Regional measurements and spatial/temporal analysis of CDOM in 10,000+ optically variable Minnesota lakes using Landsat 8 imagery

Highlights

• Remote sensing methods developed to map CDOM in 10,000 Minnesota lakes.

• Atmospheric correction and new models for Landsat 8 OLI imagery improved results.

• Ecoregions rich in wetlands and forest have higher CDOM.

• CDOM increased with increased precipitation in forest/wetland-rich ecoregions.

• CDOM decreased with increased precipitation in agricultural ecoregions.

Abstract

Information on colored dissolved organic matter (CDOM) is essential for understanding and managing lakes but is often not available, especially in lake-rich regions where concentrations are often highly variable in time and space. We developed remote sensing methods that can use both Landsat and Sentinel satellite imagery to provide census-level CDOM measurements across the state of Minnesota, USA, a lake-rich landscape with highly varied lake, watershed, and climatic conditions. We evaluated the error of satellite derived CDOM resulting from two atmospheric correction methods with in situ data, and found that both provided substantial improvements over previous methods. We applied CDOM models to 2015 and 2016 Landsat 8 OLI imagery to create 2015 and 2016 Minnesota statewide CDOM maps (reported as absorption coefficients at 440 nm, a₄₄₀) and used those maps to conduct a geospatial analysis at the ecoregion level. Large differences in a₄₄₀ among ecoregions were related to predominant land cover/use; lakes in ecoregions with large areas of wetland and forest had significantly higher CDOM levels than lakes in agricultural ecoregions. We compared regional lake CDOM levels between two years with strongly contrasting precipitation (close-to-normal precipitation year in 2015 and much wetter conditions with large storm events in 2016). CDOM levels of lakes in agricultural ecoregions tended to decrease between 2015 and 2016, probably because of dilution by rainfall, and 7% of lakes in these areas decreased in a₄₄₀ by ≥3 m^-1. In two ecoregions with high forest and wetlands cover, a₄₄₀ increased by >3 m^-1 in 28 and 31% of the lakes, probably due to enhanced transport of CDOM from forested wetlands. With appropriate model tuning and validation, the approach we describe could be extended to other regions, providing a method for frequent and comprehensive measurements of CDOM, a dynamic and important variable in surface waters.

Introduction

Research in recent decades has revealed a central role for colored (or chromophoric) dissolved organic matter (CDOM) in regulating major physical, chemical and biological processes in lakes and rivers (e.g., reviewed in Solomon et al., 2015, Williamson et al., 1999, Creed et al., 2018, and elsewhere). We now know that CDOM functions as one of a small number of “master variables, “ similar to phosphorus, pH and redox potential, that control important aspects of the composition and functioning of aquatic ecosystems and regulate their responses to environmental change (Williamson et al., 1999; Creed et al., 2018). Recent studies show that CDOM levels strongly influence: (a) light and thermal regimes in lakes (e.g., Houser, 2006; Ask et al., 2009; Thrane et al., 2014; Pilla et al., 2018; Snucins and Gunn, 2000), (b) biogeochemical cycles (e.g., Knoll et al., 2018; Corman et al., 2018), (c) food web processes and interactions (e.g., Karlsson et al., 2009; Solomon et al., 2015), (d) contaminant bioavailability (e.g., Tsui and Finlay, 2011), and (e) water clarity (e.g., Brezonik et al., 2019a). Knowledge of the sources, levels, and cycling of CDOM in freshwaters thus is important for aquatic resource management and for predicting the outcomes of environmental change.

Moderate to high levels of CDOM in freshwaters are determined largely by rates of transport from soils and wetlands in surrounding watersheds and thus are affected by a combination of factors related to vegetation and hydrology. The dependency of aquatic CDOM on dynamic external sources, combined with internal production and loss processes in aquatic systems, can lead to high variability of CDOM levels across landscapes and within lakes at time scales of seasons to years (Brezonik et al., 2015, Williamson et al., 1999). Human-driven changes in temperature, atmospheric chemistry, land use and watershed hydrology also can have strong effects on CDOM (Creed et al., 2018, Finstad et al., 2016, Kritzberg, 2017, Stanley et al., 2012, de Wit et al., 2016).

