Measuring Water Clarity and Quality in Minnesota Lakes and Rivers: A Census-Based Approach Using Remote-Sensing Techniques
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Measuring Water Clarity and Quality in Minnesota Lakes and Rivers: A Census-Based Approach Using Remote-Sensing Techniques by Patrick L. Brezonik, Leif G. Olmanson, Marvin E. Bauer, and Steven M. Kloiber Photo © The Regents of the University of Minnesota. Used with permission of the Metropolitan Design Center. The St. Croix River near Washington County, Minnesota. O ver the past few decades, marvels. Google Earth now provides production, and monitor forest health. satellite-derived information has detailed images on our television and More recently, the MODIS satellites become an integral part of our computer screens of far-off places using have been used to estimate terrestrial daily life. It is difficult to imagine how high-resolution satellite images. In addi- primary production on a continental crude and inaccurate weather forecasts tion, GPS systems based on satellites scale and to monitor oceanic levels of would be without the daily satellite now enable even “directionally chal- chlorophyll on a global scale. Satel- feeds that drive the complex models lenged” drivers to head their cars to the lites also provide essential hydrologic on which weather forecasts are based, desired destination and make it possible information (e.g., snow pack, rainfall, and of course the satellite images them- to locate and track people in the most and evapotranspiration estimates) selves have revolutionized how we view remote and inaccessible locations. for water resources management. the movement (and scale) of weather Satellite technology also has had Despite many efforts reported in patterns. Similarly, satellite technology profound impacts on the Earth sciences the scientific literature during the past has become such an essential and and on science-based management three decades, however, procedures using commonplace component of global of natural resources. For example, satellite imagery to measure surface telecommunications, television, and Landsat imagery has been used for water quality have not been adopted on Internet systems that few give a second several decades to analyze land-use a routine basis. For the past nine years, thought to the technology behind these and land-cover patterns, estimate crop we have been working with colleagues SUMMER 2007 3
and students to change this state of algal abundance. Water clarity is useful image; we found that measurements affairs and to develop workable proce- in measuring water quality because taken within a few days (±3 to 7 days) dures for routine use of satellite imagery it relates directly to both human-use of image acquisition still provide strong and related technology to assess surface perceptions of quality (clear water gener- relationships. This is because water water-quality conditions by Minnesota’s ally being preferred, especially for swim- clarity (Secchi depth) usually does not water management agencies. Most ming) and to the abundance of algae. exhibit large and rapid fluctuations of our work has focused on lakes and Thus, water clarity is an indirect measure in a given lake (although there are water clarity as a simple measure related of a lake’s “trophic state”—its status in strong seasonal patterns in clarity). to users’ perceptions of water quality. terms of nutrient concentrations and The general predictive equa- However, we also have explored the use biological productivity. Because of its tion that we found for water of satellite imagery to measure other simplicity, Secchi depth is one of the clarity estimation has the form: water-quality characteristics, such as most frequently measured properties chlorophyll (a measure of the abundance of lakes, and many citizen monitoring ln(SD) = a(TM1/TM3) + b(TM1) + c of algae in water) and humic matter programs have been developed over the (natural derivatives from plants that past quarter century to supplement the where a, b, and c are coefficients fit to produce the brown stain in wetlands monitoring by government agencies. the calibration data by the regression and water draining from forested areas), Nonetheless, only about 10% of the lakes analysis, ln(SD) is the natural logarithm and evaluated the potential of remote- in the Twin Cities metropolitan area are of the Secchi depth for a given lake, and sensing techniques to identify and map monitored for water clarity in any given TM1 and TM3 are the brightness values aquatic vegetation in lakes and wetlands. year, and in some parts of the state, the measured by the Landsat sensor in the We have explored the capabilities and fraction monitored is much lower. blue and red bands, respectively, for limitations of several satellite sensor In our early studies on poten- a pre-defined area of the lake surface. systems, as well as aircraft-mounted