1. Introduction
The central United States is host to one of the largest drainage basins in the world, the Mississippi River. Today, the Mississippi River has depths of up to 100 ft (30.5 m) and is 2350 miles (3,781 km) long. The river, however, has a long and complex history [
1,
2,
3], with some profound changes occurring during the Pliocene and Pleistocene [
4,
5,
6,
7,
8,
9].
A remnant of the ancestral Mississippi River floodplain is the (~3.2 Ma) Pliocene [
10] Upland Complex (UC) deposit, a high-level sand and gravel river terrace that is discontinuously preserved on interfluves adjacent to the modern Mississippi River in Illinois, Missouri, Kentucky, Tennessee, Arkansas, Mississippi, and Louisiana (
Figure 1). Geologists have mapped the UC and use this information to better understand the Pliocene Mississippi River [
7,
8,
9]. Sand and gravel companies are particularly interested in understanding the distribution and thickness of this deposit because it is the major commercial source of sand and gravel along the lower Mississippi River [
11].
The UC (also called the Lafayette Gravel, Mounds Gravel, or pre-loess gravel) is interpreted to be the remnant of a deposit of the Pliocene Mississippi River that once formed a continuous blanket up to 492-ft (150 m) thick, ≥62 miles (100 km) east–west from the Chickasaw Bluffs in western Tennessee to at least Crowley’s Ridge in eastern Arkansas, and 435 miles (700 km) north–south along the Mississippi River from northern Illinois to southern Louisiana (
Figure 1) [
4,
5,
7,
8,
9,
12,
13,
14,
15,
16]. Sea level decline during the Pleistocene resulted in an incision by the Mississippi River that isolated the UC as an eroded high-level terrace [
4,
7]. The UC disconformably overlies Eocene formations and is disconformably overlain by Pleistocene loess.
The UC primarily consists of brown to very pale brown (Munsell color 7.5 YR 4/6 to 10 YR 8/6) massive bedded well-rounded chert pebble gravel and coarse quartz sand (
Figure 2) [
9]. Gravel clasts are almost entirely chert with some polycrystalline quartzite (15%). Chert clasts are stained with a surficial coating of goethite (FeO(OH)), 1.0–2.0 mm thick. The red-brown goethite staining shows no percussion marks, which indicates that the brown stain is post-depositional. Smaller clasts (12 cm) are frequently brown stained throughout or are partially or completely replaced by goethite. Common unstained clasts vary in abundance, without an observable geographic pattern [
9].
The UC is of great interest, both geologically and economically. Studying the distribution and deposition of the UC has helped geologists understand the evolution of the lower Mississippi River Valley. The UC is also economically valuable, as it is the primary source of sand and gravel for Memphis, Tennessee and the surrounding region (
Figure 1). Having a local source of gravel keeps construction costs low, so it is sought by mining companies particularly near cities. This research project was undertaken to evaluate the use of drone photography for mapping the stratigraphy of the UC exposed in an active quarry in DeSoto County, Mississippi and for photographing quarry excavation throughout the life of one of the quarry’s pits (
Figure 3).
Unmanned Aerial Vehicle (UAV) technology, particularly drones, is increasingly valuable in geologic investigations [
17,
18]. Drone technology allows for safe study of difficult-to-access or hazardous areas and drones have been important in risk assessment and management [
19]. Quarry highwalls are especially hazardous due to possible collapse and falling rocks. Because physical access to quarry highwalls is a safety concern, physical sampling of the highwall sediment is often not permitted by quarry management.
This study focused on a sand and gravel quarry in DeSoto County, Mississippi. Specifically, this research was conducted in the pit of an active quarry, mined by Memphis Stone and Gravel Company in Southaven, Mississippi (
Figure 3), which is being mined in the Pliocene UC deposit. Our primary objectives were to assess the utility of drone photography to map the three-dimensional geology exposed in the quarry pit to better understand its environment of deposition and to provide information that may help in the future exploration of the sand and gravel deposit regionally.
