Monitoring Bakhtegan wetland using a time series of satellite data on the Google Earth Engine platform and predicting parameters with Facebook’s Prophet model

Subject Areas : Natural resources and environmental management

Mohsen Dastaran ¹ (MSc. Student of Remote Sensing and Geographical Information System, Faculty of Earth Sciences, Shahid Beheshti University, Tehran, Iran)
Shahin Jafari ² (MSc. Student of Remote Sensing and Geographical Information System, Faculty of Geography, University of Tehran, Tehran, Iran)
Hossein Moslemi ³ (MSc. Student of Remote Sensing and Geographical Information System, Faculty of Humanities, Tarbiat Modares University, Tehran, Iran)
Sara Attarchi ⁴ (Assistant Professor, Department of Remote Sensing and GIS, Faculty of Geography, University of Tehran, Tehran, Iran)
Seyed Kazem Alavipanah ⁵ (Professor, Department of Remote Sensing and GIS, Faculty of Geography, University of Tehran, Tehran, Iran)

Received: 2021-07-26 Accepted : 2021-09-26 Published : 2022-10-23

Keywords: Prophet prediction model, Bakhtegan wetland, Mann-Kendall test, Google Earth Engine,

Abstract :

Background and Objective Wetlands are habitats for vegetation and wildlife and because of this, they have a high environmental value. Also, wetlands reduce soil erosion, restore aquifers, store rainwater in a flood event, and provide water for agriculture or livestock. Wetlands are vulnerable to human interventions and changes such as drainage, urban sprawl, infrastructure development, and over-exploitation of groundwater resources. Prediction of the condition of wetlands in the future requires a correct understanding of the evolution of wetlands and identifying their trend of change. Nowadays, Remote Sensing technology is widely used for mapping wetlands, and its ability to monitor the changes in wetlands regardless of the diversity of wetlands has significantly increased the value of this science in this field. Remote Sensing can be an effective means of simulating and predicting wetland degradation processes by providing images at different times and through dynamic spatial modeling. In this study, the changes in the Bakhtegan wetland have been monitored. This wetland has high environmental and tourism importance and its drying affects negatively the living conditions and health of local people as well as tourism in the region. In addition, predictions of precipitation parameters, groundwater level, and temperature have been conducted. For this purpose, the Google Earth Engine platform was used to capture and process images. Google Earth Engine is a platform that can capture and process images in the shortest time and at high speed. In this regard, using Google Earth Engine, changes in the lake water area along with changes in temperature, groundwater level, and precipitation were extracted and monitored. Moreover, a comparison took place between these parameters to determine the changes that have taken place in the lake over the past two decades. To predict the parameters, the changing pattern was predicted and analyzed using the Prophet model. The most important advantage of the Prophet model is its ability to convert discrete data to continuous data to make the best predictions. This method automatically detects the trend of seasonal data and displays the trend of seasonal changes.Materials and Methods Satellite images were acquired from the Google Earth Engine platform to monitor the wetland. Landsat 7 and 8 images were used for water body extraction, GRACE Data were used for extraction of groundwater level changes, MODIS product was used for extraction of vegetation and wetland surface temperature, and TRMM image product was used to extract precipitation values. An automated water extraction index was used to extract the wetland body water. The groundwater level was extracted from the GRACE sensor. MODIS sensor product was used to obtain the surface temperature time series for the study area. For the extraction of precipitation time series, the monthly cumulative data of the TRMM (3B43V7) satellite with a spatial resolution of 0.25°C was extracted using Google Earth Engine and the trend of changes was evaluated and analyzed. The Mann-Kendall test is one of the most widely used non-parametric tests for detecting meteorological and environmental data trends, which is used to detect a monotonic trend line since this test is a non-parametric method, it does not need that the data follow a normal distribution. The Prophet predictive model is a predictive library developed by Facebook and is available in R and Python programming languages. This library supports additive modeling methods and can properly predict discrete values continuously. This feature is called "Holiday". Another feature of this library is the automatic detection of daily, weekly, seasonal and annual trends. The mean