Abstract
Current research on evapotranspiration (ET) is motivated by the growing world population and
its demand for food and hence an intensified irrigation of cultivated lands, along with the need
to better understand climate changes. ET links the land surface processes to the atmospheric
processes and is thus of importance in both hydrological, agricultural and atmospheric sciences.
Still today, accurate measurements of ET are not achieved easily. The state-of the-art method to
measure ET, the eddy covariance method, is associated with uncertainties and its footprint,
though at the order of around 1 hectare, varies much with the atmospheric stability and wind
conditions. Indirect measurements of ET are obtained with satellite imagery, as a residual of the
surface energy balance. Satellite images provide spatially distributed measurements, however
high resolution satellite products provide footprints of about 30 m, which is too coarse for many
applications. Spatially distributed measurements from Unmanned Aerial Vehicles (UAVs –
drones), can close the gap in scale between eddy covariance measurements and satellite
estimates. Lightweight cameras and sensors mounted onboard UAVs enable data resolutions of
about 2 cm and coverage of several hundreds of square meters. Further, UAVs can operate at
any time provided that winds are mild and rain doesn’t occur. Hence, not just the spatial
resolution but also the temporal resolution are highly improved with UAVs, compared to
satellites. This enables new scales of measurements and thus new understandings of ET and its
inferred parameters such as crop water stress and heat fluxes in the surface energy balance.
However, UAV data collection is a new measuring method and the lightweight sensors are
novel instrumentations. Workflows for processing UAV data, and the data quality that
lightweight sensors provide, need to be examined in order for UAVs to become efficient data
collection platforms, advantageous for various scientific fields along with commercial and
applied implementations.
Studies conducted in this thesis investigated UAVs as platforms for sensors that enable
estimates of ET, heat fluxes and crop water stress at a spatial and temporal resolution which is
not feasible using the current state-of-the-art measuring methods: eddy covariance and satellites.
It was thus investigated to what extent UAV systems can provide accurate, high resolution
(especially thermal infrared) data which are suited as input to various ET algorithm and other
purposes requiring good quality remote sensing data. In this thesis the collected UAV data were
applied to both One (OSEB) and Two-Source Energy Balance (TSEB) algorithms along with
the trapezoid, WDI algorithm. The two versions of the TSEB algorithms were developed to be
operational with satellite images and it was tested whether they can be operational with UAV
images as well. UAV images vary from satellite images in that they, among other things, enable
data to be collected in overcast weather situations. Thermal infrared (TIR) data are considered
to be better quality when collected in overcast weather situations, however the ET algorithms
were developed under the assumption of clear sky conditions. UAV based ET estimates were
compared to both eddy covariance measurements and satellites estimates.
ET estimates obtained from the UAV data were generally in good agreement with ET estimates
from the eddy covariance system and satellite products. The Root Mean Square Error (RMSE)
obtained when using the TSEB models were similar to RMSE values obtained in other studies
using the same models but with satellite data as boundary conditions instead of UAV data
(Paper 1). Further, a direct comparison between a UAV TIR mosaic and a TIR scene from the
Landsat 8 satellite showed coherent results of ET from different crop types (Paper 3). The
veracity of the sub-field variations that were revealed from the UAV data (both TIR and Red-
Green-Blue (RGB) data) were validated using patterns from the applied irrigation system along
with point measurements of soil water contents, Leaf Area Index (LAI) and NDVI (Paper 2).
Accurate, spatially distributed estimates of ET, heat fluxes and crop water stress were thus
obtained in spatial resolutions not previously seen. Additionally, Paper 1 established that the
good quality TIR measurements obtained during overcast weather situations can be combined
with advanced algorithms originally developed for satellite images and clear sky conditions.
Further, the fact that a relatively cheap UAV system can provide good quality TIR and RGB
data is of great significance for the practical implementation of UAVs, in various fields, as the
need for costly flight operations are reduced. Moreover, the workflow for UAV TIR imagery
was made more efficient by converting TIR images to unsigned 16 bit data and hence enabling
image processing (such as orthorectification and concatenation) within the Agisoft PhotoScan
software framework.
This thesis contains, to my best knowledge, the first investigation on advanced heat flux
modelling based on UAV images (Paper 1) along with the first attempt to directly compare
UAV TIR data with Landsat 8 satellite TIR data (Paper 3).
its demand for food and hence an intensified irrigation of cultivated lands, along with the need
to better understand climate changes. ET links the land surface processes to the atmospheric
processes and is thus of importance in both hydrological, agricultural and atmospheric sciences.
