Climate Change | Management and Monitoring of Soil and In Situ Sensor Data as the Key to Process Understanding

The decision to reconfigure and straighten the route of the Emscher river in 1902 was to herald the start of a dismal future for the waterway as the industrial age went on to reach its peak. While the Ruhr and Lippe became fresh water streams for the Ruhr valley, the Emscher, which was more centrally located between the other two rivers, was relegated to the role of a sewage channel for practically the entire Ruhr area (1). As a slow-draining lowland river the Emscher had always encountered problems on account of its very small drop in height between source and mouth (1, 2). Increasing industrialisation meant a further deterioration in the situation until things became untenable. Channelled into pipes and placed in a concrete strait-jacket the waterway was euphemistically referred-to as a sewage ditch (3, 4) and essentially was no longer locally recognised as a river at all.

One of the main tasks that has been set for the post-mining era involves the near-natural restructuring of this river landscape – the EMSCHERGENOSSENSCHAFT/LIPPEVERBAND (EGLV), Essen/Germany, has devoted itself to this regeneration work. In an operation costing billions of euros (2, 5) the core feature of the project – the subterranean Emscher canal – is now nearing completion. The old surface watercourse was also restored and its confluence with the Rhine relocated further north in order to increase the rate of descent (5). The river was to be reconstructed step by step, and within the available space, to create a functioning river landscape (6) of the kind that existed prior to industrialisation.

While the river was forced to undergo a major transformation during the many years of industrial expansion, even today it is faced with a set of new challenges that threaten to upset the whole conversion project. Climate change has brought with it prolonged periods of drought followed by torrential rains and these events are now casting doubt over whether the natural restructuring project can deliver the hoped-for success. The restoration effort means ensuring that any impact on water levels and on the local flora and fauna is kept to an absolute minimum so that the landscape can recover and restore itself in a natural way. It is still too early to say whether there will be sufficient water input from the catchment area via the local watercourses to maintain this and how dynamic the process will be. The recent dry spells and heavy rainfall indicate that the momentum is likely to change in the years ahead. The exact pace of this is as yet hard to define, which is why the Research Center of Post-Mining (FZN) at TH Georg Agricola University (THGA), Bochum/Germany, is working with a number of partners, including the Munster-based company EFTAS Fernerkundung Technolo-gie-transfer as well as commissioning personnel at the EGLV, to develop an understanding of the -processes involved.

This paper picks up on some aspects of a project that was undertaken by the FZN. In order to enable a detailed understanding of developments and to provide the basis for upscaling to satellite level various in situ sensors were set up to measure soil moisture levels and the findings checked against the soil data. The initial aim is to assess the rate of discharge in the catchment area and then ultimately to provide conclusive satellite-supported, area-wide evidence of what is going on.

1  Introduction

The dry years of 2018 to 2020 and the torrential rains of July 2021 suggest that climate change is also having a growing impact on post-mining issues. For the EMSCHERGENOSSENSCHAFT/LIPPEVERBAND (EGLV), Essen/Germany, as the responsible water authority, it is therefore important to understand the processes that are under way within the area under its control in order provide a basic rationale for measures such as the near-nature reconstruction of the Emscher waterway. Projects of this kind should aim to derive maximum benefit for the natural river landscape and also for the local residents. A feasibility study “Climate change – management and monitoring (C2M2)” was therefore initiated in order to establish how an area-wide assessment could be made of fresh-water inflows from the Emscher catchment zone. Extremely useful satellite data were supplied here by the EU Copernicus Earth Observation Programme (7). The Copernicus Earth-observing Sentinel satellites provide a high degree of geometrical and spectral resolution and the data are made available free of charge under a European Community scheme. Unlike in previous programmes of this kind the data from each satellite pass are stored and then made available to users. The flights that the relevant satellites make over the Emscher region typically last several days. Sentinel 1 (8), which is a C-band radar satellite pairing, and Sentinel 2 (9), which is a multispectral satellite pairing, overfly the area rotationally every five days and supply an extensive body of data that enables early reaction to area-wide changes as and when they occur.

