Multispectral Optical Remote Sensing for Water-Leak Detection

Water losses from water distribution means have a high environmental impact in terms of natural resource depletion (water, energy, ecosystems). This work aims to develop an optical airborne surveillance service for the detection of water leaks (WADI-Water-tightness Airborne Detection Implementation) to provide water utilities with adequate and timely information on leaks in water transportation mains outside urban areas. Firstly, a series of measurement campaigns were performed with two hyperspectral cameras and a thermal infrared camera in order to select the most appropriate wavelengths and combinations thereof for best revealing high moisture areas, which are taken as a proxy for water leakage. The Temperature-Vegetation-Index method (T-VI, also known as Triangle/Trapezoid method) was found to provide the highest contrast-to-noise ratio. This preliminary work helped select the most appropriate onboard instrumentation for two types of aerial platforms, manned (MAV) and unmanned (UAV). Afterwards, a series of measurement campaigns were performed from 2017 to 2019 in an operational environment over two water distribution networks in France and Portugal. Artificial leaks were introduced and both remote sensing platforms successfully detected them when excluding the unfavorable situations of a recent rain event or high vegetation presence. With the most recent equipment configuration, known and unknown real leaks in the overflown part of a water transportation network in Portugal have been detected. A significant number of false alarms were also observed which were due either to natural water flows (groundwater exfiltration, irrigation runoff and ponds) or to vegetation-cover variability nearby water-distribution nodes. Close interaction with the water utilities, and ancillary information like topographic factors (e.g., slope orientation), are expected to reduce the false alarm rates and improve WADI's methodology performance.

Estimating Evapotranspiration From Satellite Using Easily Obtainable Variables: A case study over the Southern Great Plains, U.S.A

Evapotranspiration (ET) is a critical component of the Earth's water budget, a critical modulator of land-atmosphere (L-A) interactions, and also plays a crucial role in managing the Earth's energy balance. In this study, the feasibility of generating spatially-continuous daily evaporative fraction (EF) and ET from minimal remotely-sensed and meteorological inputs in a trapezoidal framework is demonstrated. A total of four variables, Normalized Difference Vegetation Index (NDVI), Land surface temperature (T_s ), gridded daily average temperature (T_a ) and elevation (z) are required to estimate EF. Then, ET can be estimated with the available soil heat flux (G) and net radiation (Rn) data. Firstly, the crucial model variable, T_s - T_a , is examined how well it characterizes the variation in EF using in situ data recorded at two eddy correlation flux towers in Southern Great Plains, U.S.A in 2011. Next, accuracy of satellite-based T_s are compared to ground-based T_s . Finally, EF and ET estimates are validated. The results reveal that the model performed satisfactorily in modeling EF and ET variation at winter wheat and deciduous forest during the high evaporative months. Even though the model works best with the observed MODIS-T_s as opposed to temporally interpolated T_s , results obtained from interpolated T_s are able to close the gaps with reasonable accuracy. Due to the fact that T_s - T_a , is not a good indicator of EF outside the growing season when deciduous forest is dormant, potential improvements to the model are proposed to improve accuracy in EF and ET estimates at the expense of adding more variables.