tial applications of satellite imagery Once coefficients have been fitted to sensors with high spectral and spatial to measure surface water quality, we the equation, it can be used to infer resolution that provide special capa- found close correlations between water Secchi depth values for all other lakes bilities for remote measurements of clarity, as measured by the Secchi disk, in the Landsat image. The images are water quality in rivers and streams. and light in the blue and red bands of approximately 110 miles on a side and This article summarizes progress we the spectrum reflected from lake water may contain hundreds of lakes and have made in these areas with emphasis surfaces and measured as “brightness” by ponds. The spatial resolution (so-called on water clarity and provides some satellite sensors. We used this informa- pixel size) of brightness data in the perspectives on the likely future role tion to develop procedures to estimate Landsat images is 30 meters (roughly of satellite- and aircraft-based remote- water clarity in terms of “inferred Secchi 100 feet) on a side. Because we need sensing techniques in water resource depth” using images gathered routinely to use data only from pixels for water monitoring and management in Minne- (every 16 days, assuming no interference surfaces not affected by land or aquatic sota. The work described here has been by cloud cover) by the Landsat satel- vegetation, the net effect is that the supported by a variety of state and federal lites. With financial support from the procedure is capable of obtaining high- agencies, including the Metropolitan Metropolitan Council, our initial studies quality results on lakes larger than Council, Minnesota Pollution Control focused on lakes in the Twin Cities metro about 20 acres (8 hectares). For the Agency, Department of Natural Resources, area, but we later expanded our coverage seven-county Twin Cities metro area, Legislative Commission on Minnesota across the entire state. Colleagues have which constitutes part of one Landsat Resources, NASA, and the U.S. EPA. The applied our techniques for lake clarity image, this translates to approximately senior author was the 2003–2004 Fesler- monitoring in Wisconsin and Michigan, 550 lakes and open-water wetlands.1 Lampert Chair in Urban and Regional and recently we conducted pilot studies Affairs, and some of the work described in to evaluate the usefulness of the tech- Lake Water Clarity in the Twin Cities this paper was supported by that position. nique in Ohio, Indiana, and Illinois. Metropolitan Area Because atmospheric conditions (e.g., Our initial studies on the method estab- Ground and Satellite Measurements of haze and water vapor content) affect the lished that it works in a practical sense Lake Water Clarity light reflected by land and water surfaces and that it provides reliable data not In essence, clarity measures the distance as it travels back toward satellite sensors, otherwise available. Regarding prac- that light penetrates in a water body. no universal predictive relationship ticality, historical Landsat images are In lakes, water clarity commonly is exists between water clarity and satellite available from the EROS Data Center in measured by a simple device called a sensor response. Instead, it is necessary Sioux Falls, South Dakota, for relatively Secchi disk—a 20-cm (8-inch) diameter to calibrate the general relationship low cost (a few hundred dollars), and white disk—that is lowered into the applicable to Landsat and Secchi depth many images are available for free in water column until it can no longer data for each Landsat image. We do this existing local archives. Once the proce- be seen. The depth of disappearance is using standard regression procedures and dure is established for a given region, called the Secchi depth. Three types of Secchi depth measurements collected the processing time per image for an constituents affect water clarity: algae by existing monitoring programs on experienced analyst is less than a week, and algal-derived particles; natural, a few lakes (at least 20 and sometimes and the processing uses software that is colored organic matter (humic matter); as many as 100 per image) near the 1 and soil-derived clay and silt particles. In time the satellite image was taken. The The above description is a simplification of the procedures used to convert Landsat data into Minnesota lakes, soil-derived turbidity ground-based Secchi depth data do inferred lake clarity (Secchi depth) values. Details usually is not important, and water not need to be collected at exactly the on the method are found in a manual available clarity most commonly is related to same time that Landsat acquires an online at http://water.umn.edu. 4 CURA REPORTER