These satellite data are therefore important not only for assessing soil movements and humidity levels but also for identifying changes in vegetation and land use. The latter task, which has been entrusted to the company EFTAS Fernerkundung Technologietransfer, Münster/Germany, is only peripheral to the main topic under discussion here.

However, in order to be able to understand exactly what satellite data can reveal about the target area the Research Center of Post-Mining (FZN) at TH Georg Agricola University (THGA), Bochum/Germany, has undertaken to prepare and assess the in situ components using a multi-layered approach. To this effect a representative catchment zone was selected for the Emscher area. The Boye, which is one of the largest of the Emscher tributaries, seemed to offer a compromise in that it balanced an overall assessment of the Emscher with the effort required for the feasibility study. The in situ components that were installed at four different sites created a combined soil sensor system that was specifically designed for measuring soil moisture and temperature levels.

The original concept, which involved inserting a flying-drone level between the sensor system and the satellite level, could only be partially implemented. It was planned first to use a suitable upscaling process to raise the in situ results to drone level, this comprising both thermal infrared data (temperature) and multispectral data (plant health). However a crash involving one of the drones meant that this level did not become available until the end of the growing season, with the result that the intermediate level could only be created using conventional RGB images and had to be interpreted in purely optical terms. Nevertheless, the aerial photographs that were recorded on a monthly basis do give an impression of how the vegetation developed in the catchment area over the course of the year.

2  Working area

Four representative test areas were chosen within the Boye catchment zone (Figure 1). The Boye flows north west to south east on a central course between Gladbeck in the north east and Bottrop in the south west. It then has to be lifted from polder level to flow into the Emscher close to the Prosper coke works on Gladbecker Strasse (B224).

The test zones selected for the survey could never fully represent the fragmented make-up of the landscape within the Boye drainage basin, though they do give some idea of the scenic diversity of area. Given the nature of the question being posed all residential zones and industrial and traffic areas were omitted from the survey. One aspect of significant interest to emerge was that for much of its course the Boye has to be pumped from a polder that is linked to Prosper-Haniel colliery (shaft sites 9 and 10). In some sections – at its confluence and at the Hoheheidebach pumping station – the Boye is also fully enclosed in a pipe and so is not visible from the surface.

2.1 Nattbach

The Nattbach stream flows through the Natroper Feld conservation area (Figure 2) between two old spoil tips on Welheimer Strasse (Gladbeck).

It still has the V-profile of a typical Ruhr drainage ditch. The concrete shells have mostly been removed. This small Boye tributary was chosen for the survey project because of its position and profile.

The measuring equipment was installed on a south-west exposed slope.

2.2  Rainwater retention basin in the retention soil filter

Slightly upstream of the Boye, on Gladbecker Strasse L511 and south of the A2 motorway, the Emschergenossenschaft has set up a retention soil filter in the Bottrop municipal area (Figure 3).

This installation is connected to a rainwater retention basin that is favourably positioned as a “satellite test field” for the project. Here the sensors could be installed in accordance with the geometric resolution of the Sentinel 1 pixels and on a piece of ground specifically kept free for the purpose. In this case one satellite pixel can represent a single installation. The basin has been artificially created.

2.3  Boye headwaters

Even before the first pumps began to influence the flow of the Boye these upper reaches included a wetland area (Am Schleitkamp in Bottrop) that was regularly prone to flooding, and which now forms part of the near-nature restoration scheme (Figure 4).

This particular zone exhibits highly dynamic water levels and constitutes an area undergoing natural succession. In this case the sensors were installed from the slope down into the bed of the stream in order that the full sequence of events could be assessed.

2.4  Brabecker Mühlenbach

The last site to be selected was the Brabecker Mühlenbach, which is a communal stream running through Brabeck field between Rentforter Strasse and the A31 motorway (Figure 5). The test area comprised a small feeder stream for the Boye that at this particular point flows through agricultural land. The sensors were installed on both the north and south facing banks of the stream.

3  Methodical approach

In order to build an understanding of the processes involved soil moisture and temperature sensors were set up around the site and a number of soil samples taken. The latter then underwent particle size analysis and a humus content measurement. The survey sites were visited on a monthly basis in order to document developments (image database, drone flights).

