TerraEye is a tool, dedicated to the opencast mining industry, to among other things, monitor the environmental impacts of opencast mines. The functionalities we are developing are based on satellite data analysis, with the support of artificial intelligence and machine learning. In this article, we present selected functionalities relating to environmental impacts, using several Russian open-pit mines as examples.

The idea behind TerraEye system is automated monitoring of both mining areas and surrounding areas affected by mining activities. Monitoring of areas affected by mining activities is centered around the environmental impacts of open-pit mining. In an era of increasing social and environmental awareness, effective monitoring and reporting of these impacts is becoming mandatory in most countries. In addition, TerraEye system can support the decommissioning and reclamation processes of open-pit mines through quantitative and qualitative monitoring of the effectiveness of restoring use and natural values to the areas transformed by mining activities. TerraEye system provides processed information ready for analysis. Based on this alert system, in-situ sampling can be significantly reduced.

Practice, what did we analyze? Russian open-pit mines

Recently, due to the ongoing Russian-Ukrainian war, we decided to take several Russian open-pit mines under our R&D work. Russia is the world’s largest country in terms of surface area, covering more than 17 million square kilometers. This huge area determines the above-average access to raw materials, and those strategic resources play one of the key roles in the ongoing conflict. Focusing on raw materials extracted by surface mining methods, Russia ranks second in the world regarding coal reserves (175 billion tons). It also holds huge amounts of metal ores including, among others, the world’s third-largest iron ore reserves (25 billion tons). In addition, Russia is a major producer of critical minerals such as nickel, palladium, and rhodium. It is also the second producer of cobalt and the sixth-largest producer of graphite in the world.
Three areas where Russian mines are conducting or have conducted open-pit mining were selected for the purpose of this analysis

(Fig. 1) Open pit coal mine. An area located in southern Russia, in the eastern part of the Novosibirsk region, about 20 km east of the town of Iskitim, where hard coal is mined in several deep open-pit excavations increasing in size each year. In this area, large-scale external dumps are also visible.
(Fig. 2) Open pit iron ore mine. An area located in southern Russia, in the Chelyabinsk region, about 20 kilometers west of the border with Kazakhstan. This area includes a deep open – pit of iron ore and associated external dumps, a processing unit, and a tailings pond.
(Fig. 3) Decommissioned Opencast Coal Mine. The area located in the western part of Russia, in the Sverdlovsk region. It contains the final pit after lignite mining where mining has been completed and is now being filled with water. This pit is directly adjacent to the village of Volchansk. A huge external dump, reclaimed into the forest, is also visible.

In each of the selected areas, the potential for the application of TerraEye functionality was recognized.
At location (1), a determining factor in the selection was the noted progressive deforestation and the exclusion of land for the purpose of opening new parts of the deposit out, as well, as providing new dumping sites. Also observed was the possibility of deterioration of vegetation in the vicinity of the pits due to the huge, increasing scale of mining in the area.
In location (2), the deciding factor for selection was the increasing land use caused by pits and dumps,  and the presence of a tailing storage facility.
In location (3), the deciding factor for selection was the decommissioning stage of the mine and the recharging water table in the final excavation, with the presence of many natural lakes in the vicinity of the excavations.

The functionalities of TerraEye system used for these texts are the segmentation of mining area elements, water detection, and greenery detection.

Functionality – segmentation of elements in the mining areas  
The functionality we are developing allows the user to automatically detect and track changes in the land cover classes of mining areas.

Depending on the user’s needs, it can provide information on the current surface of areas transformed by mining activities and support decision-making processes by verifying long-term plans at the highest management level. TerraEye system, by tracking changes in the surface of individual elements of open-pit mining operations, refers to the most obvious environmental impact of mining activities, i.e. land use. The machine learning models we are developing are capable of recognizing as many as 15 land cover classes automatically, including the most basic ones – excavations and dumps. This is a functionality that is complementary to others targeting action at individual sites (slope stability monitoring, land subsidence on dumps). In the future, it will be necessary for the operation of an integrated, multifunctional system, to reduce the amount of input data required by the user. Moreover, we are working on the possibility of integrating this functionality with autonomous drones, which will independently select the destination of the flight to collect detailed data. 
Below for location (2), examples of excavation area predictions automatically generated by our machine learning model are presented. The comparison is between 2017 and 2021.

Fig. 1 Changes in the pit area at location 2 from 2017 to 2021, made by the ML model.

In the visible mining area, a significant increase in the surface area of excavations and external dumps is observed. Among other things, TerraEye system provided information that, as a result of the opening new parts of the deposit out and the progress of mining, the excavation area between 2017 and 2021, increased from more than 119 to about 332 hectares. As a result, about 213 hectares of land were taken out of use for iron ore mining.

Functionality – segmentation of elements of mining areas

Similarly, below, this time for location (1), examples of dumping area predictions automatically generated by our machine learning model. The years 2018-2022 are compared

Fig. 2 Changes in dumping areas at location (1) from 2018 to 2022, made by the ML model.

