Ecological Indicators for the Nation (2000)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

4 Indicators for National Ecological Assessments ' n the preceding chapters, we have discussed the desirable characteris- tics of indicators, the sources of data that underlie them, the models ~ that support them, and the criteria by which good indicators can be identified. Based on those discussions and the conceptual model we used, the committee recommends national indicators for three major categories of ecological information. These categories encompass the nation's most important ecological issues. By computing them and paying attention to them, the nation should be aware of the status of its ecosystems, be alerted to changes that may require management interventions and policy changes, and have a basis for ensuring that future generations will have access to ecosystem goods and service as rich as those enjoyed today. In some cases, noted for each indicator, some experience will need to be gained on details of the indicator's behavior, but all the indicators are based on soundly established scientific experience and principles. The proposed indicators are in general applicable to both managed (e.g., agri- cultural) and unmanaged ecosystems; the indicators of nutrient-use effi- ciency and overall nutrient balance are specific to agricultural ecosystems. · Information about the extent and status of the land use and cover types that together make up the nation's ecosystems. Information about the extent of the nation's land use and cover types informs us about the extent of ecosystem types in the nation, and it is needed to calculate several other indicators. The information and technology to calculate land cover is currently available. Land use in some ways is more informa- 64

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 65 live because it provides additional information on the status of areas and hence their ability to provide goods and services. Also, information on how land is used is predictive of future land cover and hence predicts the ability to provide goods and services. However, a land use indicator requires much synthesis of existing information and some new informa- tion, and thus will take longer to develop than a land cover indicator. Meanwhile, land cover can serve as a valuable indicator. · Information about the nation's ecological capital. This informa- tion measures the nation's natural capital, or raw materials, both abiotic and biotic. Abiotic raw materials essential for ecosystem functioning include soil (also partly biotic) and its nutrients. Biotic raw material includes the number of species still present in the country relative to their number at the time of European contact, their distribution over today's natural and human-modified environments, and the number of species present today that are exotics, those species introduced, deliberately or inadvertently, that have become established or naturalized. Indicators of biological capital are important for several reasons. Knowing what portion of our biological capital is native provides a sensi- tive measure of humans' environmental impacts, as described below. Assessing the status of the nation's biological capital is important for ethical and aesthetic reasons. Also, biological diversity is an indicator of the capacity of ecosystems to function effectively. An important but controversial theory holds that because species differ, species-rich eco- systems are more likely than species-poor ecosystems to contain some species that can thrive during an environmental perturbation (Mooney et al. 1996~. As a result, species-rich systems should be buffered against disturbances and continue to perform better in fluctuating environments than species-poor systems. Empirical studies are still few, but they pro- vide some support for the hypothesis (Tilman 1996, Tilman and Downing 1994, Naeem et al. 1994~. Further investigations will clarify relationships between biological diversity and ecosystem processes. Meanwhile, it is prudent to monitor the status of the nation's biological resources. · Information about the functioning (performance) of the nation's ecosystems and how it is changing. This information includes measures of productivity and other ecosystem processes. Changes in the produc- tivity of ecosystems are generally accompanied by changes in their ability to provide goods and services. Usually, declines in productivity are undesirable, but in freshwater ecosystems, increases in productivity asso- ciated with eutrophication can be undesirable. For each of these major categories of information we recommend indicators (Table 4.1) that are described in detail below. Although the categories apply to all ecosystems, they differ in many details for marine,

66 ECOLOGICAL INDICATORS FOR THE NATION TABLE 4.1 National Indicators of Ecological Condition Category of Recommended Reasons for EcologicalInformation Indicators Choosing Indicator Extent and Status of the Nation's Ecosystems Ecological Capital Biotic Raw Materials Land Cover and Land Use* Total Species Diversity Native Species Diversity Needed for calculation of most other indicators. Inform us about the overall extent of different ecosystem types. Measures nation's biological resources (what is present relative to what is expected). Measures the amount of biological diversity that is native. Ecological Capital Nutrient Runoff Estimator of total losses of Abiotic Raw Materials nutrients. Nutrient runoff has major effects on receiving waters. Soil Organic Matter* Best single indicator of soil condition, related to erosion. Ecological Functioning (Performance) Productivity, including Carbon Storage, Net Primary Production (NPP), and Production Capacity Lake Trophic Status Stream Oxygen Soil Organic Matter* Nutrient-Use Efficiency and Nutrient Balance Land Use* Direct measures of the amount of carbon sequestered or retained in an ecosystem (NEP), energy and carbon brought into an ecosystem (NPP), and energy- capturing capacity of ecosystems (chlorophyll). Direct measure of ability of lakes to provide goods and services. Captures the balance between instream primary production and respiration. Single most important indicator of soil quality and productivity. Inefficient use of nutrients is costly in terms of economics and damage to ecosystems to which nutrients are discharged. Provides information about ecosystem functioning. *Indicators in more than one category.