3.1   Soil and temperature sensors

On 18th December 2020 six TEROS 11 sensors (supplied by Meter Group) and a battery-buffered and solar-powered ZL6 data logger with six ports were installed in both the Nattbach and the Boye headwater areas. The project also sought to establish whether less expensive sensors from the greenhouse technology sector could be used in this case to provide area-wide information on the relevant events. To this effect it was decided to try out TELID RFID sensors from the Micro-Sensys company, namely the battery-buffered TELID 354.02 (soil moisture) and TELID 312 (temperature) units. According to the manufacturer’s specifications the battery power supply should last for up to three years. These humidity sensors use similar operating principles in that they employ a capacitive moisture measurement system where the value output, or volumetric water content (VWC), is expressed volume percent (vol. %). While the TEROS sensors also take account of the ratio between soil volume and pore volume, the more simple TELID sensors merely record the capacitive conductivity without taking account of any additionally derived information. The two sets of measurements are therefore scaled differently. While the TEROS sensors output their maximum value ranges at 55 to 65 vol. % of VWC, depending on the soil conditions, the TELID sensors regularly display values of 100 vol. % of VWC.

The TEROS sensors were installed horizontally at a depth of about 5 cm so that the measurement results would not be compromised by shadowing and also in order to produce a better correlation to the Sentinel 1 C-band radar signals (Figure 6).

As the TELID sensors were only designed for one measurement value they were normally installed in pairs, with one soil humidity sensor and one temperature sensor. In order to obtain a good mean value, however, the satellite test field in the rainwater retention basin was set up as a triple sensor arrangement with two soil moisture sensors being used in each case (Figure 7).

As the operation aimed at drawing a comparison between the two sensor types the tests carried out at the Nattbach and Boye headwater sites on 18th December 2020 first focused on the TEROS sensors, which were installed along a transect running from the streambed up into the slope. A pair of TELID units were also set up in each case at two locations in the stream and at a further sensor site that was considered representative (Figures 8 and 9).

In the event of a further drought or dry spell the set-up on the streambed was designed to indicate the point at which the stream ran dry. This situation did not occur at any time during the measurement period.

As there was no IP certificate available it was not possible to predict how the TELID sensors would behave when placed in the water and operated long-term on the stream bed, in other words below the water level. Both units have in fact been troublefree to date (one year and counting).

Despite being anchored in place with large tent spikes the TELID sensor pair in the upper reaches of the Boye were flushed away and lost during the torrential rainstorm that occurred on 14th-15th July 2021. The TEROS sensor was also washed out of its fixing point, but was held by its connecting cable and could be re-installed.

There were further interruptions to the flow of measurements at the Boye headwater site as a result of disturbance by passers-by who in some cases unintentionally detached the cables and disrupted the transmission of signals to the logger station.

The delivery of additional TELID sensors meant that by 24th February 2021 the rainwater retention basin (Figure 10) and Brabecker Mühlenbach sites (Figure 11) could also be similarly equipped.

Essential maintenance work in the retention basin twice disturbed the sensors at the site during the measurement period and this further interrupted the measuring sequence.

For the sake of completeness it should be pointed out that the TEROS sensors not only measured soil temperature but were also able to record air temperature and pressure at the logger station.

3.2  Soil sampling

Soil samples were taken at all the sensor sites and, in addition, two further samples were taken in each case at the Nattbach and at the Brabecker Mühlenbach in order to take account of the surrounding areas. The sampling work was carried out using a Purckhauer soil augur in order to accommodate the different soil types, this being supplemented by a large-sized coring sleeve of 850 cm³ capacity that provided samples for soil moisture analysis under laboratory conditions – gravimetric determination of soil water content in accordance with DIN EN ISO 11461 (10) – along with the material needed for particle size and humus measurement (Figure 12).

The particle size analysis was carried out without aggregate destruction (humus, concretions such as limestone and manganese, dispersion etc.) and in derogation of DIN EN ISO 17892-4 (11). This unusual procedure was used in order that the readings from the TELID sensors could be better assessed. The operating mode of the sensors means that the way in which the sensor is in contact with aggregates, humus or roots has a decisive impact on the measurement value. The information on the mode of action vis-à-vis the pure quartz grains that were left over after the sample preparation cannot therefore show any correlation to the soils in the test area. On the other hand it has to be assumed that the sensors will certainly react to the difference between “heavy” soil and “light” soil, that is to say between clay-loam soils and sandy soils. A basis had to be created for this comparison. That is why the results differ at least in part from the findings of the Geological Survey and drill core 50 (BK 50), although they also closely resemble them.

