	±1.8×10³⁰⁸	High-precision data
"binary"	uint8	1 byte	0 or 1	Masks

---

License

This software is licensed under a Custom License that allows free use for non-commercial applications only. See the LICENSE file for more details.