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 67 freshwater, and terrestrial ecosystems. Because marine ecosystems were not specifically covered by the committee's charge, we focus on terrestrial and freshwater ecosystems and the interface between them-wetlands. More research is needed for the full implementation of these indicators. We offer some suggestions for the order in which some components might best be initiated. Although the national-level indicators we propose are highly aggre- gated, most of them require detailed input data. If these data are archived in a disaggregated form, the indicators can be computed in a variety of ways to provide a rich array of indicators of great local and regional value. Chapter 5 describes how that can be done and describes additional regional and local indicators. THE EXTENT AND STATUS OF THE NATION'S ECOSYSTEMS To estimate the capacity of U.S. ecosystems to continue to provide the goods and services that society depends on, one needs to know the status of the different types of natural and human-modified ecosystems and how much of each major type of ecosystem remains in the country. Infor- mation is also needed on what the committee calls the matrix that the ecosystems are in, i.e., the physical aspects of the land. Thus the indica- tors in this category include land cover and land use; nutrient runoff to coastal waters (a measure of loss of an element of the matrix); and soil quality as measured by soil organic matter. Land Cover and Land Use Land cover refers to the ecological status and physical structure of the vegetation on the land surface (e.g., forests, grasslands, wetlands, crop- lands) (Meyer and Turner 1994~. However, land cover depends in part on land use, the way the land is used by people (e.g., a forest managed for timber, a forest used to conserve biological diversity, industrial areas, areas of human settlements) (Meyer and Turner 1994~. Because changes in land use often (but not always) affect land cover, land use is itself an indicator of land cover. In general, land cover can be detected and moni- tored from remotely sensed imagery, but detection and classification of land use usually requires on-the-ground measurements. Often, especially in industrialized countries, information about land use can be obtained from maps and other data sources at local and regional scales, but such compilation is more labor-intensive and expensive than getting informa- tion on land cover. Data on current land cover and trends in those values are essential for the derivation and use of most indicators the committee recommends.

68 ECOLOGICAL INDICATORS FOR THE NATION Therefore, we present the land cover indicator first and recommend that top priority be given to developing this indicator as rapidly as possible. A reliable land use indicator, although more difficult to develop, is also of great importance. The mathematical tools needed to assess land cover patterns and how they change are detailed in Appendix B. To assess this component of human impacts on the land, the commit- tee recommends a land cover indicator to track the amount of land in each of an array of land cover types, such as croplands, forestlands, wetlands, and nature reserves. A large fraction of the Earth's land surface is devoted to agriculture, and so agroecosystems ecological systems that are intensively managed for the production of food or fiber are essential components of any land cover and land use indicators. Agricultural systems are managed for high production, and typically are characterized by intensive nutrient and pesticide inputs, fast growth/harvest cycles, and low plant and animal diversity (Odum 1984, Matson et al. 1997~. Negative effects that may accompany these patterns include increases in soil erosion, groundwater contamination, eutrophication of lakes and rivers, and increased resis- tance of pest and plant pathogens to the chemicals used to kill them (Matson et al. 1997~. Nevertheless, intensively managed agroecosystems usually contain parcels of unmanaged or lightly managed areas, such as woodlots, fencerows, or riparian areas that can act both as refuges for beneficial predators of insect pests (Letourneau 1997) and as reserves for insect pests, weed seeds, plant pathogens, and alternate hosts of fungal pathogens of crops, such as cedar apple rust or crown rust of oats (Schumann 1991~. Because high-value farmland is being lost to commercial, industrial, and residential land uses (USDA 1997), the amount of agroecosystems is a component of the land cover indicator. Both land cover and land use indicators are extremely important. Although this section focuses on land cover, because data are more readily available at present, much of it is relevant to land use, which should also soon become a usable indicator. The land cover indicator can be applied to all environments, includ- ing those in which the "land" is submerged. Therefore, the land cover indicator recommended by the committee, includes rivers, wetlands, and riparian zones as well as dry land. The land cover indicator records the percentage of land in each of a variety of land cover categories. Every time land cover is computed, the proportional representations should be compared with those that existed at the previous recording time. To provide a useful indication of the status of the nation's lands, many cat- egories of land cover types must be recognized and the input data entered and stored separately for each one. Because the proportion of land in