readily available in remote-sensing labo- Figure 1. Water Clarity (SD) for Twin Cities Metropolitan Area Lakes (inferred from ratories. A search of historical archives Landsat TM image), September 1998 showed that at least one cloud-free image is available for the Twin Cities metro area nearly every year for a late- summer index period (late July to early September), which we found to be the optimal time to assess water clarity in Minnesota lakes. Regarding data reli- ability, R2 values (a measure of goodness of fit) for the regression relationships to establish the coefficients of the predic- tive equation normally are in the range of 0.8 to 0.9, meaning that they explain 80 to 90% of the variance in the relation- ship, and occasionally they are as high as 0.95 (95% of the variance). Given that ground-based measurements of Secchi depth are themselves subject to some imprecision, we consider this to be quite acceptable. In addition, excellent agree- ment between satellite-estimated and ground-observed Secchi depth trends can be achieved. For example, Secchi depth temporal trends during the 25-year record were found to be virtually iden- tical in several metro-area lakes for which long-term ground-based data were avail- able to compare with Landsat-inferred results. For Coon Lake (Anoka County), Secchi depth was found to increase at a rate of 0.062 m per year during the period 1973–1998 based on in-lake measurements, and a best-fit regression equation of the satellite-inferred Secchi depth yielded an increase of 0.067 m Note: One (1) meter equals 3.3 feet. per year during the same time period. In subsequent studies on water clarity Results for 1998, the last year in this and higher clarity in many lakes. The in metro-area lakes, we were interested in initial analysis (Figure 1), demonstrate 1996 image was acquired in mid-July, evaluating spatial and temporal patterns that at any given time lakes in the just outside the late-summer window in lake clarity. Specifically, we wanted to metro area exhibit a wide range of water that we found best represents the period answer the following questions: (1) Does clarity. Analysis of the temporal trend of lowest water clarity in Minnesota lake water clarity in the metro area vary results shown in Table 1 indicates that lakes, and this also may explain the depending on climatic conditions, such climatic conditions have an important higher water clarity values in that year. as extended periods of drought or above- influence on water clarity in the region’s Several statistical tests were used average precipitation?; (2) Can long-term lakes. For example, 1988, which was to examine temporal differences trends in water clarity of metro area lakes one of the hottest and driest summers and trends across years in lake water be related to well-documented changes on record in the region, had the highest clarity. A Kruskal-Wallis test showed no in land-use/cover within the region?; fraction of lakes in the lowest clarity significant differences in metro-area and (3) How strong are the relation- class and one of the lowest fractions regional lake water clarity between ships between lake clarity and land-use/ of lakes in the highest clarity class the years 1983, 1986, 1991, 1995, land-cover conditions in the landscape among all the study years. Hot sunny and 1998, but water clarity was below around lakes? To address these questions, weather, combined with low rainfall normal in 1973 and 1988 and above we first analyzed a series of 10 images that produced little flushing of water normal in 1975, 1993, and 1996. from the late-summer index period through the lakes, yielded heavy algal Although no simple long-term trend covering a 25-year interval (1973–1998). blooms and low water clarity in many in lake water clarity was detected at the Ground-based calibration data were lakes. In contrast, 1993 and 1996, which scale of the entire Twin Cities metro obtained for each image, primarily from were wet and relatively cool summers, area, a Kendall Tau-b analysis for each of the Metropolitan Council’s citizen moni- had low fractions of lakes in the lowest the approximately 500 metro-area lakes toring program, and Landsat-inferred clarity class and high fractions in the with sufficient data to conduct a trend Secchi-depth values were obtained by highest clarity class. The combination analysis indicates that 49 lakes (10% the procedures described above for of cooler temperatures, less sun, and of the total) had significant temporal approximately 575 lakes and open-water greater flushing of water through the trends. Water clarity improved in 34 wetlands in the Twin Cities metro area. lakes resulted in lower amounts of algae lakes since 1973 and declined in only SUMMER 2007 5