2. Materials and Methods
Drones and other forms of UAVs allow for high-definition aerial photography and are useful for imaging inaccessible or hazardous areas such as +100-ft (+30.5 m) high quarry walls (highwalls) (
Figure 2). We conducted a total of eight drone photography surveys between May 2017 and May 2018 as the sand and gravel pit was actively being mined. To obtain the best consistent light conditions for mapping, as well as to minimize shadow effects for the interior highwalls, we carried out all the flights between 11:00 and 13:00, with a clear or fully overcast sky, minimal highwall shadow, and minimal wind. In this study we used geometric-mathematical reconstructions of individual points to reconstruct the photographs in a three-dimensional (3-D) space [
19]. From UAV photogrammetry, point clouds, orthomosaics of the highwalls, and digital 3-D models were constructed (
Figure 4 and
Figure 5).
For this study, we used a drone for several reasons. Drones are a cost effective and rapid method for safely photographing large areas that allow for pre-flight elevation and flight path programming. Many areas of the pit were unapproachable because of water in the pit and the potentially unstable near-vertical highwalls. The drone allowed access to areas that would not allow a fixed-wing aircraft because of right-angle bends in the highwall. Drone technology permitted various photograph look directions, which allowed us to photograph the near-vertical highwall faces. The drone photography also allowed us to quickly and accurately create digital surface models, orthophotos, highwall photomosaics, and 3-D models from the imagery [
18].
In this research, we used a Phantom 4 Pro drone equipped with a 1-inch 20-megapixel CMOS sensor (400 nm to 700 nm RGB range; model number: CP.PT.000689) with a field of view of 84° or 8.8 mm (35 mm format equivalent: 24 mm). The Phantom model has a maximum flight time of 30 min and can be either controlled manually or operated fully autonomously using a predefined Global Positional System (GPS) path. The camera was mounted on the base of the drone and could be oriented to photograph in any direction. Vertical, oblique, and horizontally oriented photographs were taken at different distances from the highwall to determine good flight distances and look directions for geologic analysis.
All the drone flights were conducted by 901Drones of Memphis, Tennessee. Ten permanent ground control points (GCP) were placed around the exterior and the interior of the pit, with coordinates surveyed using a GNSS receiver (model: Trimble R8s) with a static maximum precision of 8 mm and 15 mm for horizontal and vertical accuracy, respectively. GCPs were used to provide topographic control of the 3-D models and to calculate photogrammetric accuracy (
Figure 4). The Root Mean Square Error (RMSE) was calculated as the differences between the x-y-z positions of the ten GCPs and their x-y-z positions on the georeferenced orthophoto. Several methods were tested to determine the most effective parameters for flying the drone. For vertical 3-D mapping of the pit, the drone was flown at 200 ft (61 m) above the quarry floor with the camera pointed directly down, and the drone was launched and controlled from within the pit. Approximately 350 images were required to cover the entire pit area and to capture the GCPs on the pit’s perimeter. For analysis of the highwalls, two photography perspectives were employed. Drone photography was taken at 90° (horizontal) angles within 25–30 ft (7.6–9.1 m) of the highwalls and obliquely at 45° at 20 ft (6.1 m) above the highwall top. A 1.5 m wooden scale was placed on the highwalls prior to horizontal and oblique aerial photographing. All the photos had a minimum of 60% overlap for subsequent 3-D modeling and stereographic viewing to ensure that every part of the area was covered in a minimum of three overlapping images.
Two software packages were used to process the drone photography in this study: 3D Survey (
https://www.3dsurvey.si) and Agisoft Photoscan Professional (
https://www.agisoft.com). The 3D Survey software is a point-cloud processing software that was used to generate vertical and oblique 3-D models and orthophotos of the quarry pit. Agisoft Photoscan was used to create orthomosaics and 3-D models of the near-vertical highwalls for geologic “cross-section” mapping (
Figure 4,
Figure 5,
Figure 6,
Figure 7 and
Figure 8).
