rindcalc 1.2.9

rindcalc is an open source python package created to calculateLandsat-8 indices, composites, and classification.

rindcalc

Raster Index Calculator

v.1.2.6 notes: Increased stability when reading input landsat bands.

rindcalc is an open source python library built on numpy and gdal aiming to provide seamless and accurate raster index calculations and composites of Landsat-8 imagery using gdal and numpy. Landsat bands are pulled directly from files downloaded from USGS containing all bands in the landsat scene . Since rindcalc uses the standard naming convention of landsat bands, it only needs the folder in which Landsat-8 bands are contained instead. This method allows for easy, quick, and consistent index calculations from Landsat -8 imagery.

Indices: AWEIsh, AWEInsh, NDMI, MNDWI, NDVI, GNDVI, SAVI, NDBI, NDBaI, NBLI, EBBI, UI, NBRI,

Composites: RGB, False Color

Unsupervised Classification: K-Means (Mini Batch)

The k-means unsupervised classification module utilizes sci-kit learn's MiniBatchKMeans which provides significantly faster computation times than the standard K-means algorithm, but with slightly worse result [1]. 'No Data' values are populated with the median value of the array as the classification algorithm does not work with numpy arrays that contain 'nan' values.

Dependencies

• GDAL (v 3.0.0 or greater)
• numpy (v 1.0.0 or greater)
• sci-kit learn (v0.22.1 or greater)

Installation

Windows

pip install rindcalc

For Windows installation gdal wheels must be installed first.

Modules

Composite Modules | rindcalc.composite_utils

• RGB(landsat_dir, out_composite)
• FalseColor(landsat_dir, out_composite)

Index Modules | rindcalc.index_utils

• AWEIsh(landsat_dir, aweish_out, mask_clouds)
• AWEInsh(landsat_dir, aweinsh_out, mask_clouds)
• NDMI(landsat_dir, ndmi_out)
• MNDWI(landsat_dir, mndwi_out)
• NDVI(landsat_dir, ndvi_out, mask_clouds)
• GNDVI(landsat_dir, gndvi_out)
• ARVI(landsat_dir, arvi_out)
• VARI(landsat_dir, vari_out)
• SAVI(landsat_dir, soil_brightness, savi_out)
• NDBI(landsat_dir, ndbi_out)
• NDBaI(landsat_dir, ndbai_out)
• NBLI(landsat_dir, nbli_out)
• EBBI(landsat_dir, ebbi_out)
• UI(landsat_dir, ui_out )
• NBRI(landsat_dir, nbri_out)

landsat_dir = Landsat-8 folder that contains all bands

*_out = out file raster will be saved as

mask_clouds = True or False

i.e. Landsat-8 folder structure:

.
|--LC08_L1TP_091086_20191222_20191223_01_RT               Landsat Folder ex. #1
|   |-- LC08_L1TP_091086_20191222_20191223_01_RT_B1.TIF
|   |-- LC08_L1TP_091086_20191222_20191223_01_RT_B2.TIF
|   |-- LC08_L1TP_091086_20191222_20191223_01_RT_B3.TIF
|   |-- ...
|-- 2019_12_22                                            Landsat Folder ex. #2
|   |-- LC08_L1TP_091086_20191222_20191223_01_RT_B1.TIF
|   |-- LC08_L1TP_091086_20191222_20191223_01_RT_B2.TIF
|   |-- LC08_L1TP_091086_20191222_20191223_01_RT_B3.TIF
|   |-- ...

K Means Classification Module | rindcalc.class_utils

• k_means(input_raster, out_raster, clusters, itr, batch_size)

clusters = Number of classes wanted

itr = Number of iterations to perform

batch_size = Size of mini batches

Example:

import rindcalc as rc
landsat_dir = './.../LC08_L1TP_091086_20191222_20191223_01_RT'
ndvi_out = './.../NDVI_1.tif'
rc.NDVI(landsat_dir, ndvi_out, False)

OR:

import rindcalc as rc
rc.NDVI(landsat_dir = './.../2019_12_22', ndvi_out = './.../NDVI_2.tif'
, mask_clouds=True)

K means unsupervised example:

import rindcalc as rc
input_raster = 'C:/.../.../NDVI.tif'
out_raster = 'C:/.../.../NDVI_K.TIF'
clusters = 2
itr = 10
batch_size = 50
rc.k_means(input_raster, out_raster, clusters, itr, batch_size)

