IDENTIFICATION OF GEOLOGICAL AND ENGINEERING CONDITIONS FOR FORGING THE BED OF A LARGE LOWLAND RIVER, BASED ON LOW-CEILING AD HOC AERIAL PHOTOGRAPHS

: The article outlines the possibilities for assessing geological conditions related to fording the beds of large lowland rivers, based on low-ceiling ad hoc aerial photos, and the current state of the art on the utilization of remote-sensing methods in geological and engineering studies. The conducted research provided a confirmation that a set of remote-sensing sensors installed onboard an aerial vehicle enables recording images, the analysis of which allows determining the geological and engineering conditions of lowland riverbed zones, to the extent adequate to make decisions on crossing them


Introduction
Most bed sections of large rivers within Poland, especially in the Polish Lowlands, (such as the Vistula, Bug, Narew) are characterized by having relatively wide beds and depths rarely exceeding 2.5 m under conditions of medium and low water levels.This is, in many cases, the permissible fording depth for most modern structures [1,15].However, many sections of river bed bottoms under medium and low water level conditions (which are dominant throughout an average hydrological year) are made of soils with insufficient resistance, especially to dynamic loads.These include easily liquefiable (thixotropy) loose sands or other weak soils, such as organic soils [5,6].Therefore, it seems that ensuring full mobility of military vehicles requires equipping mechanized infantry sub-units with devices and procedures enabling fast identification and reconnaissance of zones/location convenient for river bed fording [4,22].
A crucial element that allows quickly obtaining remote-sensing data documenting the current state of the riverbed zone is the air segment.Important criteria in terms of selecting such a system include high time resolution, mobility, low sensitivity to adverse weather conditions and low imaging cost.Such requirements are currently met by both ultralight aircraft and many Unmanned Aerial Vehicles (UAVs).Both solutions, if equipped with appropriate image sensors, enable taking low-ceiling aerial photos of a quality sufficient to implement the assumed objectives.
Based on low-ceiling images in natural colours (RGB), taken by aerial systems with image recognition sensors, it was possible to identify elements/features of the Vistula and Buk riverbeds, which are important in terms of assessing geological and engineering conditions associated with their potential fording.Four basic groups of elements impacting bed bottom stability, bank stability, bed morphology and river bed process dynamics have been identified.

Research genesis
The experience acquired based on analysing the source literature, which addresses preparation, reconnaissance and fording of water obstacles, shows that this process is timeconsuming and requires the involvement of specialists (sappers, divers) [1,7].
The scope of identifying geological and engineering conditions is also set by the technical parameters of vehicles that currently are or will be in the nearest future owned by the Polish Armed Forces.
Fording is one of the most difficult combat tasks and requires special efforts by all troop types [4].According to Malinowski P. and Golczak Z. [4], the presence of additional factors complicating the course of action should be expected when fording water obstacles of medium width (up to 150 m), depth up to 3 m and mid stream speed to 1 m/s, which are typical for Poland.Besides reasons associated with their character of subunit specificity, these factors may also result from the geological structure of river valleys [10].
Most large river sections within Poland, especially in the Polish Lowlands (such as the Vistula, Bug and Narew) are characterized by relatively wide beds, with a depth rarely exceeding 2.5 m under conditions of medium and low water levels [1].This is, in many cases, the permissible fording depth for most modern structures [7,13].The option to quickly identify such zones within a riverbed can also be useful for many services responsible for crisis management, flood protection or border protection within sections made up of rivers [15].
Despite the large morphology variability over time, Polish Lowland riverbeds also contains stable zones.Bottom relief changes within them are minor and the bed is not deepened below a specified level.The alluvial bedrock of such zones contains soils with high washout resistance -hard-scoured formation, often in the form of thresholds.Fords often functioned in such areas in the past [12].Simultaneously to the stable zones, which in the morphology of a river valley are often distinguished by a significant narrowing of the floodplain area [8,9], the Polish Lowland river valleys contain zones where the bed is formed in low-strength formations -e.g., poorly compressed organic formations [10].Minor riverbed depth in such places does not guarantee effective crossing.
The diversification of the geological structure of such river valleys and beds within the Polish Lowlands results from the lack of geomorphological formation maturity, which is the consequence of the region's genesis [2,11,20] and valley polygenesis, therefore, the share of other morphogenic factors in their formation, other than riverine.In most cases, the Polish Lowlands rivers were unable to remove all traces of their primary genesis through erosion and accumulation [11].This fact promotes identification of zones with stable (and unstable) bed based on valley floor relief.Remote sensing methods are particularly effective in this regard [3,17,18].

