The Egyptian Nile Delta covers about 24 000 km² and is the largest delta of the Mediterranean Sea. Historic textual sources witness up to seven major Nile branches that flowed through the delta, while today only the Rosetta and Damietta branches exist (Fig. 1; Bietak, 1975). In antiquity, these waterways were of high significance for intra-Egyptian trade and traffic, and major ancient Egyptian cities are exclusively found in their immediate surroundings. Hence, the reconstruction of the Holocene delta environments is also crucial for studying the human–environment interactions of ancient Egypt (Butzer, 1976; Pennington et al., 2017; Bietak, 1975). For protection against the seasonal Nile floods, settlements were either built on Pleistocene sand mounds (“geziras”) or on the embankments of the river branches. However, due to the long-term dynamics of the riverine system, the landscape of the Nile Delta has constantly changed. Water courses have been silted up and are no longer visible in the modern landscape. Knowing about the importance of localizing the route of former Nile branches, several geophysical and geoarchaeological investigations were carried out in the past, of which some also rely on remotely sensed imagery, e.g. Ginau et al. (2017) and El-Fadaly et al. (2019).

Figure 1Comparison of cloud-free summer (July/August) median RGB composites of the Nile Delta: (a) MODIS (2001–2021), (b) Landsat (1985–2021), (c) Sentinel-2 (2017–2021), and (d–f) detailed views of the city of Zagazig. Ancient river channels of the Nile are drawn according to Pennington et al. (2017).

In continuation of these efforts, the processing of remote sensing time series of Landsat (including Landsat 9 and Collection 2 data), Sentinel-2 and MODIS was conducted using the cloud-based processing capacities of the GEE. For this purpose, we developed a script which allows the generation of cloud-free surface reflectance products and various spectral indices for a user-defined period, region of interest, and for the earth observation data of the respective sensors. The source code of this tool is freely available on GitHub (https://github.com/EricMoeller96/master_thesis, last access: 9 November 2022) and comes with documentation on the most important settings. The tool requires a minimum of user inputs for the execution and can be executed for a user-defined region. Exemplarily, the datasets, specified in Table 1, were processed to generate median RBG composites for the entire Nile Delta (Fig. 1). In addition, multispectral indices, such as the normalized difference vegetation index (NDVI), NDWI (Gao, 1996), and normalized difference snow index (NDSI), are calculated and processed by default. Following the approach presented in a preceding work (Ullmann et al., 2020), in this study we exemplarily focus on two median NDWI images which were calculated from the time series: one for the winter (January/February) and one for the summer (July/August) seasons and for each sensor. These NDWI images were then differenced (summer minus winter) to draw the long-term seasonal difference (ΔNDWI; Fig. 2).

Table 1Overview of investigated remote sensing datasets which were processed using the Google Earth Engine. Bands refer to B = blue, G = green, R = red, NIR = near infrared, and SWIR = short-wave infrared. Multispectral indices and products are as follows: NDVI = normalized difference vegetation index, NDWI = normalized difference water index, NDSI = normalized difference snow index, NBR = normalized burn ratio, and LST = land surface temperature. Composites of each band, index, and product were generated using the median operator; as such, a pixel shows the median value over the entire stack of all cloud-free acquisitions within the period of investigation.

Download Print Version | Download XLSX

Figure 2Comparison of ΔNDWI anomalies of the Nile Delta: (a) MODIS (2001–2021), (b) Landsat (1985–2021), (c) Sentinel-2 (2017–2021), and (d–f) detailed views of Geziret Sineita (Tell es-Sunayta) (see van den Brink, 1987). Ancient river channels of the Nile are drawn according to Pennington et al. (2017).

The GEE script allows cloud-free summer and winter mosaics to be processed for the entire Nile Delta in a fast manner and for all sensors (Fig. 1) using analysis-ready products (i.e. surface reflectance). Processing time in the GEE was about 10 to 30 min, which would not be achievable using standard computing facilities given the high number of scenes (e.g. more than 1900 scenes for Landsat).

Thus, the analyses are not limited to the spectral indices, but the multispectral information can also be utilized, which opens possibilities for further investigations apart from the differentiation of spectral indices, e.g time series analysis or land cover classification. The ΔNDWI images of all three sensors show corresponding positive and negative anomalies on the broadest scale (Fig. 2), which follow the general systematic outlined in Ullmann et al. (2020). As such, the largest and strongest anomalies likely display distinctive features of the general (palaeo-)environmental setting as sketched by Butzer (1976). For instance, strong negative anomalies of the ΔNDWI somewhat match the proposed location of sands at or near the surface at several locations between the modern course of the Rosetta and the Damietta branches and between Tanta and Cairo. In all datasets negative ΔNDWI values are found in the western Nile Delta (south of Alexandria and west of Damanhur), the central Nile Delta (south of Tanta towards Cairo), and in the northern Nile Delta (in the vicinity of Lake Burullus). Positive ΔNDWI values are less frequent, and the largest patches are found near Lake Burullus (north) and Lake Manzala (east). Obviously, the higher spatial resolution of Landsat and Sentinel-2 allows a more detailed picture to be depicted; in both datasets spatially varying positive and negative anomalies are found in the delta west of the Rosetta branch. Overall, Sentinel-2 data deliver a better geometric resolution revealing more details (e.g. as exemplified for Geziret Sineita in Fig. 2d–f); however, due to the rather short time series of 4 years (compared to 36 years offered by Landsat), anomalies are less clear, and the ΔNDWI image shows more heterogeneities compared to the results of Landsat. This is also displayed by the higher deviation used to scale the ΔNDWI images (Fig. 2). It is likely that this issue is linked to the short time series as seasonal differences in NDWI become best visible when long timescales are analysed (Ullmann et al., 2020). As such, a long observation period is important to identify rather weak anomalies in the ΔNDWI. Thus for now, the Landsat archive appears to be the most useful and promising record for identifying ΔNDWI anomalies associated with surface and near-surface discontinuities in soil properties. Taking this further, Fig. 3 shows preliminary results from the continued mapping efforts. The visual analysis of the Landsat ΔNDWI revealed several additional linear and meandering anomalies, especially between Damanhur and Tanta and in the surroundings of Zagazig. Some of them correspond remarkably well with the general flow directions and river routes indicated by Pennington et al. (2017). These anomalies could therefore indicate former courses of the Canopic Branch and Saitic Branch.

Figure 3Preliminary mapping results of ΔNDWI anomalies in the central Nile Delta based on Landsat imagery (1985–2021). Ancient river channels of the Nile are drawn according to Pennington et al. (2017).

Data relating to this paper are available from the corresponding authors upon reasonable request. The Google Earth Engine code is available at https://doi.org/10.5281/zenodo.7313130 (EricMoeller96, 2022).

The methodology was developed by TU and EM. The formal analysis was conducted by TU and EM. TU, EM, RB, ELA and JM interpreted the results jointly. Funding was acquired by TU and JM. The manuscript was prepared by TU and JM with contributions from all authors.

At least one of the (co-)authors is a member of the editorial board of E&G Quaternary Science Journal. The peer-review process was guided by an independent editor, and the authors also have no other competing interests to declare.

Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This article is part of the special issue “Quaternary research from and inspired by the first virtual DEUQUA conference”. It is a result of the vDEUQUA2021 online conference in September/October 2021.

This research has been supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; grant no. 507687060).

This open-access publication was funded by Julius-Maximilians-Universität Würzburg.

This paper was edited by Christian Zeeden and reviewed by Janek Walk.