Many studies have analyzed channel changes with aerial imagery (many examples are reviewed in Lawler, 1993). In this study we have used Landsat images instead because they are free to use and available worldwide.

Landsat images (from 1984 and 2016), collected from the archive of Landsat available on the Earth Explorer platform of the United States Geological Survey (USGS, 2017), are used to highlight and study the destruction of river banks. The images used for this study are coherent in terms of being collected during a sunny acquisition period (between June and August). Images taken during the sunny months show a very good contrast between image objects (Jensen, 1983) and this is fundamental in landscape-change detection studies in order to display the bare soil, water limit and vegetation cover. We have presented the band ratio composites 5/7, 5/4 and 3/1 for Landsat 5 TM (30 m) and 6/7, 6/5 and 4/2 for Landsat OLI (15 m), respectively, in red, green and blue, so that the vegetation cover (various agricultural fields) in the fluvial plain appears in red, the river bed in light blue and the stream flow in dark blue.

Previous studies (Gurnell et al., 1994; Downward, 1995; Petts, 1989; Winterbottom & Gilvear, 2000) indicated the usefulness of identifying channel changes using historical maps and photographs in a geographic information system environment. However, the typical GIS approach to planform changes has the problem that it is limited to two dimensions (Rhoades et al., 2009). Measurement of lateral erosion of rivers requires the third dimension of bank height, which is unavailable (Carroll et al, 2004). To overcome this problem, some authors (Rhoades et al., 2009) have used LIDAR data to calculate bank heights. However, airborne LIDAR data not only have a high cost but also are unavailable for many areas, including our study area, which is a major challenge. In order to solve this situation, in the present study, we have used high-resolution (12.5 m) terrain-corrected data (DEM) obtained from ALOS PALSAR: Advanced Land Observing Satellite Phased Array type L-band Synthetic Aperture Radar (accessed through ASF DAAC, https://earthdata.nasa.gov/about/daacs/daac-asf, November 16, 2017). The coordinate system considered in Morocco is the World Geodetic System 1984, Universal Transverse Mercato, Zone 29N (WGS_1984_UTM_Zone_29N). We chose this data because of its high resolution (12.5 m), its geographical coverage (free and available for most zones worldwide) and its quality (post-correction), which are the main characteristics by which to judge its adequacy for our particular need: to estimate river bank heights.

We used Landsat images (Fig. 2A and B) processed in the above section (remote sensing) to quantify the average volume of sediment released through bank erosion during the period of study (from 1984 to 2016). The two images are registered with the WGS_1984_UTM_Zone_29N coordinate system and that projection was maintained for all extracted data used in this study.

Landsat images from 1984 and 2016 were used to draw bank boundaries. They were digitized as lines over the entire length of the river. To simplify bank limit digitization, banks were delimited based on the location of agricultural land cover (Fig. 3A) using heads-up digitizing methods. The potential error caused by rectification and digitization may be minimized by creating a buffer of 2 m around each bank’s position (Rhoades et al., 2009). In order to identify bank retreat (formulas 1 and 2), we calculated the difference between the 1984 and 2016 bank positions (Fig. 3B). The intersection between bank positions also provides polygons, which are named erosional areas (Fig. 4). The area and length were then calculated for each erosional area. Moreover, the channel lateral migration can be estimated by dividing each erosional area by its length as in formula 1:

The width of erosional areas calculated by Formula (1) is divided by the period of study (32 years in this study) and then provides the rate of bank retreat (Rhoades et al., 2009) (formula 2):

A Digital Elevation Model (DEM, obtained from ALOS PALSAR) processed using a GIS (ArcGIS 10 .2) tool named “Add Surface Information” was used to estimate the river bank heights.

To estimate the bank heights of the Rdat River, more than 20 (20 to 30 depending on the area of each polygon) profiles were drawn over each erosional area along the entire length of the river (Fig. 5). In fact, the number of profiles depends on the erosional area lengths. The profiles are drawn so that they are more or less perpendicular to the banks, parallel to one another and evenly spaced (10 to 15 m). The altitude of each profile was then calculated by the “Add Surface Information” tool, which reports the following for each profile (Table 1): elevation (Zmax, Zmean and Zmin) and profile length and slope (maximum, minimum and average). Bank heights were then calculated based on the profile elevation Z as in formula 3:

where H_i is the bank height calculated at the point where the profile is drawn, Z_max represents the elevation at the top of the banks and Z_min is the lowest elevation representing the river bed surface. When the bank height has been calculated for each profile, the height of the whole erosional area is then estimated (formula 4):

where H_m represents the average bank height for the entire eroded area, H_i is the bank height calculated in (3) and n is the number of profiles. Also, to find the total volume (m³) of soil removed by river bank erosion in Rdat River (Fig. 6), the bank height (H_m ) is multiplied by the erosional area (formula 5):

Different field missions to different locations along the river were also carried out in order to visualize and analyze the problem of bank erosion in the field of study. In addition, to validate the obtained results, which is the main purpose of the field research proofs, bank heights were measured manually in the field and compared with those estimated by the GIS algorithm, using a correlation test (correlation coefficient between two variables), of the extracted and measured bank heights in order to define the state of the relationship between the two measured variables.