Remote sensing methods applied in geological and engineering studies
Remote sensing techniques have been used in the geological/geological-engineering surveys since the 1930s.Mainly monochromatic air photography was used in the early days of remote sensing.Phototonal diversity of their image (shades of grey) rather accurately reflects the diversity of surface formation moisture.This feature was used to determine the outcrop boundaries of individual lithological types of rocks/soils.
The development of photogrammetric techniques broadened the possibilities of geological cartography with recording land relief features.This enables precise identification of such geological environment elements that proved the specific morphodynamics of an analysed surface [3].Similar possibilities in terms of identifying Earth's surface relief were created by the dissemination of radar images (obtained from SAR).
Another impetus behind geological remote sensing development was the emergence of satellite images.Herein, photographs were taken over various spectral ranges, and the common analysis of their images, also computer-aided, enabled identifying outcrop boundaries or tectonic structures that are hidden under several dozen meters of eluvia or other contemporary sediments.
The next stage in geological remote sensing development is associated with a significant improvement in the spatial resolution of satellite images.The availability of digital VHR (Very High Resolution) satellite images to civilian users is growing.They provided interpretative capabilities comparable to classic aerial photos, but are not burdened with many of their disadvantages.An increase in spatial resolution entails a development in the technology associated with simultaneous recording images on very many (often several hundred) spectral channels/ranges.The emergence of multi-spectral or highresolution photographs, also multi-spectral or hyper-spectral images, not only contributed to increase precision in the identification of the Earth's surface lithology, but, moreover, created the possibility to conduct geochemical remote sensing.
An excellent identification tool, often used for identifying very subtle features of Earth's surface relief turned out to be aerial laser scanning (ALS) [3,18,23].
All of the aforementioned remote sensing methods enable recording a static image of the Earth's surface.This results from their low temporal resolution and the high costs of acquiring remote sensing data.Such a situation significantly limits the possible documenting of geological and engineering conditions in zones with greater dynamics of morpho-and lithoforming processes.These drawbacks are not exhibited by terrain fact recording methods based on ultralight aerial vehicles (UAV) and unmanned aerial vehicles (ULAV).They allow conducting a comprehensive observation series using various types of sensors -among others, digital cameras continuously recording within the visible light and IR spectra.The application of such remote sensing methodologies is particularly suitable for the observation of bed zones in lowland river valleys.The specificity of the Polish Lowlands river bed zone environment is the large variability of their morphology, and the bed bottom relief, which results from the river hydrological regime, shaped as a result of climate changes and economic human activity [11].

Research subject
The research subject matter was a river understood as a natural watercourse, with a shaped bed and flowing under the action of gravity in the bed and valley, gouged due to erosive forces [16,19].
The study covered selected valley fragments of two large lowland rivers -the central Vistula and the lower Bug.
In the case of the central Vistula, the authors decided to select two sections of a different nature that exhibit diverse dynamics of contemporary bed processes resulting from the difference in their geological structure.This was a section of gorge nature -a fragment of the Małopolska river gorge near Solec nad Wisła (SOLEC section) and a Vistula section of a lowland river nature near Gołąb and Dęblina (GOŁĄB section).The 16 km long GOŁĄB section covered a Vistula valley fragment from km 379 (near Gołąb) to km 395 in the Dęblina area.The 15 km long SOLEC section covered a Vistula valley fragment from km 326 (near Piotrawin) to km 341 in the Iłżanka estuary area [11].
In the case of the lower Bug, the study covered the easternmost fragment of the Lower Bug Valley [48] that is approx.20 km long (km 80 -100).Within this section, the Bug is of a lowland river nature with a meandering bed development type and a course close to latitudinal.Within the analysed valley fragment, detailed research covered two selection sections -MAŁKINIA and BROK.The first -a 4.5 km long MAŁKINIA section (km 94.5-99) is located in the eastern part of the study area.It covers a fragment of the bed zone from the road bridge in Małkinia to Podgórze, where an approx.900 m Bug segment undercuts the upland edge.A railway bridge is located more or less in the middle of the MAŁKINIA section.In addition, there is an IMGW-PIB (Institute of Meteorology and Water Management -National Research Institute) water gauge post.The data from this post allowed to precisely determine water levels in the bed and flow rates during field research.The latter section -BROK, which is approx.3.5 km long (km 83.5 -87) is located in the western parts of the research area.It covers the part of the Bug riverbed zone from the estuary of the Brok river, to the railway bridge in Brok [21].
The sections selected for the research represent primary geological structures and morphodynamic types for riverine valleys of fluvial origin (erosive and accumulative).

