Non-optical Water Quality Retrieval from Zhuhai-1 OHS Hyperspectral
Images in Taipu River
Abstract
Hyperspectral remote sensing is thought to be a useful technology for
assessing the condition of inland waters. However, non-optically active
water quality parameters are rarely explored in hyperspectral remote
sensing applications, despite they are highly valued in the aquatic
environment condition. This study intends to evaluate the performance of
non-optically active water quality parameters using Zhuhai-1
hyperspectral imagery. Focusing on total nitrogen (TN), total phosphorus
(TP), ammonia nitrogen (NH3-N) and nitrate-nitrogen (NO3-N) in Taipu
River, we constructed empirical models to evaluate the precision of
water quality inversion from OHS by comparing with Sentinel-2, and
determined the sensitive bands of different water quality parameters.
The final results showed that the polynomial model based on OHS had the
greatest potential in retrieving TN, TP and NH3-N concentration, and the
R2 was 0.9678, 0.7924, 0.7682 respectively. The combination of
R(510)/R(820) and R(700)/R(806), R(940)/R(820) and R(806)/R(926),
R(709)/R(806) and R(746)/R(620) were most sensitive to TN, TP and NH3-N
respectively. The OHS and Sentinel-2 both had potential in retrieving
NO3-N. The R2 was 0.9791 from OHS and was 0.9513 from Sentinel-2. The
sensitive bands of NO3-N were R(596)/R(665) and R(466)/R(580) from OHS,
and Red Eage3/Blue and SWIR1/Blue from Sentinel-2. We also analyzed the
drivers of the spatial distribution of water quality in Taipu River, the
results showed negative impacts of farmland and urban land on water
quality, and beneficial impacts of forest land on water quality. This
study represented a promising step in hyperspectral remote sensing for
retrieving inland non-optically active water quality parameters
utilizing Zhuhai-1.