absolute error (MAE), by default, exists in the Prophet library. This error represents a more natural standard than the mean error and unlike the RMSE error, it is unambiguous.Results and Discussion In the present study, we monitored the Bakhtegan wetland using the Google Earth Engine platform to observe the trend of water level changes in this wetland from 2000 to 2020. In addition, Parameters were also predicted using the Prophet Prediction method which is developed and published by Facebook. By examining this trend, it can be observed that the water level of the wetland has been significantly reduced during two decades. In this regard, the trend of groundwater level, temperature, and precipitation in the area was investigated. Examining these factors, it was found that along with a 58.3% decrease in the water level of the wetland, there was a 260% decrease in the groundwater level of the region, although the amount of rainfall in the region has been less compared to other factors and has been decreased about 29%. Using Mann-Kendall statistical test, the trend of this decline was proved. To predict the parameters, the Prophet model has been able to make predictions for 1500 days as continuous data using discrete data. The output of the model has shown that for rainfall parameters and groundwater level a downward trend is predictable over the next 1500 days which is low intensity for precipitation but with high intensity for groundwater level. Temperature prediction indicated that it has a seasonal trend, and has a high amount of fluctuation within a year, but its annual trend indicates stability in the coming years. The results of the model for the water level of the wetland also show a relatively low upward trend that has a probability of change of ±12.5 Square kilometers. Also, the error of the parameters at the 95% significant level has acceptable accuracy, which indicates the validity of the prediction. An automated water extraction index was used in this study to extract the time series of the water body of the wetland. Using the mean time series extracted, the maximum and minimum wetland’s water body area belongs to 2006 with 629.23 square kilometers and 2014 with 156.82 square kilometers, respectively. The time series of changes in this wetland indicates that the water volume of the wetland has been declining in the last two decades. According to this study, it can be concluded that the trend of changes in the water level of the wetland has been decreasing. The descending changes in the lake based on the trend of changes in groundwater levels indicates a decrease in water volume in the area. Considering that the trend of precipitation changes has been stable, it can have assumed that improper management and excessive use of groundwater may be a reason for lowering the water level of the wetland. Due to the same decrease in the water level of the lake, the temperature has also decreased by about 3°C.Conclusion According to this study, it can be concluded that groundwater levels and precipitation will have a downward trend in the future, which will lead to a decrease in the water level of the wetland, which itself has the potential to fluctuate in the future, and the downward trend continues. With the current trend, the only solution is to plan properly to preserve the wetland. If this trend continues, we will face the destruction of the wetland. Given the monthly trend of the wetland surface, it is suggested not to over-exploit groundwater resources, especially in the summer. For further research, the Google Earth Engine platform can be used without the need to download the images and spend a lot of time and money, to obtain the time series of images. Regarding the prediction, in future studies, the Prophet model can be applied, since it uses discrete data and at the same time provides the desired accuracy.

References:

Alic E, Das M, Kaska O (2019) Heat flux estimation at pool boiling processes with computational intelligence methods. Processes 7(5):293

Bagheri M H, Bagheri A, Sohooli GA (2016) Analysis of changes in the Bakhtegan lake water body under the influence of natural and human factorse. Iran - Water Resources Research 12(3):1–11 (In Persian)

Bagherpour M, Seyedian M, Fathabadi A, Mohamadi A (2017) Study of Mann-Kendall test performance in detecting the series of autocorrelation. Iranian Journal of Watershed Management Science&Engineering 11(36):11–21(In Persian)

Chen L, Jin Z, Michishita R, Cai J, Yue T, Chen B, Xu B (2014) Dynamic monitoring of wetland cover changes using time-series remote sensing imagery. Ecological Informatics 24:17–26

Endter-Wada, J., Kettenring, K. M., & Sutton-Grier, A. (2020). Protecting wetlands for people: Strategic policy action can help wetlands mitigate risks and enhance resilience. Environmental Science & Policy, 108, 37-44.