Table 1. Water Clarity Trends for Twin Cities Metropolitan Area Lakes, 1973–1998 Number of lakes, by month and year Secchi Depth (SD) July Aug Sept Aug Aug Sept Aug July July Sept 1973 1975 1983 1986 1988 1991 1993 1995 1996 1998 > 4 meters 7 4 14 0 3 0 0 1 4 0 (1.6%) (1.0%) (3.2%) (0.0%) (0.6%) (0.0%) (0.0%) (0.2%) (0.8%) (0.0%) 2–4 meters 94 79 61 94 77 75 68 86 118 96 (21.1%) (18.9%) (14.0%) (19.5%) (16.4%) (15.5%) (14.0%) (17.7%) (23.9%) (19.7%) 1–2 meters 192 165 182 160 120 187 218 192 245 184 (43.2%) (39.6%) (41.7%) (33.3%) (25.5%) (38.6%) (45.0%) (39.6%) (49.6%) (37.8%) 0.5–1 meter 122 137 114 168 168 173 144 151 104 149 (27.4%) (32.9%) (26.1%) (34.9%) (35.7%) (35.7%) (29.8%) (31.1%) (21.1%) (30.6%) < 0.5 meter 30 32 65 59 102 49 54 55 23 58 (6.7%) (7.7%) (14.9%) (12.3%) (21.7%) (10.1%) (11.2%) (11.3%) (4.7%) (11.9%) Total 445 417 436 481 470 484 484 485 494 487 Note: One (1) meter equals 3.3 feet. Column percentages may not total 100% due to rounding. 15 lakes. The most common trend (43% adopted the procedure and is conducting brightness signatures from 19 lakes with of lakes with trends) was an increase annual assessments of water clarity in Secchi depth measured within three days in clarity of 4 to 8% per year. Lakes metro-area lakes. Results for 2003, 2004, of the image date yielded a strong fit with decreasing clarity generally were and 2005 are available from MCES. to the equation model described earlier small or quite shallow, suggesting that (R2 = 0.82, meaning that the relationship morphometry (the physical shape and Water Clarity in Ponds and Small Lakes explains more than 80% of the varia- size characteristics of a lake) may predis- The relatively modest spatial resolu- tion in the data). We created a pixel-level pose a lake to respond more strongly to tion of the Landsat sensor (30 m) limits map of water clarity of small lakes and changes in natural or human-induced the usefulness of Landsat imagery to ponds in Eagan, as shown in Figure 2. stresses. The finding that more lakes assess water clarity conditions in small Only 14 of Eagan’s 375 lakes, ponds, had improving rather than worsening lakes and ponds. Newer commercial and wetlands were large enough to be clarity during the 25-year record at satellites with spatial resolution (pixel included in our previous metro-area first was surprising, given the major sizes) of 1 to 4 meters overcome this assessments and 22 were included in our geographic expansion of the Twin Cities limitation and allow assessment of statewide assessments (see below), but footprint that occurred during that water clarity in neighborhood ponds the IKONOS image assessed 236 water time period. However, suburban sprawl much smaller than an acre in area. Two bodies. The wide range of water clarity in has occurred largely at the expense of such commercial satellites, IKONOS small ponds within a small geographic agricultural lands rather than more and Quickbird, are the primary sources area is striking and likely reflects both pristine forested areas, and croplands of the amazing aerial images shown natural factors (pond depth and water- often produce higher levels of nutri- increasingly in television news broad- shed area) and effects of landscape use ents in storm-water runoff than do casts and the imagery that provides the and management practices. We found suburban residential areas. Moreover, basis for Google Earth. Both satellites similar results for images obtained in the salutary effects of efforts to manage have sensors similar to the Landsat TM 2001 and 2002. The cost of IKONOS the quality of urban and suburban sensors in spectral characteristics (e.g., data ($2,500 to $4,000 per image) prob- storm-water runoff, and also to directly three broad bands in the visible spec- ably would be too large for many cities manage water-quality conditions in trum and one in the near infrared). just for lake-clarity assessments, but lakes, should not be discounted. Figure 2 shows an image for Eagan, should be justifiable if the images also Since the studies described above Minnesota, acquired on August 23, are used to monitor changes in land were completed, we have continued to 2000, by the IKONOS satellite. The use, land cover, and wetland status. assess lake water clarity across the Twin city of Eagan, located in the south- Cities metro area. Additional informa- eastern metro area (just southeast of Statewide Census-Level Assessment of tion for 2000, 2003, and 2005 is available the Minneapolis–St. Paul Airport) and Lake Water Clarity on the “Twin Cities Lake Browser” (see roughly 6 by 6 miles (about the size of Starting in 2000, we expanded our http://water.umn.edu), through which an IKONOS image), has about 375 small analysis of lake water clarity in Minne- one can locate individual lakes of interest lakes, ponds, and open-water wetlands sota from the metro region to the and obtain data for the entire period greater than one acre in area. The image entire state. Since that time, we have of record. Perhaps more significant is was processed using the basic proce- completed five censuses at approxi- the fact that the Metropolitan Council dures we developed for Landsat imagery. mately five-year intervals for the 20- Environmental Services (MCES) has Multiple regression analysis of the year period 1985–2005. Each census 6 CURA REPORTER