3.3  Aerial surveys using flying drones

Two survey drones were deployed to take aerial photographs on a monthly basis in order to record the progress of development at the test sites. The vehicles in question were a DJI Phantom 4 RTK (real-time kinematics) and a DJI Mavic 2. The positional accuracy of the aerial shots taken by RTK drones is within 2 to 4 cm as they are able to use a correction signal from the Satellite Positioning Service (SAPOS) of North Rhine-Westphalia (12). In order to derive sufficient positional accuracy for the Mavic 2 unit a number of reference points and GCPs (ground control points) were also surveyed and recorded during the sensor installation process. The software products OpenDroneMap and Drone2Map were used for preparing the orthophotos.

The use of drone-mounted multispectral and thermal sensors, as originally planned, was only possible on a trial basis towards the end of the vegetation period following the procurement of replacement equipment for a sensor system that had crashed during a mission elsewhere. The data obtained in this case proved to be of little scientific relevance and are merely described in the publication without further analysis.

The resulting dataset comprises a usable RGB aerial image for each test area as obtained by monthly drone flights. This is discussed in more detail below.

3.4  Geographic information system

All geospatial data and results were incorporated in a C2M2 GIS (geographic information system) as the project developed. This provided a basis for an exchange of data with the client and with partners EFTAS. It also served as a data pool and geo database for further work. The core software was ArcGIS Pro and data exchange was handled via the ArcGIS online platform.

4  Results

4.1 In situ sensors

One of the first positive results of the project was that both types of sensor proved reliable in operation and visual correlation was established with the recorded aerial images. Confirmation of the general usability of the TELID sensors can be considered a particular success. Data analysis work, along with a subsequent assessment by partners EFTAS to the effect that without further processing steps there would be no exploitable correlations between the measurement data and the Sentinel 1 radar data, would therefore indicate that the situation requires further detailed investigation. The two factors at play here can be best understood by referring to the example at the Nattbach (cf. Figure 8). One sensor sited just 10 m away is on the bed of the stream below the water level, two others are regularly soaked through or even flooded completely, while a further three units installed on the slope, and completely exposed to the sun, react very quickly to any changes. The Sentinel 1 pixel ultimately encapsulates this situation with a single overflight, providing a reflectance value for the entire pixel grid within which the aforementioned dynamics are played out in fragmentary fashion. The same applies to the sensor data, which while being better correlated nevertheless tend to differ when it comes down to details. This is illustrated by the example given in Figure 13, which shows the measuring sequence for the triple sensor arrangement set up at retention basin position 3.

Here it can be seen that even sensors positioned barely 10 cm apart can still yield deviating values. This can mainly be put down to the irregular nature of the substrate. If the sensor touches a root or if it intersects with the crawl-track of a soil-dwelling organism different values will be obtained when the sensor data are being analysed and evaluated in detail.

An added factor as far as the radar data are concerned is that a thick vegetation cover developed at each of the survey sites during the course of the growing season and this made it practically impossible for the radar signals to penetrate to ground level. This can be seen in one of the images shown in Figure 14.

While the radar does have a chance of reaching the ground in the winter months, this is no longer the case during the vegetation period. Furthermore, the soil sensors react directly to short-term changes, whereas the satellite cannot record certain situations in the course of its flight schedule.

For this reason an attempt was made, as part of the upscaling process, to reduce the complexity of the soil-sensor results by way of greater generalisation. The data in the box plot diagrams were amalgamated for the calendar weeks using the data from two survey sites (Nattbach sensor position 5, Boye headwaters sensor position 6). The aim here was not to interpret each actual individual value but rather to make general statements over the course of the year and in terms of the issues being addressed to define those points in time when the water input from the tributary streams feeding the Emscher was either very limited or non-existent. Interpreting individual values for solitary locations would not serve any purpose here.