FacebookResearch (2017) Prophet: forecasting at scale. [Online]. Available at: https://research.fb.com/blog/2017/02/prophet-forecasting-at-scale/

Fan D, Wu H, Dong G, Jiang X, Xue H (2019) A Temporal Disaggregation Approach for TRMM Monthly Precipitation Products Using AMSR2 Soil Moisture Data. Remote Sensing 11(24):2962

Feyisa G L, Meilby H, Fensholt R, Proud SR (2014) Automated Water Extraction Index: A new technique for surface water mapping using Landsat imagery. Remote Sensing of Environment 140:23–35

Gulácsi A, Kovács F (2020) Sentinel-1-Imagery-Based High-Resolution Water Cover Detection on Wetlands, Aided by Google Earth Engine. Remote Sensing 12(10):1614

Halabian, A. h., & Shabankari, M. (2016). Study the Trend of Temporal- Spatial Variation in Mesopotamian Marshlands and Effective Factors. Human & Environment, 14(4), 9-24 (In Persian)

Hu T, Liu J, Zheng G, Zhang D, Huang K (2020) Evaluation of historical and future wetland degradation using remote sensing imagery and land use modeling. Land Degradation & Development 31(1):65–80

Iguchi, T., Kozu, T., Meneghini, R., Awaka, J., & Okamoto, K. I. (2000). Rain-profiling algorithm for the TRMM precipitation radar. Journal of Applied Meteorology and Climatology, 39(12), 2038-2052.

Joodaki G (2014) Earth mass change tracking using GRACE satellite gravity data.. Available at: http://hdl.handle.net/11250/232785

Kaplan G, Avdan U (2018) Monthly analysis of wetlands dynamics using remote sensing data. ISPRS International Journal of Geo-Information 7(10):411

Kendall MG (1948) Rank correlation methods. ISPRS International Journal of Geo-Information

Klemas, V. (2011). Remote sensing of wetlands: case studies comparing practical techniques. Journal of Coastal Research, 27(3), 418-427.

Liu Y, Hu Y, Long S, Liu L, Liu X (2017) Analysis of the effectiveness of urban land-use-change models based on the measurement of spatio-temporal, dynamic urban growth: A cellular automata case study. Sustainability 9(5):796

Mann HB (1945) Nonparametric tests against trend. Econometrica: Journal of the econometric society 245–259

Neteler M (2005) Time series processing of MODIS satellite data for landscape epidemiological applications. International Journal of Geoinformatics 1(1):133–138

Patakamuri S K, Muthiah K, Sridhar V (2020) Long-term homogeneity, trend, and change-point analysis of rainfall in the arid district of ananthapuramu, Andhra Pradesh State, India. Water 12(1):211

Saha, T. K., Pal, S., & Sarkar, R. (2021). Prediction of wetland area and depth using linear regression model and artificial neural network based cellular automata. Ecological Informatics, 62, 101272.

Sahay A, Amudha J (2020) Integration of Prophet Model and Convolution Neural Network on Wikipedia Trend Data. Journal of Computational and Theoretical Nanoscience 17(1):260–266

Salmanpour A, Salehi M H, Mohammadi J, Naderi M (2016) Monitoring Soil salinity around Bakhtegan lake, Fars province, Iran, using Landsat data. Electronic Journal of Soil Management and Sustainable Production 6(1):177–190(In Persian)

Tabouzadeh, S., Zarei, H., & Bazrafshan, O. A. (2016). Analysis of severity, duration, frequency and zoning map of meteorological drought of Bakhtegan river basin. Irrigation Sciences and Engineering, 38(4), 109-123. (In Persian)

Vishwas B, Patel A (2020) Hands-on Time Series Analysis with Python. Apress

Willmott C J, & Matsuura K (2005) Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate research 30(1):79–82

Winter TC (2000) The vulnerability of wetlands to climate change: a hydrologic landscape perspective 1. JAWRA Journal of the American Water Resources Association 36(2):305–311

Zunic E, Korjenic K, Hodzic K, Donko D (2020) Application of Facebook’s Prophet Algorithm for Successful Sales Forecasting Based on Real-world Data. arXiv preprint arXiv:2005.07575