Tracking Environmental Changes in Manzala Lake Integrating Landsat Data Based on Interpolation Nonlinear Models

Study Area

Manzala Lake is the largest coastal lake in Egypt, a shallow brackish lake extending between the Damietta Nile River branch and the Suez Canal with a maximum length of 50 km along the vulnerable Mediterranean coast. Manzala Lake is located at the northeastern part of the Nile Delta, a sandy beach ridge separate the lake from the sea. While it connected to the Mediterranean Sea through Boughaz El-Gamil inlets 1 & 2 (Figure 1). It is typically located between longitudes 31° 45ʹ and 32° 22ʹ E and latitudes 31° 00ʹ and 31° 35ʹ N. Km2. The lake is shallow, typically ranging from 0.7 to 1.5 m in considerable depth. It connected to the North by the Mediterranean Sea, East by Suez Canal, South by El-Salam Canal and West by Damietta Branch of River Nile.

Click to enlarge

Water samples collected at two periods (2016 & 2019) and are analyzed according to Table 1.

Remote Sensing Applications

Data Collection: Three Landsat images were acquired using http://glovis.usgs.gov/, Path: 176 and Row: 38. Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+) and Operational Land Imager and Thermal Infrared Sensor (OLI/TIRS) images were acquired on 1992, 1994, 1997, 2000, 2003, 2006, 2009, 2014, 2017 and 2019, respectively have been selected for mapping and detecting environmental changes of Manzala Lake (Table 2). A field investigation was applied universally and instantly confirmed typically using a topographic map with scale 1: 50,000.

Table 2: Characteristics of Landsat images from 1992 to 2019. Image pre-processing and enhancement: Preprocessing applications as; atmospheric correction, radiometric calibration, geometric correction was typically applied to all images for further interpretation and enhancement using ERDAS Imagine. The geographical coordinates are unified by the latitude and longitude coordinate system WGS_1984_UTM_Zone_36N, and the projection mode is the horizontal axis Mercator projection. After images correction, Principal component analysis (PCA) was applied to reduce interferences for sustainable development of change detection. Image Classification: Maximum likelihood supervised classification was applied as the most accurate method in this present study to instantly detect different features of Manzala Lake. Supervised classification technique can studiously avoid digital problems resulting from the fundamental difference in the sensor and the atmosphere. Land Use and Land Cover: Land use and land cover are one of the most important natural resources. Remote sensing is one of the most important mechanisms of monitoring changes at different time periods. Vegetation cover was applied using Normalized different vegetation index (NDVI) as equation 1. While water surface area was calculated from normalized different water index NDWI (equation 2) according to the Modeler Function in ERDAS. The values of these indexes varies from -1 to 1, based on the vegetation density and water cover areas, and resembles high when the value close to one. NDWI was applied using the ERDAS molder to sufficiently reduce any noise pixels [9]. It's difficult to carefully distinguish between built-up areas and bare lands in any area; therefore, an index to map bare lands in the selected study area. The bare area can be detected using normalized built up area index (NDBI) (equation 3). The appropriate methodology comprised three arithmetic computations [10].

𝑁𝑒𝑎𝑟 𝑖𝑛𝑓𝑟𝑎𝑟𝑒𝑑−𝑅𝑒𝑑

NDVI=

𝑁𝑒𝑎𝑟 𝑖𝑛𝑓𝑟𝑎𝑟𝑒𝑑+𝑅𝑒𝑑 (1)

Green−Near infrared

NDWI =

Green+Near infrared (2)

SWIR−NIR

NDBI =

SWIR+NIR (3) Change detection analysis: Change detection determines the place and the considerable amount of fundamental change in a certain land cover at different times. Most of the change detection techniques rely on images classification, whether supervised classification or unsupervised classification. In this study, drilling and drying were applied to averaged time images between 1992 and 2019, to identify the occurred changes in Manzala Lake within different time. Chlorophyll a concentration according to MODIS- Aqua: The data of chlorophyll from 2002 to 2019 is available as the site of Giovanni v. 3.41. Acquired available data were selected in different years of 2005, 2008, 2012, 2015 and 2018 in 12 months. The selected region involved Manzala Lake with coastal area nearby the lake with coordinates of (31.7725 E, 31.0666 N and 32.3822 E, 31.5417 N). This data acquired from https://giovanni.gsfc.nasa.gov/giovanni/.