Figure 2. IKONOS Image Showing Water Clarity for Lakes and Ponds in Eagan, Minnesota, August 23, 2000 0 1 2 3 4 5 6 Miles comprises information on more than Landsat images for parts of the state spatial and temporal trends. Figure 3 10,000 lakes and open-water wetlands during the late-summer measurement illustrates the distribution of lake water across the entire state of Minnesota. period, the statewide censuses usually clarity across the state for the year 2005 The techniques used to produce these are composites of analyzed images (a similar image for 2000 is available results are similar to those we devel- acquired within plus or minus one from the authors as a large poster suit- oped for the metro-area assessments. year of the nominal year for a census. able for framing and classroom use), For the statewide assessments, we We currently are in the process of and Table 2 summarizes temporal trends generally sought Landsat images where analyzing this data set, which we believe in the numbers of Minnesota lakes in ideally each entire path across the state is unprecedented in size and scope, for different water-clarity classes for the yielded useable information (clear skies). Landsat orbits Earth such that it Table 2. Water Clarity in Minnesota Lakes Based on Landsat-Inferred Secchi Depth traverses Minnesota in southwest-to- (SD), 1985–2005 northeast paths that are approximately 110 miles wide, and the trajectory is Number of lakes, by year Secchi such that successive paths partially Depth (SD) overlap. By using entire paths across the 1985 1990 1995 2000 2005 state rather than single images, we were > 4 meters 1,043 1,213 1,059 909 1,254 able to extend the geographic range with available ground data for calibra- 2–4 meters 4,917 3,981 4,649 4,641 4,363 tion purposes; some remote areas in northern Minnesota, for example, have 1–2 meters 3,080 3,150 3,356 2,637 3,091 very limited ground-based measure- ments. The overlapping nature of the 0.5–1 meter 1,627 1,898 1,670 1,729 1,902 Landsat paths also meant that multiple satellite-inferred Secchi depth measure- < 0.5 meter 469 490 254 600 631 ments were obtained for approximately 60% of the state’s lakes in each census. Total 11,136 10,732 10,988 10,516 11,241 Because clouds sometimes obscure Note: One (1) meter equals 3.3 feet. SUMMER 2007 7
Figure 3. Landsat-TM–Based Census of Water Clarity in Minnesota Lakes, 2005 record (through 1998). However, when the entire 32-year study period is exam- ined, no simple linear trend is apparent regarding the extent to which metro-area lakes support swimming. Instead, a cyclic pattern is evident that may be related to climatic conditions. For example, the fraction of lakes in the non-supporting category reached a maximum in the late 1980s, corresponding to the end of a long-term drought, and a minimum in the mid-1990s, corresponding to a period of above-average rainfall in the region. The fraction of lakes in the fully supporting category showed the opposite trend (minimum in the late 1980s and maximum in the mid-1990s). Thus, it appears that climatic conditions play an important role in the temporal pattern of lake water clarity at the region scale. With the larger database available from the statewide censuses, we also have been able to examine relationships between lake clarity and land-use/land-cover conditions in the surrounding terrestrial landscape, a goal we were not able to achieve in our earlier studies focused only on the Twin Cities metro area. For example, contrasts in land use/land cover among the three largest eco- regions in the state (Figure 5) are obvious, and patterns of lake clarity appear to reflect these differences. The Northern Lakes and Forests (NLF) eco- region in northeastern Minnesota is more than 60% forested and only roughly 6% agricultural; the Western Cornbelt Plains (WCP) ecoregion in southern Minnesota is only 4% forested, period of record. Overall, the distribution classes based on water clarity over the and agricultural land occupies more of lakes across the water-clarity classes extended period of record, 1973–2005. than 80% of the region; and the North is fairly stable throughout the 20-year The Minnesota Pollution Control Agency Central Hardwood Forest (NCHF) ecore- period 1985–2005, but a few differences developed criteria for these use-support gion in the center of the state has inter- stand out. For example, substantially classes based on algal bloom problems mediate land-use/land-cover conditions fewer lakes were found in the lowest associated with various concentrations of (12% forested and 50% agricultural water clarity class (< 0.5 m) in 1995 total phosphorus (TP) in lakes. Based on land). Reflecting these ecoregion differ- than in all other years, and in 2000 established relationships between TP and ences, water clarity in NLF lakes aver- there were fewer lakes with