3. Results
During eight visits to the quarry, aerial drone mapping flights were conducted to track the mining progress and to photograph the temporary interior highwalls. We do not herein present the mapping history of the pit but focus on the highwall geology revealed in the drone photography.
Several 3-D models were constructed of both the entire pit and of the individual highwalls. The quarry pit models were used to track the mining progress to show newly formed interior highwalls (
Figure 6 and
Figure 7). Our geologic interpretation of the highwalls was based on our experience of mapping these same units in previous research [
4,
9]. The RMSEs for all eight 3-D models were within the accuracy of 2 cm and 12 cm, for horizontal and vertical, respectively. The 3-D models of the highwalls were converted into two-dimensional orthomosaic photographs that were used for geologic interpretation (
Figure 8,
Figure 9,
Figure 10,
Figure 11,
Figure 12,
Figure 13 and
Figure 14). The image spatial resolution (i.e. ground sample distance) for the highwalls was approximately 0.25 mm per pixel and the orthomosaic photographs (
Figure 8,
Figure 9,
Figure 10,
Figure 11,
Figure 12,
Figure 13 and
Figure 14) we present had minor holes where the overlap of the original images was not sufficient.
Photomosaics of five of the pit’s exterior highwalls and two interior highwalls were made (
Figure 7). Only two interior highwalls could be geologically mapped because most of the interior highwalls were covered by colluvium. During the mining, temporary berms were constructed immediately above the interior highwalls for safety purposes. Unfortunately, wash-over from the berms buried or obscured most of the in situ strata in the interior highwalls, so they could not be geologically mapped.
The exterior highwalls averaged 52 ft (16 m) in height and revealed the stratigraphy and structure of the pit area very well (
Figure 8,
Figure 9,
Figure 10,
Figure 11,
Figure 12,
Figure 13 and
Figure 14). The stratigraphy of the highwalls consists of tan Pleistocene loess (silt) overlying reddish brown Pliocene UC fluvial sand and gravel. The sands and gravels are essentially flat-lying river bar deposits with cross-beds (
Table 1). Only the apparent dip directions of the cross-beds can be obtained from the highwall surfaces but the cross-beds dip in all directions, indicating a full azimuth of paleo-flow directions that suggests a meandering flow pattern.
5. Conclusions
Drone aerial photography provides a highly effective imaging system for open-pit mining and research of the exposed geology. Aerial drone mapping of the Mississippi River’s Pliocene UC terrace in the Memphis Stone and Gravel active sand and gravel quarry has shown its utility in (1) recording mining progress, (2) identifying and quantifying sedimentary facies geometry, (3) estimating Pliocene Mississippi River paleocurrent directions using cross-bed apparent dip directions, (4) identifying geologic structures, including faults, fractures, folds, and liquefaction, and (5) preserving a digital stereoscopic record of the ancient Mississippi River alluvium that is removed in the mining process (
Figure 7,
Figure 8,
Figure 9,
Figure 10,
Figure 11,
Figure 12,
Figure 13 and
Figure 14 and
Supplementary Materials).
During the acquisition of the drone photography, no geologic structure was evident in the quarry highwalls. However, the orthomosaics of the highwalls provided an excellent mapping base wherein a fault became evident. The reverse faulting, folding, and liquefaction in highwall 1 demonstrates a deformation that we believe is tectonic (
Figure 8). An interpretation of tectonic faulting is preferred over geomorphic slumping because the fault has compressional reverse movement, drag folding in both the hanging wall and the footwall, cross-cutting relationships that demonstrate two periods of faulting and adjacent sediment liquefaction, which suggests coincident earthquake shaking. The fault in highwall 1 is interpreted to continue south and pass through highwall 4. Thus, this study identified northerly-striking late Pliocene or Pleistocene faulting that had not been previously documented in DeSoto County, Mississippi. Based on the relationship between fault displacement and earthquake magnitude, an 8-ft (2.4 m) fault displacement could produce a moment magnitude 7.3 earthquake [
23].