Remote Sensing of Inland Lake Harmful Algal Blooms Linda Novitski Cooperative Institute for Limnology and Ecosystems Research Ann Arbor, Michigan May 2, 2014
Talk Outline Introduction to harmful algal blooms Introduction to remote sensing Inland lake harmful algal bloom detection and successes in the Great Lakes
Harmful algal blooms (HABs) What is an algal bloom and why are they so troublesome? Often photosynthetic bacteria, called cyanobacteria Often blue-green in color Produce chlorophyll-a (photosynthetic pigment) Can produce toxins Can reduce oxygen Smelly, unsightly Lois Wolfson
Harmful Algal Blooms (HABs) Ford Lake Silver Lake
ill conduct a gap analysis of existing indicators for marine and rmful algal blooms throughout the country, including an analysis d new State of the Lakes Ecosystem Conference (SOLEC) and mmission indicators for HNABs. A preferred suite of indicators sted using the GLRI remediation efforts for toxic cyanobacteria ake Ontario. Researchers will also assess community these indicators by surveying members of the local bay ve Our Sodus, Inc. The end product of this effort will be a suite hat can be used throughout the Great Lakes basin to track g the remediation efforts targeting harmful algal blooms. In also help us design better monitoring programs for HNABs no they may occur throughout the Great Lakes basin. Harmful (nuisance) Algal Blooms
Harmful algal blooms (HABs) Common genera: Microcystis (P) Anabaena (P) Aphanizomenon (P) Gloeotrichia (P)
Harmful algal blooms (HABs) Other cyanobacterial genera: Cylindrospermopsis (P) Oscillatoria (P) Planktothrix (P/B) Calothrix (B) Lyngbya (P/B) Nodularia (P) Nostoc (P/B) Scytonema (P/B) P=planktonic B=benthic
Nuisance algal blooms Example genera: Cladophora (B) Chlorococcus (P) Ceratium (P) Rhodomonas (P) Prymnesium (P)
Harmful Algal Blooms (HABs) What kinds of toxins do HABs produce? Hepatotoxins (microcystin, nodularian, cylindrospermopsin) Neurotoxins (anatoxin and saxitoxin) Various skin and gastrointestinal toxins Mediators of microbial interactions HAB genera can often produce more than one toxin
Harmful Algal Blooms (HABs) Why are they so successful? Toxic Hard to eat and not nutritious Buoyant Superior competitors for nutrients
Harmful algal blooms (HABs) Where are they located? All inland lakes are potentially susceptible llll
Harmful algal blooms (HABs) How to avoid them? Check for beach advisories Avoid water that looks green, has an odor Avoid swallowing lake water or disturbing the water s surface, or touching the water Ensure that children and animals do not drink water How to prevent or control them? Identify lakes with bloom problems Reduce nutrient pollution Riparian buffers (lower nutrients & provide shade) Mixing of water column Algicides
HABs and climate change HABs are likely to become more frequent and last longer with climate change due to: Increased precipitation events Higher water temperatures
Remote sensing of HABs What is remote sensing? Use of satellites to see Earth s surface Often images are free! Why use it to detect algal blooms? Has the potential to be more time efficient and cost effective than one-time, in-lake sampling. Satellite imagery can help locate, forecast algal blooms
Blue band Green band Red band Infrared band Thermal band Reflectance values
Step 1: Model building Reflectance values + in-lake measurements chlorophyll model Step 2: Chlorophyll estimation New reflectance values + chlorophyll model new chlorophyll estimates
Chlorophyll-a: photosynthetic pigment Secchi depth: measure of water clarity disk lowered until no longer visible
Remote Sensing of HABs Historically: Only visible bands and sometimes infrared bands used in models Simple linear regression models Only useful for one lake, one date, one state, etc. Not useful for lakes with high dissolved organic carbon
Remote sensing of HABs Landsat 7 ETM+ Reflectance data Landsat 7 ETM+ In-lake measurements 2007 Environmental Protection Agency National Lakes Assessment data chlorophyll-a or Secchi depth Chlorophyll models: Linear regression: national vs. regional models Linear regression vs. boosted regression tree
Remote sensing of HABs Landsat 7 ETM+ Available for free! Has been operational since early 1999 Small pixel size Can be obtained from: http://earthexplorer.usgs.gov/
Remote sensing of HABs Landsat 7 ETM+
Landsat band 1 (blue band) Landsat band 2 (green band) Landsat band 3 (red band) Landsat bands in visible light spectrum
2007 US EPA National Lakes Assessment Sample Sites
Remote Sensing Model Results Model Water variable Bands used Linear regression Chlorophyll Blue, red 0.19 R 2 Linear regression Secchi depth Blue, red 0.49
Lake Type A B C D E F G Description Eastern highlands: Low to moderate elevation, warm reservoir Eastern highlands: Low to moderate elevation northern lakes and reservoirs Northern central plains and lowlands: Low elevation lakes and reservoirs with small surface area Southern and coastal plains and lowlands: Low elevation, shallow, warm lakes and reservoirs with large surface area Northern central plains and lowlands: Low to moderate elevation, shallow lakes and reservoirs North-western mountains: Moderate to high elevation, cold, deep lakes and reservoirs South-western mountains: moderate to high elevation, cold lakes and reservoirs