The box plots also confirm that the TEROS and TELID sensors showed a similar pattern right from the outset, even though the individual values could also differ significantly (cf. comments on the data scales).

Figures 15 and 16 show example recordings taken from the Nattbach. When interpreting the box plots it should be borne in mind that each stick diagram covers the entire range of values, some of which exhibit extreme variations, over a calendar week.

The box is defined as a contrast between the median und arithmetic mean. If the box is black then the arithmetic mean is greater than the median. This suggests a right-skewed statistical distribution of the measurement values in the relevant calendar week (the readings are mainly “underrated”). Conversely, a white box indicates a left-skewed distribution, as the median is greater than the arithmetic mean (the values are mostly “overrated”).

On the basis of this knowledge two interim conclusions can immediately be drawn from the curve plots:

1. Both types of sensor produce a similar trajectory for soil humidity over the course of the year, even though the actual values are differently scaled. The correlation coefficients are around 0.79 both for the arithmetic mean and for the median.
2. The TEROS sensors tend to produce an underrated picture of the measurements while the TELID sensors tend to overrate them (with some exceptions).

The picture for the data from the Boye headwaters is far less pronounced with the measurement period of almost four weeks not being fully covered due to the damage inflicted by passers-by on the data transmission line. A similar pattern for the sensors is discernible here over the course of the year. At the Boye headwaters site too the TEROS sensors tend to underrate the soil humidity levels while the TELID sensors are prone to overrating them.

The work of processing and comparing the information with the satellite data is still to be done and has now been held over to a follow-up project.

4.2  Soils

The soil samples more or less confirmed the picture of the test area as depicted in the NRW Geological Survey’s 1:50 000 scale soil map (13). The different soil types identified mainly comprised gley, brown-earth gley and similigley. One exception was the rainwater retention basin, which was not separately mapped in BK 50. The ground surface in this area has been fashioned by man and can be described as loosely compacted soil (or “kultisol”).

At the Nattbach it is easy to see man’s imprint and the small-scale disparities caused by anthropogenic alteration. In the immediate vicinity of the streambed (sensor position GlaS02, Figure 17 a) the profile appears to have been impacted by building work. There is no distinct Ah horizon present – just, at best, a thin litter layer with grasses and weeds growing through it (L).

To a depth of about 15 cm there is a highly marbled horizon with humus inclusions and oxidation features in some parts, which is designated as Gho. Below this horizon is a 10 cm-thick transition zone that though marbled and interspersed with rust-stained streaks also exhibits reducing, grey-green bleaching. The renaturation process has for some time enabled the fluctuating groundwater level to take effect here. Beneath this can be found a clearly reduced and bleached horizon without concretions, which is labelled as Gr. The authors are of the opinion that this is composed of anthropogenically influenced gley soil that has been marked by re-wetting.

The anthropogenic alteration caused by the old upgrading of the stream and the roadbuilding work above it is more clearly discernible at sensor position GlaS04 (Figure 17 b). This point is located on the slope of the former V profile of the Nattbach. A distinct humus overlay has not been formed and a litter layer is recognisable (approximately 1 cm). Beneath this there is an apparently raised layer of loose substrate that is designated here as yA. Alternatively, the horizon could be addressed as R (deeply furrowed, trenched). The designation Y/K probably better reflects the reality of the situation. The upper layers to a depth of about 6 cm gradually become enriched with humus and a number of rusty streaks are also present, though these are weakly defined. This is labelled a yAho horizon. The profile then ends simply with a pale sandy substrate with rusty brown streaks. It is not apparent whether these are the result of colour variations in the raised sandy substrate or are to be interpreted as oxidation due to water accumulation or the process of water flowing off the slopes. The horizon has been designated as anthropogenic IC, denoting terrestrial kultisol or loosely compacted soil (terrestrially anthropogenic soil.)