water clarity Secchi depth, we converted the criteria ages slightly more than 3 m, and about values in the 1 to 2 m range but more into equivalent Secchi depth values. For 70% of the lakes have values in the 2 to lakes with values in the 3 to 4 m range example, a TP concentration of about 80 4 m range. In contrast, water clarity than in other years. Information on micrograms per liter (µg/L) corresponds in WCP lakes averages just less than trends in lake clarity for individual coun- to a Secchi depth of 0.62 m (or 2.0 ft) 1.0 m, and only about 8% of the lakes ties and across Minnesota’s ecoregions in lakes where phosphorus controls the in this region have clarity values in can be found at http://water.umn.edu, growth of algae. Lakes with higher TP the 2 to 4 m range (and very few have and the “Lake Browser” on this site concentrations (hence, lower Secchi values greater than 4 m). Lakes in the allows users to locate individual lakes depth values) than that are considered NCHF exhibit a wide range of clarity, across the state and obtain water to be “non-supporting” (that is, unfit) but the average value is about 1.5 m, clarity and other information. for swimming because they have a high and roughly 40% of the lakes have Of interest regarding the statewide frequency of nuisance algal blooms. clarity values in the 1 to 2 m range. census information is how it extends The results in Figure 4 suggest a gradual Of course, other factors aside the earlier information on lake clarity trend in the fraction of metro-area lakes from land use/land cover affect lake within the Twin Cities metro area. in the non-supporting category during clarity—water column depth, under- Figure 4 shows the distribution of TCMA the first half of this decade, compared lying geology, and population density lakes in four “swimming use-support” with results for the end of the earlier in lake catchments also play important 8 CURA REPORTER
Figure 4. Swimming Use Support for Twin Cities Metropolitan Area Lakes, 1973–2005 100% 90% 80% 70% 60% Full 50% Full (Marginal) Partially Impaired 40% Non-Supporting 30% 20% 10% 0% Note: Swimming use support based on water clarity (Secchi depth, SD). SD values defining the use-support classes are as follows: full, > 1.60 meter; full (marginal), 1.21–1.60 meter; partially impaired, 0.80–1.20 meter; non-supporting, < 0.80 meter. One (1) meter equals 3.3 feet. roles. In general, lakes in the NLF tend Aircraft-Based Remote Sensing of Water offer an attractive possibility for moni- to be deeper than lakes in the WCP, Quality in Rivers and Streams toring river water quality and obtaining and geology and climate are important Satellite imagery works well to assess more complete spatial coverage than is factors that predispose the NLF to be water clarity and related optical proper- possible with ground-based sampling. much more forested than the WCF. ties of lakes (such as chlorophyll and To explore the usefulness of this tech- Striking relationships between humic color), but geometric consider- nique, we conducted a series of studies lake water clarity and land use/cover ations make satellite imagery a poor in the summers of 2004 and 2005 on also were found when the data were choice for remote sensing of rivers and rivers in the Twin Cities metro area examined at the county level (Figure streams. As they pass through the Twin and nearby areas using a small plane 6). Across Minnesota counties, average Cities metro area, only large rivers like and sensor system operated by the lake clarity increases with increasing the Minnesota and Mississippi Rivers are University of Nebraska. In contrast to percentages of forested land and wide enough to have pixels representing the Landsat TM and IKONOS sensors, decreases with increasing percentages water surfaces that are not affected which have only a few, broad spec- of agricultural and urban land. The by shoreline and land surfaces. High- tral bands, the sensors we used in the rate of decrease with degree of urban resolution satellites like IKONOS do have aircraft studies collected reflected light land is especially dramatic; average lake the required spatial resolution, but acqui- in many (narrow) spectral bands, and clarity declines to only about 1.0 m sition of such images for a particular this provided much better opportunities (3.3 ft) at 10% urban land in a county. date and time is not a straightforward to develop stronger predictive relation- Others have reported similar effects of proposition, and the relatively small ships for various optically related water- urbanization and extent of impervious areas these images cover lead to impracti- quality characteristics. Support for this area on the biotic integrity of streams. cally large costs per river mile monitored. work was provided by the Legislative Digital map layers delineating indi- Multispectral sensors similar to Commission on Minnesota Resources vidual lake catchments are becoming those in satellites like Landsat and through the Minnesota Pollution available for Minnesota, and it will be IKONOS (or with higher spectral reso- Control Agency (MPCA), and collec- interesting to see whether relationships lution) that are deployable in small tion of ground-based calibration data with lake clarity improve when land- aircraft are commercially available. was achieved by a collaborative effort use analyses are conducted at the scale Because aircraft can be flown precisely of the MPCA and the Metropolitan of individual lake catchment areas. on flight paths along river reaches, they Council Environmental Services. SUMMER 2007 9