Landsat-8 Bands

Band Number	Name	µm	Resolution
1	Coastal/Aerosal	0.433-0.453	30 m
2	Blue	0.450-0.515	30 m
3	Green	0.525-0.600	30 m
4	Red	0.630-0.680	30 m
5	NIR	0.845-0.885	30 m
6	SWIR 1	1.560-1.660	30 m
7	SWIR 2	2.100-2.300	30 m
8	Panchromatic	0.500-0.680	15 m
9	Cirrus	1.360-1.390	30 m
10	TIR 1	10.6-11.2	100 m
11	TIR 2	11.5-12.5	100 m

Cloud Masking Algorithm

Cloud masking takes the landsat QA band and reads it as a numpy array. Values classed as clouds and cloud shadows are then given the value of 0. Values not equal to zero are then given the value of 1. This mask array is then reshaped back into it's original dimensons. The reshaped array is then multiplied by each input band of the index calulation. This ensures all pixels where clouds and cloud shadows are contained are replaced with 'nan ' and all other pixels retain their original values.

mask_values = [2800, 2804, 2808, 2812, 6986, 6900, 6904, 6908,
               2976, 2980, 2984, 2988, 3008, 3012, 3016, 3020,
               7072, 7076, 7080, 7084, 7104, 7108, 7112, 7116]

m = np.ma.array(qa_band,
                    mask=np.logical_or.reduce([qa_band == value for value 
                                                in mask_values]))
np.ma.set_fill_value(m, 0)
m1 = m.filled()
m1[m1 != 0] = 1

m1.reshape(qa_band.shape)

Indices

Composites RGB = (Red, Green, Blue)

Water

• AWEIsh = ((Blue + 2.5 * Green - 1.5 * (NIR + SWIR1) - 0.25 * SWIR2)) / (Blue + Green + NIR + SWIR1 + SWIR2) [1]

• AWEInsh = ((4 * (green_band - swir1_band) - (0.25 * nir_band + 2.75 * swir1_band)) / (green_band + swir1_band + nir_band)) [1]

• MNDWI = ((Green - SWIR1) / (Green + SWIR1)) [3]

Moisture

• NDMI = ((NIR - SWIR1) / (NIR + SWIR1)) [2]

Vegetation

• NDVI = ((NIR - Red) / (NIR + Red)) [4]

• Green NDVI (GNDVI) = ((nir_band - green_band) / (nir_band + green_band))

• ARVI = ((nir_band - (2 * red_band) + blue_band) / (nir_band + (2 * red_band) + blue_band)) [5]

• VARI = ((green_band - red_band) / (green_band + red_band - blue_band))

• SAVI = ((NIR - Red) / (NIR + Red + L)) x (1 + L)

  • L = Soil Brightness Factor

• MSAVI2 = (((2 * nir_band + 1) - (np.sqrt(((2 * nir_band + 1)**2) - 8 * (nir_band - red_band)))) / 2)

Urban/Landscape

• NDBI = (SWIR1 - NIR) / (SWIR1 + NIR)

• NDBaI = ((SWIR1 - TIR) / (SWIR1 + TIR))

• NBLI = ((Red - TIR) / (Red + TIR))

• EBBI = ((swir1_band - nir_band) / (10 * (np.sqrt(swir1_band + tir_band))))

• UI = ((swir2_band - nir_band) / (swir2_band + nir_band))

Fire

• NBRI = ((nir_band - swir2_band) / (nir_band + swir2_band))

References

[1] Feyisa, G. L., Meilby, H., Fensholt, R., & Proud, S. R. (2014). Automated Water Extraction Index: A new technique for surface water mapping using Landsat imagery. Remote Sensing of Environment, 140, 23-35

[2] Gao, B. C. (1996). NDWI—A normalized difference water index for remote sensing of vegetation liquid water from space. Remote sensing of environment, 58(3), 257-266.

[3] Xu, H. (2006). Modification of normalised difference water index (NDWI ) to enhance open water features in remotely sensed imagery. International journal of remote sensing, 27(14), 3025-3033.

[4] Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetation. Remote sensing of Environment, 8(2), 127-150.

[5] Kaufman, Y. J., & Tanre, D. (1992). Atmospherically resistant vegetation index (ARVI) for EOS-MODIS. IEEE transactions on Geoscience and Remote Sensing, 30(2), 261-270.