To supplement this, comparison samples were taken above the sensor installation on a grassy site close to the Nattbach (Gla01/Gla02, Figure 17 c). This location also contains a small rivulet and exhibits a humic topsoil that has probably been ploughed to a depth of 15 cm. Beneath this, and clearly differentiated, there is a distinctly marbled and 10 cm-thick Go horizon with pronounced oxidation features. Below this horizon can be found a significantly reduced and bleached layer, designated as Gr, though slight concretions and rust-stained streaks suggest that the conditions here are changing, due presumably to drainage processes. This Gr was extremely waterlogged between the 25 and 30 cm levels, the area beneath being wet. The soil is designated as anthropogenically altered gley that is now drier than it originally was owing to drainage activities.

As mentioned, the soils in the retention basin are designated as “kultisols” as they have been established artificially and have developed slowly through cultivation. Each of the samples exhibited a distinct humus layer – designated as yAh – that is developing similarly to that of the grassy site. Below each of these layers there is simply an even thicker lC horizon.

The Boye headwaters in turn feature gley soils in conformity with BK 50. Unlike those found at the Nattbach, however, the bleaching here is more pronounced and has the classic pale-grey coloration.

There was greater variety to be found at the Brabecker Mühlenbach. The transect started off in agricultural land in the north and touched two sampling points directly alongside the stream (north and south banks) before ending at a fourth sampling point south of the stream. Sampling point Bra01 (Figure 18a) lies to the north of the waterway in an area that for the last two years has been used for growing maize. Here a ploughing horizon almost 30 cm thick exhibits a humus content that decreases with depth and then transits into a Bv horizon. The substrate of the Bv horizon can be seen to contain light reddish-brown streaks that indicate some degree of transient waterlogging. This could also point to a transition to an Sw horizon of similigley soil. However, as the profile was dry (no waterlogging) at the time the samples were taken (end of meteorological winter) this possibility was dismissed. It was therefore designated as brown earth altered by cultivation (arable farming).

Gley soil is again present at the stream itself, though this had been much altered by ploughing. The profile on the south side of the stream (BraS02, Figure 18b) has a fairly simple structure. Here can be found a ploughing horizon (up to 20 cm deep) that has compromised the Go area (ApGo between 20 and approx. 25 cm). This is followed consistently to the end of the profile by a rust-stained horizon that is interfused with rust-red streaks, which is designated as Go. The water table is undeveloped. The profile is designated as gley soil that has been altered by cultivation (arable farming). There is one distinctive feature present on the north side right by the bank of the stream. Here a highly organic horizon was exposed in the profile at a depth of around 50 cm. This is generally designated as an O horizon, though it may also be associated with an old peat layer (H).

Profile Bra02 (Figure 18c) lies to the south of the stream in an area that had been covered with green manure the previous year and had remained fallow in 2021. This location includes a ploughing horizon nearly 30 cm deep that is consistently humus-based. The following horizon has distinct redoximorphic features. Here in particular can be found clear concretions around which the soil has become bleached. This horizon has therefore been designated as Sw. The opinion is that this is a similigley soil that has been altered by cultivation (arable farming).

Even though the particle size analysis was not carried out in compliance with DIN standards there was still a fairly good concordance with the data from BK 50. The analysis generally showed the presence of sandy silt or silty sand, some of which contained large gravelly pebbles (Table 1).

According to the soil appraisal BK 50 refers more frequently to loams and silts, which are then given the attributes “very sandy” or “very silty”. As the aggregate material must therefore be larger due to the absence of any pre-treatment process (cf. soil sampling), the clay fraction cannot be reliably determined using the selected method. For designation as loam, however, the clay fraction plays a fairly significant role, with the result that the discrepancy appears plausible.

The humus content is also important for the water retention capacity. In this respect the samples yielded widely varying results. As expected, the content found at the anthropogenically disturbed slope of the Nattbach was the lowest at 3 to 5 % (GlaS02 – GlaS06, Table 2), whereas the grassy location further upstream showed the highest values at around 15 %. The profiles investigated at the Brabecker Mühlenbach also exhibited high values where they had been homogenised by cultivation and ploughing. These values of between 8 and 12 % were still significantly higher than those for the Boye headwater site, which were fairly well fixed at 6 to 7 %. The comparatively high values found at the rainwater retention basin stood out from the rest and are likely to have been artificial enhanced by the building work.

Both factors – soil composition and humus content – are crucial for the water regime at each site. The findings obtained for the survey areas indicate that the water capacity tends to be fairly low as a function of the sand and silt content and the vegetation soon suffers from drought stress.