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 69 each land cover category changes relatively slowly, land cover needs to be reported only once every five years, but its values need to be computed annually as inputs to other indicators. Supporting Models and Data Requirements. The conceptual model for land cover is simple. Land cover measures the proportion of the land- scape (and waterscape) occupied by each member of a set of land cover types that must, by definition, add up to the total area of the nation. The major decisions concern the number of land cover categories to recognize, how to account for their spatial configurations, and how to accommodate changes in the number and kinds of categories that are recognized. Change in the Proportional representation of various land cover cate~o- ries is the variable of interest. Visually representing proportional changes for a large number of land cover categories is difficult, but pie or star diagrams may serve that purpose. More sophisticated analyses of changes in land cover categories are also possible through the use of Markovian models (Appendix B). Those models assemble the transition probabilities between categories into matrices. From the matrices the steady-state distribution of categories and the rate of approach to the steady state can be calculated. As a result, one can obtain an indication of what the landscape will eventually look like under selected policies, and how long it will take to reach that state. Data Needs. Current capabilities of satellite imagery are sufficient for identifying a large number of categories of land cover. Very complex classification schemes are possible by combining imagery from several sensors with scenes from different seasons. As pointed out in Chapter 2, such techniques depend either on the Advanced Very High Resolution Radiometer (AVHRR), because of its more frequent sampling, or on a combination of the coarse spatial resolution of AVHRR and the finer reso- lution of Landsat Thermatic Mapper (TM). The forthcoming map by the U.S. Geological Survey (USGS) with a 100 meter resolution, based on Landsat TM data, also may serve as the basis for establishing categories. The condition of the nation's flowing waters is clearly an important element of the land cover indicator. An aggregated measure of flow patterns of the nation's rivers would be useful, but the complexities of river-flow dynamics make such a measure computable and understand- able only at the level of river basins. For the land cover indicator, a simpler measure percent free-flowing (as explained below) captures enough of the pattern to serve as a useful surrogate. Dams, regardless of their purposes, change discharge patterns, in- hibit movement of fishes, and impound sediments and their contained organic carbon, nutrients, and contaminants (Poff et al. 1997~. Water is

70 ECOLOGICAL INDICATORS FOR THE NATION retained in reservoirs to alter the pattern of peak flows and rates of flow during storms. The timing of water release determines the timing and magnitude of seasonal runoff maxima, which often differ from those in free-flowing streams in the region. Therefore, for this reason it will be desirable for the land cover indicator to include a measure of percent free- flowing, or the length of free-flowing parts of streams and rivers divided by their total length. The percentage can be computed for each river basin and then aggregated into a single nationwide value. Percent free-flowing can change in response to policy and manage- ment decisions. Benke (1990) estimated that of the 5,200,000 km of streams in the contiguous 48 states, only 42 streams flowed unimpeded for more than 200 km. This is much less than 1 percent of the length of the nation's rivers. Of these 42 rivers, only six, all of which are in the southeastern United States, flow to the sea. Damming of rivers may continue, at least for hydropower generation, with the greatest number of undeveloped sites in the Pacific, mountain, and northeastern states. On the other hand, sentiment is increasing for removing some dams. Data on the length of large rivers impounded behind dams can be obtained using satellite mea- surements, but information on dams on small rivers can be gathered only by field surveys. However, because the number of dams built or destroyed over a short period is a very small percentage of all dams, the percentage free-flowing would change extremely slowly. Updating the data base would be relatively simple once it had been compiled. As mentioned previously, the U.S. Department of Agriculture's National Resources Inventory (NRI) provides a comprehensive assess- ment of the state and performance of natural and agricultural ecosystems on 800,000 sites on private lands every five years. The NRI is a compre- hensive sampling of land cover, land use, soil erosion, prime farmland, wetlands, and other characteristics on nonfederal lands in the United States. These data show that although soil erosion and agricultural wet- lands loss have decreased, 6 million acres of prime farmland were con- verted to nonagricultural uses between 1982 and 1992. NRI data can form part of a system to confirm ground truth of land use categories identified by satellite imagery. The U.S. Geological Survey (USGS), through its Biological Resources Division, and in collaboration with other federal agencies, is creating a vegetation map for the United States at 1:100,000 scale from Landsat TM data. The USGS is also more than halfway through a major effort to map existing land cover for the United States, at approximately 100 m resolu- tion, also using Landsat TM data. Several national- and continental-scale data sets acquired by federal agencies over the past few years to pro- mote land cover studies for the United States, the humid tropics, and North American boreal forests are now available to the scientific com-