Digitization of the bank boundaries indicated that there are many (more than 100) individual eroding banks along the Rdat River from upstream to downstream. These erosional areas range between 93.16 m² (minimum) and 37311.94 m² (maximum) and provide an estimated total erosion of 537601.22 m² (Fig. 7). In addition, we noticed that further down the river course, eroded banks become more noticeable. The width of the eroded area becomes larger. Thus, the high peak of erosional areas is located at the lower reaches.

Comparing the total integration of the chord length (Fig. 8A), defined as the longest line within the polygon of erosion, of each erosional area with the total length of Rdat River, we concluded that at least 31% of the river length was eroded between 1984 and 2016. The width of erosional areas (formula 1 and Fig. 8B) is highly variable and reaches critical values (higher than the average, which is 109.8 m) especially in the lower part. The width of erosional areas reaches up to 217.26 m downstream.

River bank retreat rates along the 50-Km study reach over the 32-year study period are remarkably variable and range from 0.05 m yr^-1 to a high peak of 5.05 m yr^-1 (Fig. 9).

The observed bank heights in the field in the south of the river (downstream) were measured and compared to bank heights extracted using the correlation test (significance of correlation). The results, illustrated in Fig. 10, indicate that there is a significant positive relationship between the bank heights observed in the field and the extracted bank heights, with r (29) = 0.55 and P = 0.015. (R: coefficient of determination; P: p-value).

Bank height is highly variable from one erosional area to another. It ranges between 0.24 and 4.05 m with an average of 2m. Obviously, spatial variation in bank height tends to be important at Rdat River, with a large frequency of the higher values; 2 to 3.5 m in a total of 58 erosional areas and 1.5 m in 25 erosional areas (Fig. 11). We calculated the bank height for each eroded area, which made the volume calculated more accurate than that obtained by defining an average bank for the entire river.

The multiplication of the bank heights by the surface of eroded areas gives us the volume of sediment eroded along the studied river. It varies from 0.13 × 10³ to 94.12 × 10³ m³, with a mean of 12.33 × 10³ m³ and a total of 1159.33 × 10³ m³ (Fig. 12A and B). The volumetric erosion rates range from 3.05 m³ yr^-1 to more than 2188.95 m³ yr^-1, with an average of 286.82 m³ yr^-1. The volume of eroded sediment increases significantly from the upper to the lower reaches of the river: it starts to increase after 5 Km downstream (Fig. 13), while the highest peaks of eroded volume (Fig. 12A) occur around 30 Km downstream.

The considered changes in bank erosion rates are mainly located in the downstream part of the river, as mentioned by Lawler et al. (1999). In this study, the results also show that the erosion of river banks becomes more important downstream: the erosional areas become larger (Fig. 8B) and the volume of sediments released into the channel (Fig. 13) becomes higher compared to the upstream part of Rdat basin. The slopes (Fig. 14) play an important role in this process. In Rdat basin, for example, in the upper reaches, where the slopes are very high (40° to 65°), weathering processes tend to be concentrated in the mid-winter periods (Lawler et al., 1999). The slopes in the middle part of the basin range between 21° and 39°, while the lower course has very gentle slopes but the valley becomes flat and wide. At this point, many meanders have developed over time.

Field study indicates that the erosion is not the only process causing the degradation of river banks. We found that the landslide after a quick drop in the water level and the collapse may also affect river banks. In addition, when overbank flow occurs, considerable damage is detected due to flooding along the studied river including erosion of banks as well as damage to crops planted on the alluvial plain (Fig. 15A and B). This area is an important agricultural region for the residents who live along the river.

Landsat images helped us to define the river bank limits based on detection of changes in vegetation cover. GIS data allowed us to extract the maximum of the erosional area and to calculate the river bank retreat and heights along the entire Rdat River during the last 32 years. Subsequently, field measurements were used to validate the extracted bank heights based on our approach. The bank retreat rates reach a maximum of 5.05 m.yr^-1, with a large spatial variation in bank heights. The height of river banks is very variable, ranging from less than 1 m to several meters. Given these large numbers, we estimate that river bank erosion is widely contributing to the sediment supply of the studied river, with an average volumetric erosion rate of 286.82 m³ yr^-1. Unfortunately, there are no gauges on the river of Rdat basin to validate this value. Actually, our main motivation to study this river is to understand more this issue of river bank river erosion and its impacts in order to help the deciders make a move and maybe then we might have a chance to install a gauges in this river and manage this river.

The sediments released in the river when eroded from banks might have several environmental impacts. These can be a source of contaminated sediments; for example, a source of phosphorus (P) results from the loss of agricultural lands. Thus, the accumulation of sediments in the river may increase environmental problems (Roslan et al., 2017). Moreover, taking into consideration the important volume of sediments (1159.33 × 10³ m³ between 1984 and 2016) released into the channel, this study has also demonstrated that the sedimentary loads resulting from river bank erosion contribute to the filling of the river channel, which may lead to a narrow river section. When heavy rains arrive in the basin, the river will need to evacuate, under all and any circumstances, its liquid and solid flow. To do so, it needs more space, so it will adopt a new geometric shape and then erosion of banks and the phenomenon of flooding of the alluvial plain will increase.