The main aspects of the water holding capability are defined by the air and field capacity (14). The field capacity is normally determined in the laboratory (pressure vessel) and represents the amount of water retained after gravitational drainage when a water-saturated soil has been left for two days. In this case the water content is determined gravimetrically. The general practice is to use a quantity of water that can be held against gravity at a pF value¹ of 1.8.

¹ pF = potential of the free energy of the water (non-dimensional)

Under laboratory conditions this value is set after water saturation in a pressure vessel. This is referred-to as retained water. All the water circulating in the larger pores of >50 µm can, e. g., be carried into the aquifer under the force of gravity and is then available for inflow into the watercourses.

This information can therefore be interpreted to mean that according to the scientific consensus in-soil water can only be transferred to watercourses and to the water table when the field capacity of the soil is exceeded (14). However, even retained water is not completely accessible to plants. Due to the adhesion and cohesion present in soil components and aggregates part of the water is so firmly embedded in place by the capillary forces that this “dead water” (in pores of < 0.2 µm) cannot be absorbed by the plants via their root system. The available portion is referred-to as the usable field capacity (uFC).

If the water content then reaches a level that can no longer be extracted by the plant from the in-soil water reserve the result will be drought stress. The in-soil water level reaches the permanent wilting point (PWP), which will happen at pF 4.2, and the plant reacts by suffering drought damage, which may be indicated by a reduced chlorophyll content.

As the Sentinel 1 radar data did not exhibit a high correlation with the direct measurement results in the first series of tests the authors assume that dry stress is potentially a means of obtaining an area-wide assessment. One approach could work by determining plant health and the relevant indices, such as the Normalised Difference Vegetation Index (NDVI) (15):

Here NIR represents the near infrared and RED the red channel from the RGB range. The reflectance values of the near infrared and red channel are therefore correlated in the NDVI, this being derived using multispectral data gathered by flying drones and area-wide Sentinel 2 data. If the plant goes into drought stress this will be reflected in these indices. The soil values are therefore crucial for recognising when this point has been reached. Table 3 presents the data for the survey site in accordance with BK 50.

Even though no pressure-vessel tests could be carried out as part of the project, the Soil Science Textbook (14) does contain information on how these values can be estimated from the particle size analyses (Figure 19).

If the results of the particle size analyses are transferred to a graph it becomes apparent that a high air capacity and a large macropore volume has to be assumed in each case. This means that the survey area can drain a high percentage of the rainwater into the receiving streams. However, only about 20 vol. % of plant-useful in-soil water remains in the sandy and silty soils of the survey area. The uFC range is already exhausted at a 10 % proportion of residual soil water. Between 10 and 5 % vol. % the soils then reach the PWP, with the result that at the very least those grasses and herbaceous plants that cannot be supplied from the ground water will react by suffering drought stress.

If the data obtained over the measurement period are merged with the depicted soil parameters (Figure 20) it can be seen that 2021 was a positive year as far as the supply of water from the Boye catchment area is concerned.

Most of the measurement results lay within a range of values that clearly suggest an involvement of the macropore volume, and so by implication the capillary forces were not able to fix the entire in-soil moisture reserve. Taking the Boye headwater sensor BotS06 as an example it can be seen that at no time was a value reached that might have been considered critical for the supply of water to the plants.

By contrast, it can be envisaged that in a dry year the in-soil stock of moisture will quickly become exhausted and levels will more frequently fall below the point at which water is not readily available to plants, i.e. the permanent wilting point. It has to be assumed that in such a year the macropores in particular will dry up more quickly and water will be retained in the drainage area.

4.3  Orthophotos

As it was not possible to implement the original plan for thermal and multispectral drone shots the survey sites were flown over on a monthly basis using an RGB camera. This at least allowed a visual assessment to be undertaken of the vegetation conditions (Figure 21).

Images of this kind provide an early impression of plant health, though no nearer infrared channel is available. The green band, as a replacement for the NIR, is used along with the red channel and the NDVI is calculated in the conventional way (here designated as “green NDVI”). Tests yielded plausible results for the phenological succession as far as vegetation development was concerned (Figure 22). Artificial landscape elements and dead trees were also clearly recognisable.