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 71 munity for analysis. At a global scale, the first complete 1 km resolution global land-cover product was released by the International Geosphere Biosphere Program (IGBP) in Summer 1997. A land use indicator will need to distinguish among forms of a gen- eral category of land use depending on the criteria by which they are managed. For example, forest lands should be segregated into categories such as primary forest, unmanaged second-growth forest, forest man- aged primarily for timber production (referred to here as timberland), recently burned forest, forest managed for biodiversity preservation (e.g., forests in Safe Harbor agreements), and forest reserves, as well as into categories such as deciduous forest, boreal forest, and Pacific coast rain- forest. Such categories can be aggregated into fewer categories as needed, but to identify forests in which the human impact is being reduced, is not changing, or is increasing, the disaggregated input data will be needed. As an example, we discuss wetlands in some detail. Categories of aquatic habitats include wetlands, fresh and saline lakes, reservoirs, rivers, and bays. Most difficult to identify and monitor are wetlands water- logged landscapes that are inhabited by distinctive biotas (NRC 1995b). Wetlands cover 26 percent more area in the coterminous United States than all other categories of aquatic habitat combined (Frayer 1991~. Their biotic communities respond chiefly to the influence of hydrology, driven by topography and climate, but also to nutrient supply related to geology and soils. In many locations, because of anaerobic conditions, wetlands accumulate substantial deposits of organic detritus. Wetlands are called peatlands when accumulations of partly decomposed organic matter reach a depth of 30 cm. Many diverse oxidation/reduction reactions mediate elemental fluxes between the atmosphere and wetlands (Mitsch and Gosselink 1993~. Wetlands are of major significance for the cycles of carbon, nitrogen, and sulfur, and peatlands are a reservoir of at least 400 billion tons of carbon worldwide (Woodwell et al.1995~. Significant losses of carbon from that reservoir are projected if the global climate warms (Gorham 1991, 1995a). Therefore, monitoring the extent and status of various categories of wetlands is extremely important. Many physical, chemical, and biological criteria have been used to categorize and monitor wetlands (Adamus and Brandt 1990, NRC 1995b), and a land use indicator will need to recognize a substantial number of wetland types. As in the case of terrestrial land use types, specific catego- ries of wetlands should be separated according to the criteria by which they are managed. A particularly important category is wetlands created as part of mitigation settlements under Section 404 of the Clean Water Act (Kusler and Kentula 1989), because created wetlands seldom fully replace naturally functioning wetlands (Erwin 1991, Gorham 1995b, NRC 1995b, Race and Fonseca 1996~. It is also important to recognize as a separate

72 ECOLOGICAL INDICATORS FOR THE NATION category wetlands restored after years of tile drainage and crop produc- tion under the government's Wetlands Reserve Program, which since 1990 has been administered by the Natural Resource Conservation Service (formerly the Soil Conservation Service) under the Food and Security Act of 1985. Wetland types can be distinguished using data from the National Wetland Inventory (NWI) carried out by the U.S. Fish and Wildlife Service since the mid-1970s, mostly by means of 1:60,000 color-infrared photogra- phy. By late 1991, 70 percent of the coterminous states and 22 percent of Alaska had been mapped (NRC 1995b), the basic mapping units being the set of wetland categories devised by Cowardin et al. (1979~. The NWI has prepared a report on wetland status as of the mid-1980s, and on the trend of losses since the mid-1970s (Dahl and Johnson 1991), based on a repre- sentative sample of U.S. wetlands. Future reports are planned at 10 year intervals. Other NWI products include reports to accompany each 1:100,000 scale wetland map, reports on state wetlands, and a wetland- plant database (Mitsch and Gosselink 1993~. NWI maps are being digi- tized for a computerized geographic information system that will facilitate both analysis and display of the data. In only ten states, however, was the process near completion in 1994 (NRC 1995b). Reliability. Calculating changes in the proportions of land cover in the United States with fully replicable techniques will require integrating remotely sensed information with data from statistically based, in situ sampling programs. A growing number of examples exist in which such information has been collected on regional scales, especially using remote sensing. Of particular interest to the scientific community are studies of tropical deforestation (Skole and Tucker 1993; Laporte et al. 1995; lanetos et al. 1997 [GOFC]), in which changes in proportions of land cover types have been calculated and full transition-probability matrices have been derived. Temporal and Spatial Variability. The land cover indicator itself does not retain the spatial information in the underlying data. However, if the underlying data are archived with all their spatial information intact, then more complex, spatially explicit measures of land cover changes can be computed (Appendix B). These measures are likely to be of substantial regional and local interest. Many natural processes cause changes in land cover, but typically they result in relatively small annual changes. Appreciable changes typi- cally occur only over decades. However, some anthropogenic changes, such as clearing forests for rangeland or farmland, and large fires, happen

