Water quality estimation by optical remote sensing in boreal lakes

Abstract

Monitoring of lakes is mainly based on collection of water samples, which are transported to a laboratory for analyses. In lake-rich regions gathering of water quality information is challenging, because only a small proportion of the lakes can be assessed each year, often only a few times a year. One of the techniques for improving the temporal and spatial representativeness of lake monitoring is remote sensing.The main objectives of this study were to investigate and test remote sensing interpretation algorithms for water quality estimation in Finnish lakes, to develop optical models for the needs of interpretation and for the estimation of light attenuation, and to study the advantages of the use of remote sensing data as compared to the conventional monitoring methods. The experimental material included detailed optical measurements in 11 lakes, remote sensing measurements with concurrent in situ sampling, automatic raft measurements and a national dataset of routine water quality measurements. Remote sensing data consisted of airborne and satellite measurements (ETM+, ALI and MERIS).The analyses of the spatially high-resolution airborne remote sensing data of eutrophic and mesotrophic lakes showed that one or a few discrete water quality observations of conventional monitoring can yield a clear over- or underestimation of the overall water quality. The use of TM-type satellite instruments in addition to routine monitoring results substantially increases the number of lakes for which water quality information is obtained. The results indicated preliminarily that coloured dissolved organic matter (CDOM) can be estimated with TM-type satellite instruments, which could be possible utilised as an aid in the estimation of the role of lakes in global carbon budgets. Based on the results of reflectance modelling and experimental data, MERIS satellite instrument has optimal or near-optimal channels for the estimation of turbidity, chlorophyll a and CDOM in Finnish lakes. MERIS images with 300 m spatial resolution can be utilised in production of water quality information in different parts of large and medium-size lakes, and in filling the gaps of conventional monitoring. Regional algorithms that would not require simultaneous in situ data for algorithm training would increase the amount of remote‑sensing-based information available for lake monitoring. The MERIS Boreal Lakes processor, trained with the optical data and concentration ranges provided by this study, enabled turbidity estimation with good accuracy without need for algorithm correction with in situ measurements, while chlorophyll a and CDOM estimation requires further development of the processor. The accuracy of interpretation of chlorophyll a via semi‑empirical algorithms can be improved by classifying lakes prior to interpretation by CDOM level and trophic status, and by creating lake-type-specific algorithms. The results of optical modelling showed that spectral diffuse attenuation coefficient can be estimated with reasonable accuracy from the measured water quality concentrations. This provides more detailed information on light attenuation from routine monitoring measurements than is available through the Secchi disk transparency.This study improves the interpretation of water quality by remote sensing and encourages the use of remote sensing in lake monitoring.