Remote sensing work methodology
The research included three stages.The first stage involved selecting test sites.Location-related decisions were based on the methodology of identifying the stability of Polish Lowlands rivers formulated in the work by T. Falkowski, which concerns natural riverbed stability [9].The identification of zones that are potentially suitable for fording rivers was based on analysing the flood plain surface relief.The authors were particularly looking for locations where contemporary freshet erosions traces were concentrated on the surface of the flood plain.Such formations indicate the lack of any possibility to expand the bed cross section within a given fragment through lowering the bottom level and frequent damming of water when a freshet wave is passing or due to the formation of an ice jam [16,21,47].Such a relief of a flood plain belt (so-called proximal flood plain; [24]) adjacent to the riverbed may suggest morphological bed stability, which means the presence of hardto-wash-out alluvia substrate surface in the culmination bed [9,10,20].
The study involved taking photos/images using an ultralight aircraft and an unmanned aerial vehicle.One of the most important aspects of the in-house studies was to prepare for analysis, the remote sensing materials acquired by a Flying Laboratory of the Air Force Institute of Technology (Fig. 1) [15,23].
Source data comprised of video footage in natural colours, recorded by a digital HD camera.The first stage of the work was to review the video material and select frames where the Vistula riverbed zone image within field test covered areas was recorded.In the case of particularly interesting sections, the authors decided to develop photo-sketches made up of several (from 4 to 12) image frames.For this purpose, individual frames were combined and then subjected to tessellated and saved as .TIF files.The files were then uploaded to the GIS database and calibrated based on permanent elements found in the Vistula riverbed zone.A digital surface model (DSM) was used for calibration due to the fact that Lidar measurements (ALS) were taken for both analysed sections at a very low water level.Elements of the riverbed hydrotechnical structures (heads and groins), as well as (to a smaller degree) anthropogenic elements in its direct vicinity (buildings, power lines or ferry crossings) recorded on both types of remote sensing materials turned out particularly suitable for calibrating aerial materials.A spatial resolution of 0.15 m was obtained for the recorded materials [57].In addition, the remote sensing work included analysing thermal images of the riverbed zone acquired during research flights of the AFIT Flying Laboratory.
After analysing both types of remote sensing materials -a video footage in natural colours and a video taken with an IR camera -it was concluded that thermal imaging was characterized by a significantly lower spatial resolution than an HD video in colour, which results from the form of recording (lossy compression) data acquired by this remote sensing detector type.This is why, to identify the geological and engineering conditions in the riverbeds of the Vistula and Bug, the authors decided to mainly use HD materials in natural colours [3,18].
The basic material for remote sensing of the riverbed zone consisted of photo-sketches based on videos taken in natural colours by the AFIT "Flying Laboratory".Due to the nature of conducted work -the need to identify bed formations of complex morphology, as well as the nature of land relief formations located at various depths under water -it was decided to implement a manual classification method, which is deemed the most accurate, despite being very time-consuming.
The remote sensing work mainly appertained to identifying emerged and submerged riverbed formation, elements of riverbed hydrotechnical structures or their remains, course of the mainstream line, as well as elements that could indicate the stability of these facilities over time, e.g., vegetation therein.The authors also documented riverbed processes impacting geological and engineering conditions, as well as features directly indicating their course of occurrence.In case of doubts, auxiliary material (additional information source) was based on LIDAR data (NMPT), topographic maps and orthoimages from a national geoportal (www.geoportal.gov.pl).However, the latter exhibited limited suitability due to the fact that they were obtained at high water levels in the riverbed [15].
The remote sensing study also involved an attempt to assess the geological and engineering conditions significant from the perspective of river fording.The most important included the type and degree of compaction in the sediments making up the riverbed bottom and depth.This interpretation was verified based on the results of bathymetric measurements, geological soundings and field charting conducted at the same time [14].