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 73 quickly. Therefore, land cover needs to be computed annually but it need not be formally reported more often than once every five years. Statistical Properties. The statistical properties of land cover are clearly a function of the reliability of identification of land cover types from remote sensing and of the statistics of the in-situ sampling program. Because the use of remote sensing to quantify land cover change is an active area of research and application within the scientific community and operational agencies, the statistical properties of land cover should be clarified in the near future. The classification of land cover types in the land cover indicator must be sufficiently comprehensive for all the additional indicators that are derived from it. Classification issues can be addressed in two ways. One would be to determine the number of classes required by the most demanding indicator and use that classification in all input data. This method would force the unification of classification schemes for all indi- cators. The other would be to have a more flexible, hierarchical classifica- tion scheme that would allow the unfolding of additional classes of land cover from a smaller number of aggregated classes from the same under- lying data. The latter approach has been used by the IGBP (IGBP 1992b; Defries and Townshend 1994; Townshend et al. 1994) in its global 1 km land cover product. A flexible, hierarchical classification system would best serve the needs of a variety of environmental indicators. The rate of land cover change determines the most appropriate sam- pling intervals, but there are also practical limitations on sampling fre- quencies. For example, analyses of satellite data are constrained by both the technical features of the satellite itself and by the ability of investiga- tors to handle and interpret the very large volume of data. In addition, the return time for sampling NRI plots or CFI plots is largely determined by the availability of field crews. However, the mean resampling time for CFI plots is approximately 10 years, which compares favorably with theo- retically determined appropriate sampling frequencies for forest produc- tivity (Appendix A). Scientists who use remote sensing have addressed issues of sampling intervals at some length (Skole et al. 1997; IGBP 1992b; Justice and Townshend 1988~. As a rule of thumb, they have suggested that five-year intervals of complete remote-sensing surveys at national, continental, and global scales are generally adequate. Complete surveys could be inter- spersed with annual stratified samples to detect rapid changes without overwhelming the capacities of investigators and systems to perform the analyses. Necessary Skills. The collection of remotely sensed data obviously

74 ECOLOGICAL INDICATORS FOR THE NATION demands a high degree of familiarity and experience with the instru- ments and data-handling capabilities. However, technological barriers to handling remotely sensed data are shrinking as computer technology improves and costs decline. The greatest barriers to developing and using the land cover indicator are probably conceptual: developing the detailed techniques for classifying, combining, and interpreting changes in land cover categories by means of which remote-sensing and in situ data are evaluated. Data Quality Control, Archiving, and Access. The input data for land cover should be archived at the most highly resolved and disaggregated levels, and the techniques used to generate land cover classes need to be described clearly and documented. Only in this way will the land cover indicator be replicable and real changes detectable. Sources of error in both measurements and classifications should also be clearly defined and documented. Comparing maps derived at different times, although fea- sible, is not a good method of documenting changes in land cover because it confounds measurement errors, interpretation errors, and cartographic errors in ways that would be extremely difficult to quantify. It is more straightforward and desirable to detect change in the underlying data themselves, use those differences for quantitative analyses, and then derive maps for presentation purposes. Robustness. Although tremendous strides have been made in recent years, the use of remotely sensed information for ecological analyses is still in its infancy. For many years, the technical challenges of simply handling and processing the data were so large that they inhibited the use of the systems by all but the most sophisticated laboratories. Rapid improvements in cost and performance of computer hardware and soft- ware are removing many of these technical impediments, but other issues remain. The most important ones include the care, maintenance, and accessibility of data archives, and the intercalibration of the remote- sensing instruments themselves. The land cover indicator is likely to be robust to reasonable sources of interference, especially if the original data are archived carefully. How- ever, the time series of measurements can be compromised by technologi- cal changes unless sufficient care is taken to ensure that new instruments are cross-calibrated, and that the calibration of instruments is maintained and monitored carefully over their lifetimes. Achieving calibration precise enough for these quantitative scientific measurements is difficult, but it can be achieved, as the Landsat data record shows. For data sets that last longer than the lifetime of any one instrument, successive instruments must be flown and cross-calibrated for a period of overlap. In this way,