Still today, accurate measurements of ET are not achieved easily. The state-of the-art method to
measure ET, the eddy covariance method, is associated with uncertainties and its footprint,
though at the order of around 1 hectare, varies much with the atmospheric stability and wind
conditions. Indirect measurements of ET are obtained with satellite imagery, as a residual of the
surface energy balance. Satellite images provide spatially distributed measurements, however
high resolution satellite products provide footprints of about 30 m, which is too coarse for many
applications. Spatially distributed measurements from Unmanned Aerial Vehicles (UAVs –
drones), can close the gap in scale between eddy covariance measurements and satellite
estimates. Lightweight cameras and sensors mounted onboard UAVs enable data resolutions of
about 2 cm and coverage of several hundreds of square meters. Further, UAVs can operate at
any time provided that winds are mild and rain doesn’t occur. Hence, not just the spatial
resolution but also the temporal resolution are highly improved with UAVs, compared to
satellites. This enables new scales of measurements and thus new understandings of ET and its
inferred parameters such as crop water stress and heat fluxes in the surface energy balance.
However, UAV data collection is a new measuring method and the lightweight sensors are
novel instrumentations. Workflows for processing UAV data, and the data quality that
lightweight sensors provide, need to be examined in order for UAVs to become efficient data
collection platforms, advantageous for various scientific fields along with commercial and
applied implementations.
Studies conducted in this thesis investigated UAVs as platforms for sensors that enable
estimates of ET, heat fluxes and crop water stress at a spatial and temporal resolution which is
not feasible using the current state-of-the-art measuring methods: eddy covariance and satellites.
It was thus investigated to what extent UAV systems can provide accurate, high resolution
(especially thermal infrared) data which are suited as input to various ET algorithm and other
purposes requiring good quality remote sensing data. In this thesis the collected UAV data were
applied to both One (OSEB) and Two-Source Energy Balance (TSEB) algorithms along with
the trapezoid, WDI algorithm. The two versions of the TSEB algorithms were developed to be
operational with satellite images and it was tested whether they can be operational with UAV
images as well. UAV images vary from satellite images in that they, among other things, enable
data to be collected in overcast weather situations. Thermal infrared (TIR) data are considered
to be better quality when collected in overcast weather situations, however the ET algorithms
were developed under the assumption of clear sky conditions. UAV based ET estimates were
compared to both eddy covariance measurements and satellites estimates.
ET estimates obtained from the UAV data were generally in good agreement with ET estimates
from the eddy covariance system and satellite products. The Root Mean Square Error (RMSE)
obtained when using the TSEB models were similar to RMSE values obtained in other studies
using the same models but with satellite data as boundary conditions instead of UAV data
(Paper 1). Further, a direct comparison between a UAV TIR mosaic and a TIR scene from the
Landsat 8 satellite showed coherent results of ET from different crop types (Paper 3). The
veracity of the sub-field variations that were revealed from the UAV data (both TIR and Red-
Green-Blue (RGB) data) were validated using patterns from the applied irrigation system along
with point measurements of soil water contents, Leaf Area Index (LAI) and NDVI (Paper 2).
Accurate, spatially distributed estimates of ET, heat fluxes and crop water stress were thus
obtained in spatial resolutions not previously seen. Additionally, Paper 1 established that the
good quality TIR measurements obtained during overcast weather situations can be combined
with advanced algorithms originally developed for satellite images and clear sky conditions.
Further, the fact that a relatively cheap UAV system can provide good quality TIR and RGB
data is of great significance for the practical implementation of UAVs, in various fields, as the
need for costly flight operations are reduced. Moreover, the workflow for UAV TIR imagery
was made more efficient by converting TIR images to unsigned 16 bit data and hence enabling
image processing (such as orthorectification and concatenation) within the Agisoft PhotoScan
software framework.
This thesis contains, to my best knowledge, the first investigation on advanced heat flux
modelling based on UAV images (Paper 1) along with the first attempt to directly compare
UAV TIR data with Landsat 8 satellite TIR data (Paper 3).
| Originalsprog | Engelsk |
|---|
| Forlag | Department of Geosciences and Natural Resource Management, Faculty of Science, University of Copenhagen |
|---|---|
| Antal sider | 104 |
| Status | Udgivet - 2016 |
FN’s Verdensmål
Dette resultat bidrager til følgende verdensmål