Chlorophyll: National vs. lake type models Model Bands used R2 National Blue, red 0.19 A Green, red 0.16 B Blue, green 0.33 C Blue, green 0.12 D Blue, green 0.10 E Blue, infrared 0.25 F Thermal 0.15 G Blue, red 0.18
Secchi depth: National vs. lake type models Model Bands used R2 National Blue, red 0.49 A Blue, red 0.36 B Blue, green 0.41 C Blue, red 0.47 D Blue, red 0.34 E Blue, red 0.35 F Blue, green 0.63 G Blue, red 0.60
Remote sensing of HABs Landsat 7 ETM+ Chlorophyll model: Boosted regression tree Picks the bands for you Better predictive accuracy
Remote Sensing Model Results Model Water variable Bands used Linear regression Chlorophyll Blue, red 0.19 R2 Boosted Regression Tree Chlorophyll All 0.44 Linear regression Secchi deph Blue, red 0.49 Boosted Regression Tree Secchi depth All 0.52
Model to predict chlorophyll a RS Predicted Chlorophyll ln(µg/l) 0 1 2 3 4 5 6-2 0 2 4 6 Measured Chlorophyll ln(µg/l)
Chlorophyll models from the literature Year Authors Satellite Geographic extent Bands used R 2 (in prep.) Novitski et al. ETM+ USA All 0.44 2008 Kabarra et al. ETM+ Coast of Triploi Blue, green, red 2005 Brezonik et al. TM Fifteen Minnesota lakes Blue, green, red, infrared 2001 Giardino et al. TM Lake Iseo, Italy Blue, green 0.99 2001 Brivio et al. TM Lake Garda, Italy Blue, green, red 1995 Mayo et al. TM Lake Kinneret, Israel Blue, green, red 1995 Gitelson et al. TM* Haifa Bay, Israel Blue, green 0.37 1994 Pattiaratchi et al. TM Cockburn Sound, Australia Blue, green, red 1991 Khorram et al. TM August Bay, Italy Blue, green 0.84 0.72 0.88 0.68-0.82 0.49 0.73-0.77 1983 Carpenter & Carpenter MSS Lake Burley Griffin, Australia Green, red, infrared 0.50-0.85
Model to predict Secchi depth RS Predicted Secchi Depth (m) 0 2 4 6 8 10 0 2 4 6 8 10 12 14 Measured Secchi Depth (m)
Secchi depth models from the literature Year Authors Satellite Geographic extent Bands used R 2 (in prep.) Novitski et al. ETM+ USA All 0.52 2008 Olmanson et al. MSS/TM/ET M+ State of Minnesota Blue, red 0.54 2008 Kabbara et al. ETM+ Coast of Triploi Blue, green 0.54 2004 Chipman et al. TM/ETM+ State of Wisconsin Blue, red 0.42-0.88 2003 Nelson et al. ETM+ State of Michigan Blue, red 0.43-0.82 2002b Kloiber et al. TM Five hundred Minnesota Lakes 2002 b Kloiber et al. MSS Five hundred Minnesota Lakes Blue, red 0.72-0.93 Blue, green 0.60-0.79 2001 Giardino et al. TM Lake Iseo, Italy Blue, green 0.85 1994 Pattiaratchi et al. TM Cockburn Sound, Australia Blue, green, red 0.72-0.78 1991 Khorram et al. TM August Bay, Italy Blue, green 0.83 1983 Lillesand et al. MSS Twenty-eight Green, red, 0.88-0.94
Remote sensing of HABs Landsat 7 ETM+ Chlorophyll is more challenging to detect than Secchi depth More data on chlorophyll vs. other color producing agents More measurements per lake. In-lake measurement & satellites passover date Incorporate other data
Successes in the Great Lakes Landsat 1-7 and MODIS satellites Great Lakes National Program Office data Boosted regression tree models
MODIS Operational 2000- present Daily passover 250 m 1 km pixel resolution Can be obtained from: http://modis.gsfc.nasa.gov/data/dataprod/index.php http://oceanmotion.org/guides/fd_3/fd_student_3.htm
MODIS bands in visible light spectrum
2007-2009 Great Lakes Sample Sites
Cross-validated R 2 =0.69, RMSE=0.55 µg/l Great Lakes Measured Chl versus Landsat ETM+ and TM Predicted Chl Landsat ETM+ and TM Predicted Chl (µg/l) 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Measured Chl (µg/l) Measured chlorophyll versus Landsat ETM+ & TM predicted chlorophyll for 2007-2009 Great Lakes sample sites
Great Lakes research
Great Lakes research
Great Lakes research
Cross-validated R 2 =0.85, RMSE=0.10 µg/l llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll Great Lakes LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL Measured Chl versus MODIS Predicted Chl MODIS Predicted Chl (µg/l) 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Measured Chl (µg/l) Measured chlorophyll versus MODIS predicted chlorophyll for 2007-2009 Great Lakes sample sites
Saginaw Bay August Landsat Image Time Series MSS 1975 MSS 1979 TM 1985 LLLLLLLLLLLLLLLLLLLLLLLLLLL MSS August 15, 1979 Chlorophyll-a (µg/l) Land Clouds TM 1990 TM 1996 ETM+ 2000 0-1.00 1.01-2.00 2.01-3.00 3.01-4.00 4.01-5.00 ETM+ 2005 ETM+ 2010 5.01-6.00 > 6
2001 2003 2005 2002 2004 2006 Saginaw Bay August MODIS Image Time Series LLLLLLLLLLLLLLLLLLLLLLLLLLL MODIS August 13, 2006 Chlorophyll-a (µg/l) Land Clouds 0-1.00 1.01-2.00 2.01-3.00 3.01-4.00 4.01-5.00 5.01-6.00 6.01-7.00 7.01-8.00 8.01-9.00 9.01-10.0 > 10.0
Final Thoughts Remote sensing is useful tool in algal bloom detection Cost effective Easy to do Can do long-term water quality assessments
Questions?