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 75 trends due to degradation of instrument performance and offsets due to changing satellite orbital geometry or new technology can be identified and quantified before additional analyses are performed. International Compatibility. International programs within the IGBP, the Committee of Earth Observing Satellites (CEOS), and the Global Ter- restrial Observing System (GTOS) have made similar recommendations. There is also substantial international activity through programs designed to use large amounts of remotely sensed and other information to track changes in land use experience (IGBP 1992a; IGBP 1995; lanetos et al. 1997; Justice et al. 1993; Justice et al. in press; Kirchoff 1994; Turner et al. 1993~. In principle, the land cover indicator could be derived similarly from any of these other international efforts. It will certainly benefit from all of them. INDICATORS OF ECOLOGICAL CAPITAL: BIOTIC RAW MATERIALS The United States has repeatedly affirmed its commitment to preserv- ing its biological resources (see for example the Endangered Species Act [ESA] of 1973 and its various amendments). Currently, the only account- ing of our success (or failure) is the number of species listed as endan- gered or threatened on the ESA Endangered Species List. This list contains only species for which there is some minimal amount of information and interest, and it is influenced by political and economic factors, as well as by information about biology and the status of populations. For these reasons, the list is not an accurate reflection of the number of species at risk of extinction, or, more generally, of how well the nation's biological resources are faring. The committee recommends two indicators of ecological capital. The first, total species diversity, is a measure of the ecological capital actually present. The second, native species diversity, compares the number of native species an area of land supports with the number it would support in the absence of human impacts. Two other indicators, which measure the original ecological capital that remains after human impacts, and the eco- logical capital borrowed from somewhere else, are proposed in Chapter 5, because they are most useful as indicators at local and regional scales. Gathering the data needed to compute these indicators will require sub- stantial investments of human and financial resources, but a major effort is needed to achieve the nation's commitment to preserving its biological resources. Both indicators depend ideally on a land use indicator, but while that indicator is being developed, these should be developed using the information available, whether via land use or land cover indicators.

76 ECOLOGICAL INDICATORS FOR THE NATION Total Species Diversity The simplest measure of species diversity is species richness, an unweighted list of the species present in any unit of land. Because loss of a species is irreversible, species richness is especially important to moni- tor. Weighted indices of species diversity are valuable for many pur- poses, but because they usually discount rare species, which are often our primary concern, we recommend indicators based on unweighted mea- sures (species richness). The land use indicator, when it is developed, can be used to build an indicator of total species diversity by assigning a score to each category of land use, representing its contribution to preserving species, and then computing the average score for the nation as a whole. That average the total species diversity indicator can be computed by multiplying each score by the number of square kilometers in its land use category, summing scores, and dividing the total by the number of square kilometers in the nation. Until that indicator is developed, land cover should be used in its place. Assigning scores for each land cover category is the difficult problem. The simplest way would be to use the number of species in each land cover category as the score. The total species diversity indicator would then be the average number of species in all land cover categories. How- ever, this method fails to consider differences in the areas covered by each land cover category. The number of species in an area depends on its extent; larger areas have more species than smaller ones (Arrhenius 1921~. For example, if a decision to recognize a new land cover category is made, a former category will need to be split. Because each part is smaller, each will have fewer species than the old category. As a result, the total species diversity indicator would decline without a change in the status of any species. Therefore, any useful indicator must adjust for area differ- ences. Often, the value is normalized by dividing the diversity by the area to obtain a new measure of the "density of species." However, this new measure is inappropriate because diversity does not increase linearly with area. Ecological experience for almost two centuries has shown that, within a continent or biogeographic province, the relationship between land area and total biological diversity fits a power law called the species- area curve: S = cAZ, where S is the number of species, A is area, and c and z are coefficients of the equation (Arrhenius 1921, Preston 1962~. If z were 1, then diversity would be linearly related to area and no adjustment would be necessary. But z is close to 1 only when similar

