rindcalc

Raster Index Calculator

rind calc is a python package created to allow raster index calculation using Landsat-8 using gdal and numpy. Landsat bands are pulled directly from file downloaded from USGS containing all bands in landsat scene. Since rindcalc only requires the file in which Landsat-8 bands are contained instead of each individual band to be specified, it allows for easy, quick, and seamless index calculations from Landsat-8 imagery.

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

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 ( v (0.22.1 or greater))

Installation

Windows

pip install rindcalc

For Windows installation gdal wheels must be installed first.

Modules

Index Modules

• AWEIsh(landsat_dir, aweish_out)
• AWEInsh(landsat_dir, aweinsh_out)
• NDMI(landsat_dir, ndmi_out)
• MNDWI(landsat_dir, mndwi_out)
• NDVI(landsat_dir, ndvi_out)
• GNDVI(landsat_dir)
• 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

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

• 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

EX:

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

OR:

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

KMEANS EX:

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

Indices

Water

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

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

• NDWI = ((nir_band - swir1_band) / (nir_band + swir1_band))

• MNDWI = ((Green - SWIR1) / (Green + SWIR1))

Moisture

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

Vegetation

• NDVI = ((NIR - Red) / (NIR + Red))

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