Figure 5. Land-Use/Land-Cover Distribution and Water Clarity in Three Minnesota Ecoregions Northern Lakes North Central Western Cornbelt and Forest Hardwood Forest Plains Ecoregions Water Wetland 1.2% 0.9% Forest Deciduous Shrubland Urban Shrubland 3.7% 0.6% Wetland Urban 0.9% Agriculture 4.6% 5.2% 4.6% Forest Coniferous Urban 6.1% 0.1% 1.6% Wetland Grassland 4.1% 5.4% Water 6.1% Shrubland Grassland 13.5% 9.2% Land Uses Water 8.9% Forest Coniferous Forest Deciduous 18.0% 10.9% Agriculture 49.5% Grassland Agriculture 18.2% 82.7% Forest Deciduous 42.9% Forest Coniferous 0.8% 4000 1600 350 Number of Lakes 3500 1400 300 3000 1200 250 2500 1000 200 2000 800 150 1500 600 100 1000 400 500 200 50 0 0 0 > 4.0 2.0–4.0 1.0–2.0 0.5–1.0 < 0.5 > 4.0 2.0–4.0 1.0–2.0 0.5–1.0 < 0.5 > 4.0 2.0–4.0 1.0–2.0 0.5–1.0 < 0.5 Water Clarity Ranges (meters) Source: Land-cover data are from 1990 Minnesota Land Cover GAP. Secchi depth data are from 2000. Note: One (1) meter equals 3.3 feet. 10 CURA REPORTER
Figure 6. Relationship between Average Secchi Depth (SD) for Lakes in each Minne- Procedures used to obtain and sota County and Percentage of Forested, Agricultural, or Urban Land analyze data from the aircraft-mounted sensor were generally similar to those 4 we have been using for satellite data. Because of the high degree of spatial Countywide Average Secchi Depth (m) 3.5 and temporal variability in optical properties (such as turbidity, water 3 clarity, and chlorophyll concentra- tions) of flowing waters, collection 2.5 of water samples from the rivers was y = 0.28x + 0.97 coordinated to occur near the same 2 R² = 0.64 time as the aircraft overpass. Water clarity was measured on river samples 1.5 using a Transparency Tube (T-tube) that provides data analogous to Secchi depth 1 but is easier to use in flowing waters. Statistically strong relationships were 0.5 found between ground-based measure- ments of water-quality characteristics 0 and the brightness of light at specific 0 20 40 60 80 100 wavelengths collected by the aircraft- Percent Forest mounted sensor (R2 = 0.94 for T-tube water clarity and 0.72 for chlorophyll a). 4 The exploratory studies were successful and yielded maps for several Countywide Average Secchi Depth (m) 3.5 optically related water-quality character- istics, including turbidity, water clarity, 3 suspended solids, and chlorophyll, along several major river reaches in the 2.5 region. Typical results are shown in Figure 7, which illustrates the distribu- 2 tion of water clarity (a) and chlorophyll (b) along a stretch of the Mississippi y = -0.023x + 3.17 1.5 River starting at highly turbid Spring R² = 0.73 Lake (above Lock and Dam No. 2) 1 near Hastings, Minnesota. The influ- ence of water low in suspended matter 0.5 from the St. Croix River on clarity in the Mississippi River is apparent, and 0 the figure shows that water clarity 0 20 40 60 80 100 improved further as suspended matter Percent Agriculture + Urban settled out over a distance of about 3 to 4 miles from the confluence of 4 the St. Croix. Chlorophyll levels were very high in parts of Spring Lake (but Countywide Average Secchi Depth (m) 3.5 spatially highly variable) and the Missis- sippi River downstream of the lake, 3 y = -0.29x + 3.31 but concentrations decreased mark- R² = 0.56 edly after confluence with the cleaner 2.5 waters of the St. Croix River. As illus- trated in Figure 7, in typical ground- 2 based routine monitoring programs on rivers, one or perhaps a few sampling 1.5 locations along the river would be sampled and the high spatial variability 1 of water quality would be missed. 0.5 Concluding Comments: Prospects for the Future 0 Effective management of natural 0 20 40 60 80 100 resources depends on accurate and Percent Urban complete information, as well as an informed public. We view the remote- Note: One (1) meter equals 3.3 feet. Trend line for urban area is based only on counties with urban area < 11%. sensing techniques for information SUMMER 2007 11
Figure 7. Spatial Distribution of (a) Water Clarity Measured by a T-tube and (b) Total Chlorophyll a in the Mississippi River from Spring Lake to Downstream of the Confluence with the St. Croix River Note: Based on aircraft-based imagery collected August 19, 2004. 12 CURA REPORTER
