Hyperspectral Sensing from Unmanned Aerial Systems for Water Quality and Quantity in Terrestrial and Aquatic Systems

Christian Josef Köppl

Research output: Book/ReportPh.D. thesis

143 Downloads (Pure)

Abstract

Water is essential for all life on planet earth. Through anthropogenic impacts and climate change, water resources are under an increasing stress. Globally,water quality is decreasing, while the frequency of floods and droughts is increasing. In order to mitigate that stress and to manage water resources sustainably, they need to be monitored to be part of decision making and policies. Traditionally, water resources are monitored by in-situ measurements,but in-situ monitoring is cost and labour intensive and only gives information in a few discrete points in space/time. Satellite earth observation can monitor water resources cost-efficiently and spatially distributed, but spectral sensors cannot penetrate cloud cover and their spatial resolution is too coarse to monitor systems like small agricultural plots or small headwater streams.

Hyperspectral (HS) remote sensing from unmanned aerial systems (UAS) can collect data under clouds and provide high-resolution spectral data of the land surface at a high spatial resolution. HS remote sensing from UAS is an emerging technology and current challenges include: no standardized operation procedures to map land surface reflectance; miniaturized HS imagers have a low signal to noise ratio and are in most cases provided by manufactures without thorough radiometric and spectral calibration; standard incoming light sensors (ILS) data must be corrected for sensor tilting effects.

The aim of this thesis was (1) to advance the operability and accuracy of HSremote sensing from UAS correcting for intermittent clouds and sensor tilt and vibrations and (2) to demonstrate applications for quantitative mapping ofwater in terrestrial and aquatic ecosystems.

For the first objective, a novel data-driven method, based on the Lambertian cosine law and the diffuse fraction of irradiance, to correct downwelling irradiance under all-weather conditions was developed, which decreased the error in downwelling irradiance from 15.9 % to 2.8 %. A spectral and radiometric calibration of a Cubert UHD 185 HS imager was carried out, achieving a radiometric accuracy for radiance measurements of < 6 % and revealing the rapid increase in spectral bandwidth with increasing wavelength. The calibration and the downwelling irradiance correction allowed mapping land surface reflectance based on the direct reflection calculation method in an operational manner with high accuracy.

Two types of applications were demonstrated for the second objective: quantifying water flows in the soil-plant-atmosphere continuum on an upland rice field and estimating parameters relevant for ecosystem health in optically complex, small streams (contaminant mixing and physical-chemical water quality parameters).

Amending soil in water-limited regions with biochar is expected to increase the soil water availability and crop resilience to drought events. The effect of biochar amendment on rice was evaluated in a field experiment in Costa Rica with a combination of thermal and HS remote sensing based models. It was found that bamboo based biochar (BC1) increased available soil moisture in the root zone by 17.7 %; while sugarcane based biochar (BC2) increased soil moisture by 10.8 %. As the gross primary production increased more than the evapotranspiration (BC1: 41.9 %; BC2: 17.5 %), the crop water use efficiency also increased (BC1: 40.8 %; BC2: 13.4 %). This demonstrated the positive effect of biochar amendment to soils for rice crop in water-limited regions and showed the potential of monitoring water quantity parameters such as root zone soil moisture, evapotranspiration, soil matric potential, and plant water use efficiency in terrestrial ecosystems based on UAS borne HS remote sensing.

Small streams are important ecosystems sustaining high biodiversity, but anthropogenic stressors lead to a worldwide decline of ecological status in streams. The application of UAS for evaluating chemical contamination linked to groundwater sources, as well as additional parameters related to physical - chemical water quality were investigated and are discussed in terms of their relation to ecosystem health.

Many streams are affected by the discharge of pollutants and an important factor for the impact on the stream ecosystem is how fast the pollutant is diluted by mixing. Therefore, characterizing the mixing in the stream is important for management purposes. Tracer tests are a well-known tool in hydrology and monitoring tracer transport by optical remote sensing from UAS is a way to characterize the mixing with unprecedented spatial detail. A tracer test with the fluorescent tracer Rhodamine WT was conducted in a small stream, impacted by contamination stemming from a contaminated site and the tracer test was monitored by a photo camera as well as a HS camera from UAS. While the photo camera could only predict tracer concentrations in parts of the stream, exposed to direct sunlight (all calibration points R2: 0.17), the tracer concentration could nevertheless be estimated based on HS data under all conditions (R2: 0.78). The HS data allowed visualizing and quantifying the spatial distribution of the tracer plume in high spatial detail. This was the first time a tracer experiment was successfully monitored by optical remote sensing in a small optically complex stream and the accuracy of tracer concentration estimation was comparable to experiments conducted in optically simpler systems.

To implement measures to improve ecological status, the drivers and stressors of the current ecological status need to be understood. One important pressure is water quality. To monitor water quality in an 11 km long, small and optically complex Danish peri-urban stream, the stream was mapped by HS remote sensing from UAS seven times during the course of a year, while taking concurrent in-situ water quality samples at eleven established sampling stations. A statistical partial least squares regression approach was used to predict water quality parameters from the remotely sensed HS data. The parameters chlorophyll-a, total suspended solids and turbidity were predicted with R2 values of 0.85, 0.58 and 0.93, respectively. The water quality mapping performed well for a stream section impacted by an algal bloom in the lake, from which the stream originates. For conditions with lower variability in water quality parameters, the effects of optical complexity of the stream had a detrimental effect on the accuracy of the mapped concentrations.

These findings highlight the great potential of monitoring water quality and quantity in small-scale terrestrial and aquatic systems using HS remote sensing by UAS. This method can deliver information with unprecedented detail to decision makers about systems, which were previously inaccessible for monitoring by remote sensing. Mapping soil moisture, evaporation and primary productivity can be used in agriculture as presented here, to investigate the effects of new farming practices, but it can also be used in an operational precision-agriculture approach, where for example water and nutrients are applied to the field efficiently when and where it is needed, by monitoring the fields with UAS. High resolution mapping of water quality in streams could be used as a monitoring tool to detect extreme events and to respond accordingly, or to identify drivers of poor ecological status, so that measures for improving ecological status could be implemented more efficiently.
Original languageEnglish
Place of PublicationKgs. Lyngby
PublisherDTU Environment
Number of pages176
Publication statusPublished - 2022

Fingerprint

Dive into the research topics of 'Hyperspectral Sensing from Unmanned Aerial Systems for Water Quality and Quantity in Terrestrial and Aquatic Systems'. Together they form a unique fingerprint.

Cite this