Statistical Models

Regression lines are helpful to know what interstitial values we could not practically know. It is useful to predict future values theoretically and can be trusted in statistical ways. The main advantage of these models is that they allow the reproduction of large-scale systems and long timescales. Statistical models vary from simple linear to complexity nonlinear depending on the nature of the phenomenon to be modeled. Four models were tested in this study for prediction of environmental changes: Linear, quadratic, Exponential and logarithm using Origin software 8.5. Linear model: Linear model is used in different ways according to the context. The most common occurrence is in connection with regression models, and the term is often taken as synonymous with linear regression model. They can help you understand and predict the behavior of complex systems or analyze experimental, financial, and biological data. Linear regression is a statistical method used to create linear model.

A quadratic regression: It is the process of finding the equation of the parabola that best fits a set of data. Exponential model: It is exhibited when the change the change per instant or unit of time of the value of a mathematical function of time is proportional to the function's current value, resulting in its value at any time being an exponential function of time. Exponential decay occurs in the same way when the growth rate is negative. A logarithmic Model: It is a model that measures the magnitude of the thing it's measuring. It can also be seen as the inverse of an exponential model. Assessment Methods: The accuracy of spatial interpolation models was assessed depending on three Statistical criteria: Root Mean Square Error (RMSE), coefficient of determination (R2) and residual sum of squares (RSS). The model can be validated with low RSS, low RMSE and with R2 value near to one. RMSE: It is used as an essential indicator that assesses the accuracy of spatial analysis in geographic information system and remote sensing (Lower (RMSE) is mainly to give most reliable values in the areas with no data. The minimum (RMSE) calculated by Cross Validation can be used to find the best spatial interpolation model control parameters [11]. The root mean square error (RMSE) was calculated for each model prediction using the formula:

RSS: it is also known as the sum of squared residuals (SSR) or the sum of squared errors of prediction (SSE) is the sum of the squares of residuals (deviations predicted from actual empirical values of data). It is a measure of the discrepancy between the data and an estimation model. A small RSS indicates a tight fit of the model to the data. It is used as an optimality criterion in parameter selection and model selection.

$$ \mathrm {R S S} = \sum_ {\mathrm {i} = 1} ^ {\mathrm {n}} \left(\mathrm {f} _ {\mathrm {i}} - \mathrm {O} _ {\mathrm {i}}\right) ^ {2} $$ i=1 Where n is he number of predicted value and 𝑓𝑖 denoted to our estimation of observed value𝑂𝑖. And so, 𝑂̅ and 𝑓̅ are mean values of observed and estimation respectively.