Photo courtesy of the Minnesota Pollution Control Agency Science Foundation since August 2004. His research focuses on understanding human perturbations on biogeochemical cycles of nutrients and metals in aquatic systems, and applications of remote sensing to regional-scale assessment of water quality. Leif G. Olmanson is a research fellow in the Department of Forest Resources and the Remote Sensing and Geospatial Anal- ysis Laboratory at the University of Minne- sota, and a Ph.D. student in the Natural Resources Science and Management Program. His research focuses on devel- opment of satellite imagery and aircraft- based remote-sensing tools to measure the status and trends of ecological condi- tions in lakes, rivers, and wetlands. Marvin E. Bauer is a professor in the Department of Forest Resources and director of the Remote Sensing and Geospatial Analysis Laboratory at the University of Minne- sota. His research focuses on applications of digital remote sensing to inventory, monitor, and analyze land, vegetation, and water resources. Steven M. Kloiber This photo shows the dramatic difference in water clarity at the confluence of the St. Croix River (blue water) and the highly turbid Mississippi River. is a senior planner with the Metropolitan Council Environmental Services in St. Paul. gathering on surface water described in tion. Development of techniques for His Ph.D. dissertation in civil engineering this article as tools that will help achieve automated correction of satellite data on the application of satellite imagery to both of these goals. Satellite imagery to account for atmospheric effects measurement of lake water clarity trends provides a cost-effective means to obtain should be possible, and this would in the Twin Cities metro area provided a comprehensive spatial coverage on the eliminate the need to gather ground- foundation for our research. His current status and trends in key characteristics based calibration data for every image. research focuses on understanding the of surface-water quality, and the results Our recent work with aircraft- impacts of urban and suburban devel- can be made readily available to the mounted sensors is especially promising opment on surface water quality and public in easily understood formats. in that it provides a means of gath- nutrient export from watersheds. Of course, these methods cannot tell ering spatially comprehensive infor- The research upon which this article us everything we would like to know mation on water quality in rivers and is based was supported in part through about water quality; satellite sensors streams heretofore not practical with funds provided by the Fesler-Lampert cannot detect the presence of poten- ground-based sampling programs. Such Chair in Urban and Regional Affairs, one tially harmful metal pollutants, organic comprehensive information, as shown of four endowed chairs and two named pollutants, or bacterial pathogens in for the Mississippi River in Figure 7, professorships made possible through a water. However, the sensors do provide could be used to provide much more generous contribution to the University accurate information about water clarity robust validation of models used to of Minnesota by David and Elizabeth and the occurrence of algal blooms in forecast water-quality conditions than Fesler. The Fesler-Lampert Endowment water, and these are common issues current ground-based methods allow. in Interdisciplinary and Graduate Studies affecting the usability as well as the Moreover, the compelling nature of was initially established in 1985 to economic value of Minnesota lakes. the visual images themselves should stimulate interdisciplinary research and Scientists in other states have help both scientists and the public to teaching through the appointment of adopted the methods we developed to gain a deeper appreciation and better distinguished, broadly learned schol- study Minnesota lakes, and resource understanding of our aquatic resources. ars to endowed faculty positions at the management agencies in Minnesota University of Minnesota. Additional sup- and elsewhere are starting to adopt Patrick L. Brezonik is a professor in port was provided by the Metropolitan these same methods as routine. Pros- the Department of Civil Engineering at Council, Minnesota Pollution Control pects for future improvements of the the University of Minnesota and held Agency, Department of Natural Resources, techniques are bright, because satellite the Fesler-Lampert Chair in Urban and Legislative Commission on Minnesota sensors likely to be developed during Regional Affairs in 2003–2004. He has been Resources, National Aeronautics and Space the next decade should provide better on leave serving as program director for Administration (NASA), and the U.S. spectral, spatial, and temporal resolu- environmental engineering at the National Environmental Protection Agency (EPA). SUMMER 2007 13
You can also read