In the exposed mining area, there is a significant increase in size of pits and dumps. TerraEye system has provided information that, because of the progress of external dumping, the total area of the 4 existing external dumps has increased from nearly 900 to more than 1,378 hectares between 2018 and 2022. As a result, about 417 hectares of land have been taken out of use for waste rock placement alone. The heap in the eastern part of the mining area increased most significantly, increasing its area by more than 200 hectares.
Similar analyses of area changes were made for other predicted classes of mining area elements. The effectiveness of the prediction systematically increases as the amount of input data for the machine learning models increases.

Functionality – water detection

Developed water detection functionality allows the user to monitor both the changing surface of water bodies and selected water quality indicators.

In open-pit mines, due to the possible influx of groundwater and rainwater into the pits, it is necessary to adequately dewater the deposits to allow mining. Dewatering of deposits with deep wells leads to the formation of a depression cone, the radius of which can reach up to several tens of kilometers. This leads to a lowering of groundwater and surface water levels in the vicinity of the mine. Furthermore, open-pit mines consume large amounts of water in technological processes related to the extraction and ore processing. Mines are required to monitor the impacts of their water management on surrounding surface water bodies. In addition, the most common reclamation method for final pits of open-pit mines is the creation of pit lakes. It is necessary to properly plan, control, and monitor the processes of pit lakes creation, both in terms of the peed of flooding and continuous monitoring of water quality.
Below for location (3), changes in the surface of the water table in the final pit after lignite mining. The analysis was performed for a time series of images from 2017-2022 (the month of August). What can be seen is the successive flooding of the final pit – due to the groundwater recharge and recipitation caused by the cessation of dewatering. The area of the water table in the final pit indicated by our system increased from more than 56 hectares in 2017 to almost 110 hectares in 2022.

Fig. 3 Changes in the surface of the water table in the final pit after lignite mining at location 3. View from the TerraEye application.

Moreover, the cessation of dewatering of the pit resulted in the restoration of the water table throughout the extent of the depression cone. Our system was able to detect a successive increase in the water table area of the surrounding lakes, which had been negatively affected by the depression cone in the past

Fig. 4 Changes in the surface of the water table in the final pit after lignite extraction at location (3). 

TerraEye system is also a tool for calculating remote sensing indicators describing surface water quality in water bodies. Thanks to the indicators already implemented in our system, it is possible, among other things, to determine the amount of chlorophyll in water, the increase of which indicates the deterioration of chemical, physical and biological parameters of water (CHL_A; NDCI); to determine the amount of dissolved organic matter, including dissolved organic carbon, the increase of which indicates the progressive degradation of biota in water (CDOM; DOC); to determine the turbidity of the water, which is a consequence of a large number of solid particles in the water body that can negatively affect water quality (TURB). Our team of specialists is constantly working on expanding the service by adding more water quality indicators.

The figure below shows an example of analytical results for dissolved organic carbon (DOC) in 2019 at location (3) and surrounding natural reservoirs. It should be noted that the amount of dissolved organic carbon in anthropogenic post-mining reservoirs that fill with water is at a much lower level than in natural lakes.

Fig. 5 Comparison of total organic carbon (DOC) in the tailings pit and surrounding lakes in 2019; map of DOC indicator hot-spots in the tailings pit.

TerraEye system can be used to support in-situ monitoring and research by indicating so-called hot spots and cold spots, i.e. points where the greatest changes have occurred compared to past data, or where a trend is evident. Based on their values, TerraEye system alerts the user to anomalies and potential risks.

Water detection functionality can also be useful in monitoring the surface of the water table in tailing – storage facilities, where the water table shouldn’t approach the upper edge of the inner slope. An example analysis was performed for the tailings pond at location (2), where the water table systematically changed, depending on the amount of waste discharged, water recovered, and weather conditions.

Fig. 6 Water table detection for the tailings pond for location (2) in 2017. 

Functionality – greenery detection

The developed greenery detection functionality is used to inventory the vegetation cover in the area affected by mining activities by indicating the extent of each vegetation cover class. Our machine learning model, as part of this functionality, can distinguish as many as 11 land cover classes, including trees, shrublands grasses, or croplands. TerraEye system can quantify losses in vegetation cover caused by, for example, deposit opening works. On the other hand, it can be used to inventory vegetation in areas rehabilitated in forestry or agricultural directions.
Below is an example of vegetation cover classification by the machine learning model for part of the location (1)

Fig. 7 Example of vegetation cover classification, for the northwestern part of the location area (1); green – trees, yellow – grasses, pink – croplands, and blue – water.

In addition, in the areas determined by the machine learning model, green quality analysis can be carried out for each pixel. Based on the remote sensing index NDVI (Normalized Difference Vegetation Index), it is possible to determine the developmental state and condition of vegetation. It is also possible to monitor the water content of foliage, using the Normalized Difference Infrared Index (NDII), and the chlorophyll content, using the Normalized Difference Red-Edge Index (NDRE), low values of which will suggest disease and plant damage. The indices are calculated separately for each class of greenery predicted by the machine learning model. In addition, it is possible to compare changes in the size of the indices over time. Below (Fig. 8), the results of comparing the size of the NDVI index are shown for fragments of forests in the central part of the location (1), where a significant decrease in its value is evident in the vicinity of the newly transformed area between 2017 and 2022.

Fig. 8 Fragment of the location area (1) with the newly transformed area visible in 2021 (left side); comparison of NDVI index values between 2017 and 2022.