Super Pit Gold Mine Land Cover

Introduction

Land Cover could change by many different reasons. The widely known are: climate, geological changes and human factor. Since, there were no significant climate or geological changes in the study for the research time frame, this project is seeking answer about any significant land changes because of Big Pit was suppressed by new gold mine pit.

Study area and test data

The Fimiston Open Pit, colloquially known as the Super Pit, was Australia’s largest open cut gold mine until 2016 when it was surpassed by the Newmont Boddington gold mine also in Western Australia. The Super Pit is located off the Goldfields Highway on the south-east edge of Kalgoorlie, Western Australia. It is is approximately 3.5 kilometers long, 1.5 kilometers wide and over 600 meters deep. The study area is located around the Super Pit approximately 2710 km sq. A Landsat-8 OLI image with processing level 1T of the study area were acquired on Jan 23, 2015 and Jan 1, 2019. The images have been geometrically corrected with reference to topography. Landsat-8 OLI image contains eight multispectral bands with a spatial resolution of 30m and one 15-m panchromatic band. The radiometric resolution is 16 bits, which is an improvement over preceding Landsat sensors. The improvement of radiometric resolution can effectively avoid grayscale over-saturation in extreme dark region and facilitate discriminating subtle features with extremely low reflectivity (such as water bodies). Climat data raster were retrieved on Jan, 2015 and Jan, 2019 in order to avoid uncertainty about main reason of the land cover changes. Elevation data raster (NASADEM Merged DEM Global 1 arc second V001) for the study area was retrieved (granule on Feb, 2000) from NASA DEM repository.

Data Source:
• USGS Earth Explorer
• World Climate Data
• Earth Data

Method

At its core, this project involves two broad components: a satellite image classification and a ModelBuilder-based analysis.

1. Climate data comparing

To avoid uncertainty if land cover change could be due to weather condition change, the climate data between given two years should be compare firstly. The table below shows average minimum temperature (°C), average maximum temperature (°C) and total precipitation (mm) for Jan,2015 and Jan,2019 in the study area.

Data	Jan, 2015	Jan, 2019
min temp (°C)	19	17
max temp (°C)	35	33
precipitation (mm)	11	13

Obviously, Jan in 2015 was hotter then in 2019 but not so significantly to say that any land cover change occurs because of the different weather conditions.

2. Satellite image classification

Training samples

In this project the area of interest is defined by six (6) features (classification classes):

• Bare soil with scattered tree
• Vegetation
• Bare land
• Salt lakes
• Mining sites
• Urban and build up

In order to perform land classification, training samples for each class of classification should be created and tested first. The image below shows a closeup the samples of each feature type of interest. False-color image by using SWIR band as red, NIR as green, and deep blue as blue (technically, a 7-5-1 image) were chosen to identify feature classes as details of vegetation, water, different type of soil and geology become clear in this band combination.

Supervised classification

Signature file was created on base of tested training samples. Next step is to run a supervised (trained) classification on satellite image for each year in the Landsat series. Results are classified rasters (Land Cover Classification) for two analyzed years.

3. ModelBuilder-based analysis

Land Cover Analysis Tool was created using ModelBuilder in ArcMap.

Main steps of this model’s workflow are:
• Get the study area: extract the Land Cover Classification rasters by mask
• Reclassify both output classification rasters to reset classification values to order values from 0 to (n-1), where n = number of land categories. Model is designed for 6 categories.
• Summarize the area of different land classes in the study area using Tabulate area tool.
• Rename fields in the tables to get intuitive field’s names that represent Land Category number and year. Land category number matches unique value in the training sample and could be easily detected.
• Join table with old year (2015) values to the new year (2019) table.
• Add new fields to joined table.
• Calculate new fields to get difference by meters square in each category.
• Copy joined table to the final table that represents only land cover changes.

The graph below represents the result of this analysis.

Conclusion

The result’s table and graph show only slightly changes in mining sites and salt lakes categories (5-7 km²). Increased urban and build-up areas also are not showing big numbers. However, there is significant change in favor to a vegetation category. Where in 2015 year was mainly bare soil with scattered tree, in 2019 these areas have more healthy vegetation. So, there is might be some relationship between new opened big gold pit in 2016 that leads to improved environment conditions. But, it would be to early to state it without some additional analysis (such as urban planning dates, etc).

Final cartography

Due to analysis shows only significant change in the vegetation category it was decided to present only vegetation category change in the final map for the better visualization purposes.

Following additional steps were performed in order to create final map. In ArcMAP:
• Create hillshade (using an elevation raster).
• From the land category raster remove all categories except vegetation (for both analyzed years).
• Add graticule, north arrow, scale, Super Pit point and a few populated places in the study area.
• Export map to Adobe Illustrator
In Adobe Illustrator:
• Trace image of vegetation raster for both years to a vector.
• Combine vegetation vectors for both years, hillshade raster and additional details (graticule, arrow, etc) to one layout.
• Apply different symbology to the vegetation category in order o visually separate 2015 and 2019 years.
• Correct graticule visualization.
• Apply transparency for vegetation layer to give it fine hillshade look.
• Add title, subtitle, inset, data sources.

Full PDF cartography map is here.