GDAL 手册上frmt_gtiff.html文件上找到这句话，感觉很有用，以备查阅：
Creation Options
TFW=YES: Force the generation of an associated ESRI world file (.tfw).See a World Files section for details.

RPB=YES: Force the generation of an associated .RPB file to describe RPC (Rational Polynomial Coefficients), if RPC information is available. If not specified, this file is automatically generated if there’s RPC information and that the PROFILE is not the default GDALGeoTIFF.

RPCTXT=YES: (GDAL >=2.0) Force the generation of an associated _RPC.TXT file to describe RPC (Rational Polynomial Coefficients), if RPC information is available.

INTERLEAVE=[BAND,PIXEL]: By default TIFF files with pixel interleaving (PLANARCONFIG_CONTIG in TIFF terminology) are created. These are slightly less efficient than BAND interleaving for some purposes, but some applications only support pixel interleaved TIFF files.

TILED=YES: By default stripped TIFF files are created. This option can be used to force creation of tiled TIFF files.

BLOCKXSIZE=n: Sets tile width, defaults to 256.

BLOCKYSIZE=n: Set tile or strip height. Tile height defaults to 256, strip height defaults to a value such that one strip is 8K or less.

NBITS=n: Create a file with less than 8 bits per sample by passing a value from 1 to 7. The apparent pixel type should be Byte. From GDAL 1.6.0, values of n=9…15 (UInt16 type) and n=17…31 (UInt32 type) are also accepted. From GDAL 2.2, n=16 is accepted for Float32 type to generate half-precision floating point values.

COMPRESS=[JPEG/LZW/PACKBITS/DEFLATE/CCITTRLE/CCITTFAX3/CCITTFAX4/LZMA/ZSTD/LERC/LERC_DEFLATE/LERC_ZSTD/WEBP/NONE]: Set the compression to use. JPEG should generally only be used with Byte data (8 bit per channel). But starting with GDAL 1.7.0 and provided that GDAL is built with internal libtiff and libjpeg, it is possible to read and write TIFF files with 12bit JPEG compressed TIFF files (seen as UInt16 bands with NBITS=12). See the “8 and 12 bit JPEG in TIFF” wiki page for more details. The CCITT compression should only be used with 1bit (NBITS=1) data. LZW, DEFLATE and ZSTD compressions can be used with the PREDICTOR creation option. ZSTD is available since GDAL 2.3 when using internal libtiff and if GDAL built against libzstd >=1.0, or if built against external libtiff with zstd support. LERC/LERC_DEFLATE/LERC_ZSTD are available since GDAL 2.4 when using internal libtiff (and for LERC_ZSTD, see above mentioned conditions). None is the default.

NUM_THREADS=number_of_threads/ALL_CPUS: (From GDAL 2.1) Enable multi-threaded compression by specifying the number of worker threads. Worth for slow compressions such as DEFLATE or LZMA. Will be ignored for JPEG. Default is compression in the main thread.

PREDICTOR=[1/2/3]: Set the predictor for LZW, DEFLATE and ZSTD compression. The default is 1 (no predictor), 2 is horizontal differencing and 3 is floating point prediction.

DISCARD_LSB=nbits or nbits_band1,nbits_band2,…nbits_bandN: (GDAL >= 2.0) Set the number of least-significant bits to clear, possibly different per band. Lossy compression scheme to be best used with PREDICTOR=2 and LZW/DEFLATE/ZSTD compression.

SPARSE_OK=TRUE/FALSE (From GDAL 1.6.0): Should newly created files (through Create() interface) be allowed to be sparse? Sparse files have 0 tile/strip offsets for blocks never written and save space; however, most non-GDAL packages cannot read such files. Starting with GDAL 2.2, SPARSE_OK=TRUE is also supported through the CreateCopy() interface. Starting with GDAL 2.2, even an attempt to write a block whose all pixels are 0 or the nodata value will cause it not to be written at all (unless there is a corresponding block already allocated in the file). The default is FALSE.

JPEG_QUALITY=[1-100]: Set the JPEG quality when using JPEG compression. A value of 100 is best quality (least compression), and 1 is worst quality (best compression). The default is 75.

JPEGTABLESMODE=0/1/2/3: (From GDAL 2.0) Configure how and where JPEG quantization and Huffman tables are written in the TIFF JpegTables tag and strip/tile. Default to 1.

0: JpegTables is not written. Each strip/tile contains its own quantization tables and use optimized Huffman coding.
1: JpegTables is written with only the quantization tables. Each strip/tile refers to those quantized tables and use optimized Huffman coding. This is generally the optimal choice for smallest file size, and consequently is the default.
2: JpegTables is written with only the default Huffman tables. Each strip/tile refers to those Huffman tables (thus no optimized Huffman coding) and contains its own quantization tables (identical). This option has no anticipated practical value.
3: JpegTables is written with the quantization and default Huffman tables. Each strip/tile refers to those tables (thus no optimized Huffman coding). This option could perhaps with some data be more efficient than 1, but this should only occur in rare circumstances.
ZLEVEL=[1-9]: Set the level of compression when using DEFLATE compression (or LERC_DEFLATE). A value of 9 is best, and 1 is least compression. The default is 6.