The multispectral sensor, which has its own red edge and NIR channel, offers a more practical solution. Such a device rather belatedly became available to the project in October and so could only be deployed on a test basis (Figure 23).

Sensors of this type are much better suited to tasks such as registering the aforementioned systematics, recording the impact of drought on vegetation and logging the decline in plant health due to drought stress.

More information on changes caused by drought can be obtained by applying infrared imaging technology. Here too the sensor only became available for use at the end of September following the crash of the original aircraft.

The image recorded on 28th September 2021 (Figure 24) e. g., and especially when compared with the multispectral recording in Figure 23, shows that the open areas along the stream heat up much more than the surrounding trees or the coppice-type vegetation in and around the soil sensors. This can also be clearly seen in the grassy areas in the south west of the infrared recording. It must be assumed that a detailed assessment of the entire area, as undertaken by two sensors, makes for much improved interpretation results.

5  Summary and outlook

The Climate Change – Management and Monitoring project, which was carried out within the given framework of a feasibility study, provided early information on methods that could help towards a better process understanding of water supplies to the Emscher river. The use of in situ components, as described here, and the derivation of the initial results and further research strategies all show that sensor technology is certainly conducive to a better appreciation of satellite data. The poor correlation found between the results of the radar evaluation of the Sentinel 1 data and the readings from the soil sensors clearly indicate that this kind of information cannot be interpreted with sufficient certainty unless the on-site conditions are properly understood. The challenges involved were therefore examined in greater detail and with the help of (derived) soil data a plan was drawn up that could enable an area-wide assessment of the issue at hand. The relevant pF curves are very closely linked to plant health and can be analysed using simple, well-proven indices based on multispectral data obtained from the Sentinel 2 sensor.

A multi-level system can therefore be proposed as follows:

• Sentinel 1 data can be used to derive soil moisture values for open land and areas with minimal vegetation cover.
• Using data from Sentinel 2 it is possible to assess plant health on an area-wide basis while allowing for phenological factors. Here it is assumed that areas that are further removed from any watercourses will react more quickly to drought conditions than those that are closer to a river or stream.
• Based on evidence-triggered systems, and in the event of an incident occurring, flying drones can be used to undertake a detailed assessment of the catchment area and can verify or refute the satellite data results in high resolution. It has been found that a combination of RGB, thermal infrared and multispectral sensors allows a comprehensive evaluation to be made.
• However, in situ data must be available if the findings are to be properly understood. It was demonstrated that soil investigations carried out with in situ soil moisture sensors were in fact conducive to a more accurate interpretation of the data gathered from the satellite and drone operations.

One of the indirect benefits of the feasibility study was that it additionally served as a test bed for determining whether simple, low-cost sensors were also suitable for obtaining evidence for upscaling and process understanding. It was demonstrated that the seasons could be captured very effectively by generalisation. However there is still a lack of understanding surrounding the practical scaling of the sensors. This problem is currently the subject of a bachelor dissertation at Georg Agricola University, which is being prepared in collaboration with the manufacturers as part of a research partnership. The work is due to be completed in early 2022.

The same applies to the question of the suitability of the “plant health” factor when it comes to conducting area-wide assessments. A master’s thesis is currently investigating whether there are certain plants or plant communities that could assume the role of indicators in this regard. It is quite conceivable that certain associations of plant species react more quickly to drought as their root systems are to be found in the upper soil layers. High temperatures and levels of solar radiation will use up the water resources more quickly in these areas, with the result that the field capacity of the soil will soon tend towards a permanent wilting point. The observation of these so-called indicator plants can in future give an early indication of the changes taking place in the catchment area and this can enable targeted measures to be taken.

The partnership must continue to pursue this line of research and should in future place greater focus on the opportunities offered by satellite image analysis.

Acknowledgement

The authors wish to thank colleagues at the Emschergenossenschaft for the support they have given to this project. They also want to take this opportunity to express their gratitude to Micro-Sensys for having responded so quickly and directly in providing sensors for the bachelor thesis. Thanks are also due to those members of staff at the FZN for their forthright personal support with the data collection work.