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 77 biological provinces are compared (such as North America and Eurasia). Typically z is about 0.15 for different areas of a single land cover category in a single continent (Rosenzweig 1995~. That is, an area 10 times larger than another will not have 10 times the species diversity of the smaller, but only 1OZ times as many species. Using the typical z-value of 0.15, 10° ~5 is only 1.4. Dividing S by A would make the larger area appear to have a density only 14 percent as high as that of the smaller area, despite both areas' being samples of the same whole and thus having the same intrin- sic species densities. The power law is neither precise nor entirely accurate (Leitner and Rosenzweig 1997), but it gives sufficiently good fits in a wide variety of circumstances to provide the basis for a quantitatively defensible adjust- ment for area. Di= Si/AiZ, where Di is the adjusted species density of land cover type i. In other words, divide the number of species, S. by a function of area, AZ, because AZ is linearly related to diversity. If a land cover category sampled at two different scales is adjusted in this manner, its species density will be the same. To compute the total species diversity indicator, multiply each D times Pi, the proportion of i in the land cover indicator. The sum of these products would then be an estimate of total species diversity, although this estimate needs to be related to some reference state, namely, the amount of biological diversity expected in that land cover type. The power law can be used for this purpose as well. For example, to determine the expected wildflower diversity in north- western timberland, one first measures the wildflower species-area rela- tionship on northwestern forest reserves (i.e., places set aside as parks and wildernesses that can serve as standards for the natural ecosystem). This yields the reference values of c and z. Second, the number of wild- flower species are measured in an average square kilometer of timber- land, Sour. Substituting 1 km2 for A, and the values of c and z obtained from the forest reserves, the referent or standard Sn for the square kilometer of timberland can then be calculated. Sir can be compared with Sn using a simple ratio, S~mbr/Sn. However, doing so would imply that the more species, the better the state of diversity. Instead, unusually high diversity, that is, values of S~mbr/Sn greater than 1, are likely to presage a decline of diversity. To correct for this problem, the score assigned to the category "timberland" should be based on the absolute value of the dif- ference between it and a forest in reserve. This value can be standardized by dividing it by Sn. For ease of interpretation, the new measure can be scaled by subtracting it from 1:

78 ECOLOGICAL INDICATORS FOR THE NATION Mtmbr—1 - { 1Stmbr— Sn l/Se} This scoring standard has a maximum value a perfect M of 1. Mathematically, M can drop below 0 if a land cover category has more than twice Sn, but it would fall to 0 only in a sterilized environment. Therefore, M can be treated as if it were a true proportion. The assessment can be repeated with a second taxon, such as butterfly species, a third, such as birds, and a fourth such as phytoplankton in lakes. In principle, there is no limit to the number of taxa that can be used. However, for many taxa, lack of data or taxonomic knowledge prevent their use at present. The quickest, cheapest, and most readily interpret- able scores will be achieved if the diversity surveys are limited to a small number of easy-to-survey taxa with high aesthetic or recreational value to Americans. The final score will be the average of the separate scores from the different taxa. In aquatic ecosystems, similar procedures can be used to calculate M. Reference lakes and streams are needed to establish the expected num- bers of species in different taxa, such as phytoplankton, zooplankton, periphyton, insects, and fishes. To compute total species diversity, this process is repeated in each category of terrestrial and aquatic land cover, to calculate a separate score for each. Then each score is multiplied by the proportion of the nation devoted to that particular land cover, Mi x Pi (where Pi is the proportion of i in the land cover indicator). Total species diversity, which is the sum of these products, will be a proportion with values from 0 to 1. It would give the nation an overall view of its biodiversity. The indicator's chief value is in providing a measure of total species richness. It can reflect human impacts, especially severe ones, and it also reflects many other environ- mental variations. Thus it allows one to compare the species richness in various land cover types as well as the effects on species richness of various natural environmental and human-caused changes. Because species counts are the backbone of any ecological survey, lists of species are available for many areas. The areas sampled need not be the same size, because, using the power law, they can all be adjusted to a fixed area. Once appropriate land cover categories are determined for the land cover indicator, the referent standard species-area relationship for each land cover category's natural ecosystem is readily calculated, if large enough samples are available. Data from very small samples fit the power law poorly; they badly underestimate diversity. Underestimates errone- ously reduce the estimate of c and increase the estimate of z, changing Sn. The errors in estimating c and z tend to offset each other, but they do so to