ZSTD_LEVEL=[1-22]: Set the level of compression when using ZSTD compression (or LERC_ZSTD). A value of 22 is best (very slow), and 1 is least compression. The default is 9.

MAX_Z_ERROR=threshold: Set the maximum error threshold on values for LERC/LERC_DEFLATE/LERC_ZSTD compression. The default is 0 (lossless).

WEBP_LEVEL=[1-100]: Set the WEBP quality level when using WEBP compression. A value of 100 is best quality (least compression), and 1 is worst quality (best compression). The default is 75.

WEBP_LOSSLESS=True/False: (GDAL >= 2.4.0 and libwebp >= 0.1.4): By default, lossy compression is used. If set to True, lossless compression will be used.

PHOTOMETRIC=[MINISBLACK/MINISWHITE/RGB/CMYK/YCBCR/CIELAB/ICCLAB/ITULAB]: Set the photometric interpretation tag. Default is MINISBLACK, but if the input image has 3 or 4 bands of Byte type, then RGB will be selected. You can override default photometric using this option.

ALPHA=[YES/NON-PREMULTIPLIED/PREMULTIPLIED/UNSPECIFIED]: The first “extrasample” is marked as being alpha if there are any extra samples. This is necessary if you want to produce a greyscale TIFF file with an alpha band (for instance). For GDAL < 1.10, only the YES value is supported, and it is then assumed as being PREMULTIPLIED alpha (ASSOCALPHA in TIFF). Starting with GDAL 1.10, YES is an alias for NON-PREMULTIPLIED alpha, and the other values can be used.

PROFILE=[GDALGeoTIFF/GeoTIFF/BASELINE]: Control what non-baseline tags are emitted by GDAL.

With GDALGeoTIFF (the default) various GDAL custom tags may be written.
With GeoTIFF only GeoTIFF tags will be added to the baseline.
With BASELINE no GDAL or GeoTIFF tags will be written. BASELINE is occasionally useful when writing files to be read by applications intolerant of unrecognized tags.
BIGTIFF=YES/NO/IF_NEEDED/IF_SAFER: Control whether the created file is a BigTIFF or a classic TIFF.

YES forces BigTIFF.
NO forces classic TIFF.
IF_NEEDED will only create a BigTIFF if it is clearly needed (in the uncompressed case, and image larger than 4GB. So no effect when using a compression).
IF_SAFER will create BigTIFF if the resulting file might exceed 4GB. Note: this is only a heuristics that might not always work depending on compression ratios.
BigTIFF is a TIFF variant which can contain more than 4GiB of data (size of classic TIFF is limited by that value). This option is available if GDAL is built with libtiff library version 4.0 or higher. The default is IF_NEEDED.

When creating a new GeoTIFF with no compression, GDAL computes in advance the size of the resulting file. If that computed file size is over 4GiB, GDAL will automatically decide to create a BigTIFF file. However, when compression is used, it is not possible in advance to known the final size of the file, so classical TIFF will be chosen. In that case, the user must explicitly require the creation of a BigTIFF with BIGTIFF=YES if the final file is anticipated to be too big for classical TIFF format. If BigTIFF creation is not explicitly asked or guessed and the resulting file is too big for classical TIFF, libtiff will fail with an error message like “TIFFAppendToStrip:Maximum TIFF file size exceeded”.

PIXELTYPE=[DEFAULT/SIGNEDBYTE]: By setting this to SIGNEDBYTE, a new Byte file can be forced to be written as signed byte.

COPY_SRC_OVERVIEWS=[YES/NO]: (GDAL >= 1.8.0, CreateCopy() only) By setting this to YES (default is NO), the potential existing overviews of the source dataset will be copied to the target dataset without being recomputed. If overviews of mask band also exist, provided that the GDAL_TIFF_INTERNAL_MASK configuration option is set to YES, they will also be copied. Note that this creation option will have no effect if general options (i.e. options which are not creation options) of gdal_translate are used.

GEOTIFF_KEYS_FLAVOR=[STANDARD/ESRI_PE]: (GDAL >= 2.1.0) Determine which “flavor” of GeoTIFF keys must be used to write the SRS information. The STANDARD way (default choice) will use the general accepted formulations of GeoTIFF keys, including extensions of the values accepted for ProjectedCSTypeGeoKey to new EPSG codes. The ESRI_PE flavor will write formulations that are (more) compatible of ArcGIS. At the time of writing, the ESRI_PE choice has mostly an effect when writing the EPSG:3857 (Web Mercator) SRS. For other SRS, the standard way will be used, with the addition of a ESRI_PE WKT string as the value of PCSCitationGeoKey.