Although CDOM is easily measured in the laboratory, the availability of in situ CDOM data is surprisingly limited relative to its importance, even in states like Minnesota, where monitoring of its >10,000 surface waters is a major focus of many state, tribal and local agencies. Several recent, large-scale assessments of regional U.S. lake monitoring efforts (Stanley et al., 2012; Ross et al., 2019) showed that far fewer data were available for CDOM and related variables such as DOC compared to nutrients, chlorophyll, and water clarity, despite the strong effects of CDOM on those and other physicochemical variables. The spatial and temporal variation in CDOM in surface waters suggests the need for more CDOM data to improve understanding of drivers and better predict lake responses to stresses ranging from local land cover changes to global climate change. Some countries with large numbers of CDOM-rich lakes have incorporated routine monitoring of CDOM or a related parameter such as DOC (e.g., Sobek et al., 2007). The relative lack of CDOM data for U.S. lakes (Stanley et al., 2019) may stem from the fact that many monitoring programs initially started in relatively low-CDOM regions but also from the fact that the importance of CDOM as a driver of ecological conditions has been appreciated only recently.

Whatever the cause, the availability of CDOM data remains deficient compared to its importance. Remote sensing using satellite-based sensors could play an important role in providing CDOM data at high temporal and spatial resolution. Recent studies show that the Landsat sensors (Kutser et al., 2005; Brezonik et al., 2005; Kutser et al., 2009; Olmanson et al., 2016a), and Sentinel-2/MSI sensors (Toming et al., 2016; Chen et al., 2017) can provide such data at scales relevant for inland lakes as small as 4 hectares (ha).

Recent improvements in Earth-observing satellite sensors have expanded the capabilities to measure optically-related water quality characteristics, including CDOM, in lakes (Olmanson et al., 2016a; Tyler et al., 2016; Pahlevan et al., 2019; Page et al., 2019). Specifically, the Landsat 8 Operational Land Imager (L8/OLI) and the European Space Agency (ESA) Sentinel-2 MultiSpectral Imager (S2/MSI) have improved spatial, spectral, radiometric and temporal resolution compared with earlier sensors. With the L8/OLI and S2/MSI constellation collecting imagery every 3 to 5 days, frequent satellite-based measurements of a variety of key water quality variables on lakes are now possible.

The use of satellite imagery to measure CDOM at large regional scales and over multiple time periods requires analysis of multiple images. Unless ground-based data are available to calibrate each image (a requirement difficult to achieve), accurate methods are needed for atmospheric correction of images to produce surface reflectance data directly representative of optical signals from waterbodies. Although various approaches have been reported to accomplish this (e.g., Pahlevan et al., 2017a, Pahlevan et al., 2017b; Vanhellemont and Ruddick, 2015, Vanhellemont and Ruddick, 2016), we have found that many of them yield unreliable results for inland lakes (Olmanson et al., 2011; Page et al., 2019). The recent availability of surface reflectance products from the EROS Center appears to have overcome this obstacle for Landsat 8 imagery (Kuhn et al., 2019), and Page et al. (2019) described a workflow process to atmospherically correct and harmonize S2/MSI and L8/OLI satellite imagery in Google Earth Engine (GEE) (Gorelick et al., 2017).

This paper describes application of these advances to measure CDOM on all waterbodies larger than 4 ha across a large geographic region (the state of Minnesota) that encompasses >226,000 km² and contains officially 11,842 lakes 4 ha or larger in area (https://www.dnr.state.mn.us/faq/mnfacts/water.html). The paper describes a robust semi-empirical approach for routine monitoring of CDOM using L8/OLI imagery. We demonstrate the consistency and reliability of two atmospheric correction methods to generate remote sensing reflectance (R_rs) products and use these products to assemble a CDOM database on >10,500 lakes for both 2015 and 2016. We assess the accuracy of retrieved CDOM data for both low- and high-CDOM waters and summarize distributions of CDOM in Minnesota lakes at the ecoregion level.