Research results
Low-ceiling photographs of the Vistula and Bug riverbed taken in natural colours (RGB) and IR) enabled identifying the following elements of the riverbed geological environment (geological and engineering conditions), important in terms of evaluating their potential fording (geological and engineering forecast): − Bathymetry and bed dynamics: • Midstream line layout (Fig. 2); • Whirl occurrence zones (evorsion -Fig.3); • Zones of water speed deceleration and stagnation (associated with the presence of weak soils); − Diversification of alluvial soil load strength; • Weak soil occurrence zones; • Zones of non-cohesive compacted and moderately compacted soils; − Zones on soils susceptible to thixotropic liquefaction within riverbanks -zones of intensive groundwater inflow to the riverbed.The mainstream zone exhibits the darkest photons in the images.High flow velocities cause specific deposition in this zone.The photographs contain identifiable streaks of sandy sedimentation in the form of elongated, minor-height outwash or sandy streaks.The outwash is shown against a background of dark, sometimes grainy phototone that can be interpreted as bed pavement outcrops (Fig. 2).
The evorsion background area is usually surrounded by a narrow, circular sandy deposition zone.Clear sandy streaks can also be identified near clusters of small whirls arranged into lines (Fig. 3).
Thermal images of the bed zone using the AFIT Flying Laboratory were acquired simultaneously to the natural colour images.The data in the form of video footage with a 320x240 resolution in a colourful, unnatural colour scale ("iron" palette) was recorded using a thermal imaging camera.Despite the use of a relatively simple detector with non-cooled, low spatial resolution sensors, the obtained images show they have the potential to be used as input data.The analysed thermal images can be an important supplement to the obtained natural-colour data, since they enable identifying bed zone features invisible within the visible light band.Figure 4 shows a fragment of the Bug riverbed zone recorded with an IR camera (A) and a natural colour camera (B).The thermal image clearly shows inactive overflow beds (1), poorly visible under natural colours and a sandy outwash (2), which is visible on the natural colour image, while being practically invisible in the IR image.
Simultaneous acquisition of both types of images is advisable.They should be recorded in the form of single frames of a remote sensing image (photos) by sensors of possibly similar technical parameter of the lens (the observation angle and focal length, in particular), which will enable their even faster and more precise comparison.

Conclusions and summary
In the low-ceiling, natural-colour (RGB) and IR images of the Vistula and Bug beds, the authors identified elements of the riverbed geological and engineering conditions significant in the perspective of assessing its fording possibilities, such as bathymetry, riverbed dynamics (midstream line layout, evorsion zones, water stagnation zones), variable load strength of alluvial soils (delta depositions zones, outwash, stabilized islands) and areas of weak soil occurrence (including zones of intensive groundwater inflow to the riverbed and the inflow of groundwater indicating their possible liquefaction impacted by dynamic stress).
The conducted research enabled developing initial assumptions for the methodology of an aerial system for identifying geological and engineering conditions in lowland riverbeds, and assumptions for constructing a photointerpretation key for riverbed zone features indicating the presence of geological and engineering conditions that favour fording locations.
The option to record riverbed conditions with a high frequency (temporal resolution) will enable identifying the dynamics of fluvial processes and forecasting river behaviour under various hydrological conditions.That information may be useful for assessing the safety of flood protection structures and other elements of the riverbed hydrotechnical development.The developed method also has the potential to be applied in crisis management.
Identification of geological and engineering conditions of fording water obstacles should be based on a quick and relatively simple data acquisition method.Such information should be of indicative nature.The catalogue of features identifying the geological and engineering conditions can be relatively easily transformed into an interpretative key that can be used even by people without thorough geological knowledge.

Fig. 1 .
Fig. 1.Flying Laboratory of the Air Force Institute of Technology[23]

Fig. 4 .
Fig.4.Bug river bed zone recorded with an IR camera (A) and a visible light band (B)[15]