Lab Analysis

Click to enlarge

Table 5: Comparison of study results (2016 and 2019) with other different studies in other years. Trans: Transparency of Secchi Disk, DO: Dissolved Oxygen, Temp: Temperature, EC: electrical conductivity, Sal: salinity, TDS: Total Dissolved Solids, Chl: Chlorophyll According to the topographic map and Google earth investigation, supervised classification was typically applied to show the unique features of the study area. In the present study, three classes were gently applied in 1992 and 2019. It was correctly classified into vegetation, water and bare/sand areas (Figure 3). The purpose of supervised classification is to assess the spatial distribution of each land cover class. Lake Manzala served as a significant source of biodiversity, but changes due to anthropogenic impacts have reduced the biodiversity of the lake. Remote sensing applications were applied for studying and detecting environmental changes. Table 6 shows the percentage of water, vegetation and bare/sand areas from 1992 to 2019. For vegetation cover, NDVI was other hand, bare lands have been decreased from 28.19 to 19.32 % (Figure 6). The marked decrease in bare land is a good indicator for the biodiversity of the lake and bird migration. This decrease in bare lands is probably due to increasing the area of vegetation cover. This decrease is due to continuous land reclamation projects along the southern and eastern shores of the lake and fish farming processes. Bare lands were like a result of the clearing of the lands for agricultural and building purposes. The gradual decrease in bare land is a good indicator for biodiversity of the lake and bird migration. This decrease in bare lands is probably due to increase tremendously the area of vegetation cover [20]. The government of Egypt drained substantial portions of the lake in an effort to convert its rich Nile deposits to farmland. The lake which was the biggest coastal wetland along the Mediterranean coast is moving toward its disappearance by two opposite forces; one of them is the shrinking of the water body by siltation of sediments typically coming from agricultural lands and the abundance of weeds and swamp vegetation as well as the drying practices for agriculture, whereas the other force incorporates the removal of the coastal sand bar carefully separating the lake from the Mediterranean Sea by erosion, which should eventually lead to the conversion of the lake into a coastal embayment instead of being a closed coastal lagoon. Land reclamation has also reduced the lake surface by half, and, despite the declining quality of life and standards of living near the lake, human populations are increasing, exacerbating the lake’s problems. According to change detection from 1992 to 2019, there are some gradual changes in this modern period (Figure 7). 70% of the total area not changed. Some drilling and drying processes applied in this period as a percentage of 12% and 14 %, respectively. According to vegetation cover increase, vegetation appears prominently in this period as a percentage 4%. According to pixel count and size two models for water and bare/sand areas were applied using ArcGIS 10.5. These models were assessed to predict the change detection from 1992 to 2019 (Figure 8). It showed that the coefficient of determination was approximately 0.92 which is close to 1.

Click to enlarge

Click to enlarge

Click to enlarge

Click to enlarge

Click to enlarge

Click to enlarge

Click to enlarge

Click to enlarge

Click to enlarge

The data of chlorophyll recorded by MODIS-Aqua aid in the identification of area-averaged of Chlorophyll a concentration monthly. Also it aid in the knowledge of high and lower pollution impacts within different years. It is noticeable that the lowest concentration in selected years was at September, 2018 with concentration of 2.411 mg m-3, while the highest concentration was at November, 2012 with concentration of 8.675 mg m-3. Also the lowest mean annual concentration was 3.725 mg m-3 in 2018 and the highest mean annual concentration was 4.372 mg m-3

Click to enlarge

Statistical Models

To compare the spatial interpolation models and the extensive selection of the best, the cross-validation technique provided by Origin software 8.5 was used to judge statistically the accuracy and statistical equations for different models were represented for accurate prediction of environmental changes in the study area as shown in Tables 7 to 10. The highest R2 value and the lowest RMSE and RSS values of the quadratic model occurred. It was found that quadratic model is the best method to accurately predict the status of environmental changes in Manzala Lake. Quadratic model is the best model to obtain high quality in the proportional representation of changes data and typically produce the best equations for water, vegetation and urban in the study area. The results also indicated that the Exponential model is the lowest statistical model in the accuracy the prediction of environmental changes, and the models in terms of preference in the following order: quadratic > logarithm > Linear > Exponential. According to the sufficient accuracy of these models, the quadratic model was selected for accurately assessing the environmental changes in the study area. By applying the best equations for water, vegetation and bare areas, the lake can be expected for 20 coming years as show in Table 10. The change in the area of the lake will be stopped when the sum of the surface area of the water lake with the area of the urban area equal to the total area of the study area. This naturally occurs according to these statistical models at the end of 2042, and the Green areas are almost completely disappeared in the study area. After that, the models fail to describe the change in the area of the Lake. The results in the table 10 show that the absolute error in these reasonable expectations is less than 0.1% (Figure 10).

Click to enlarge