Author(s): Laura Wermuth; Luiz Felipe Machado Faria De Sousa; Lino Augusto Sander De Carvalho; Nilo Nascimento; Talita Silva
Linked Author(s): Nilo Nascimento, Talita Silva
Keywords: Google Earth Engine; Pampulha Reservoir; Brazil; Statistical Mono-Window algorithm; Split-Window algorithm
Abstract: Reservoirs play an important role, especially in urban tropical regions, where they often serve as water supply, energy production, flood prevention and recreational areas. The Pampulha Reservoir (surface area 1.97 km²; maximum depth 16.17 m), located in Belo Horizonte, Brazil, is an artificial reservoir. Since the 1970’s, the frequency of algal blooms, particularly cyanobacteria, has increase, mainly due to eutrophication caused by the contamination by the input of nutrients and heavy metals from the ongoing urbanization of its tributaries. Water temperature is an important driver for water dynamics, physical and biochemical processes in inland water bodies and its monitoring is desirable for a better understanding of water quality management processes. Remote sensing techniques can be used to monitor surface water temperature and provide the opportunity of monitoring surface water temperature (SWT) and complement in situ data. This study aims to assess two methods to retrieve surface water temperature from the Pampulha Reservoir, the Statistical Mono-Window (SMW) algorithm and the Split-Window (SW) algorithm for SWT-retrieval from Landsat 8 (LS8) images using Google Earth Engine (GEE) in Pampulha Reservoir. A provided code repository of the SMW algorithm within GEE was used directly, whereas the SW algorithm was implemented individually. Both algorithms retrieve water temperature values for dates whenever an applied cloud mask admits it and CSV-files holding this data are downloaded directly from GEE. In situ data of Pampulha Reservoir are available from January 2015 until November 2017 and are compared to LS8 retrieved SWT values. Results from both SMW (RMSE = 2.35 °C, R² = 0.20) and SW (RMSE = 1.48 °C, R² = 0.63) algorithms are similar to results found in other studies, however, the later method shows more satisfactory results. Using the SW algorithm, a time series of SWT in Pampulha Reservoir of available LS8 information from 2013 up to 2020 and thermal maps showing the mean monthly SWT are generated to characterize temporal and spatial thermal behaviour of the reservoir. The time series shows a seasonal pattern with higher water temperature values during the wet and warm season (November to March) and lower water temperature values during the dry and cold season (June to August). Lower water temperatures can generally be found in more central areas. Derived SWT values of Landsat 8 using the Split-Window algorithm are representative of in situ SWT values and can be used for monitoring urban tropical reservoirs, in particular, eminently polluted ones. Especially in countries with a lack of investment in in situ monitoring such as Brazil, Landsat 8 derived SWT depicts a great source of complementary information regarding water body and resource management.