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 79 an extent that depends on the particulars of each case. Hence, it is best to avoid small samples as much as possible. The scale at which M is measured also affects its value. Some native species are rarely and irregularly found in a small proportion of a land cover type. Such species are neither typical nor sustainable under these conditions. They actually depend on their native, natural ecosystems. Nevertheless, if one counts all the native species in the entire area devoted to a land cover category, rare species will be scored. This is a problem because the indicator M depends on simple counts of species, and so it counts a rare species found in only a small number of samples exactly the same as a species with a large, widespread, healthy population in that land cover category. It would be desirable to solve this problem by replacing the number of species, S. with a diversity index that takes abundance into account (e.g., Simpson's index or the Shannon-Weiner index), but not enough is known about how species abundances are distributed in natural and dis- turbed ecosystems, or how diversity indices behave as functions of area to do so. Therefore, replacing S with a weighted index would not improve the indicator. As knowledge of abundance patterns improves, use of weighted indices might become practical. Meanwhile, an appropriate adjustment for rare species is to estimate M using many subsamples of the land cover category. Each subsample should be large enough to have a good proportion of the native species, but small enough so that rare and unstable populations are recorded in only a few of the subsamples. The few subsamples that have such rare populations will be appropriately diluted in the average M of the set of subsamples. When calculating M, land cover types should not be aggregated into a single category. To see why, consider streams and their fishes. Ignoring the effect of aridity on the area of freshwater ecosystems, there is a slightly negative (but nonsignificant) species-area curve among the native species found in each state. This relationship occurs because big states tend to be arid or, in the case of Alaska, cold and with low ecosystem productivity. Because arid areas have fewer and smaller streams, the area of a state is a poor measure of the extent of its fresh waters. Our expectation of fish diversity, our referent state, needs to be computed from subsamples whose spatial variation in aridity and productivity fall within narrow enough bounds to keep them in the same climatic zone. If the effects of human activity on the environment continue to increase, many species are likely to become extinct. If the referent standard values of Sn are adjusted to that new situation, a steady decline in Sn and a concomitant steady but fallacious rise in total species diversity would result. Recalculating Sn would be an unfortunate example of the hazards of the shifting baseline phenomenon (Pauly 1995~. Sn needs to stand as a

80 ECOLOGICAL INDICATORS FOR THE NATION fixed standard for each sort of broadly defined ecosystem. Once mea- sured well, it should not be changed even if diversity declines in pro- tected ecosystems. Only if the standard is maintained will the United States have a sound measure of how well it is doing in maintaining its biological capital. Native Species Diversity This indicator reflects human impact on the land. Land that has been so transformed by people that it cannot support most of the native species that would otherwise be there is land that carries a heavy burden caused by human activities. In contrast, the human impact on lands that still support a diverse assemblage of native species is light. We call our indicator the Native Species Diversity indicator. The total species diversity indicator counts all the species of a taxon native and exotic. The native species diversity indicator covers only natives, because its purpose is to measure human impacts. If humans cause a native species to be replaced by an exotic, native species diversity counts that as an impact; total species diversity does not. Human population pressure is the basic cause of reductions in envi- ronmental quality, including loss of native species, but population size is not the sole determinant of those losses. Many damaging practices, such as release of toxic materials and overexploitation of renewable resources, have accompanied even low-density human populations. Therefore, human population size itself is not a good environmental indicator, even though environmental threats would be fewer if there were fewer people. The human population has environmental impacts by converting natural ecosystems into places to live and work, although dwelling places, shopping centers, office buildings, factories, and public utilities constitute only a small part of the impact. In the United States, only 35 percent of the total land area impervious to water is covered by places for people to live and work; the other 65 percent is related to transportation, primarily roads and parking spaces for cars. The total area of the United States that is impervious (e.g., roofs, concrete, and asphalt surfaces) is still a small percentage of the total land area, but it has doubled since World War II. In general, when the fraction of impervious area increases above 10-20 percent of the total land in a given area (e.g., a watershed), hydrologic flow patterns change markedly from natural conditions and diffuse source water pollution problems tend to occur. In addition, much larger areas are used to grow the food, medicines, and construction materials that sustain health and well-being. These materials are also transported over distances that would have amazed people a century ago. For example, Folke et al. (1996) estimated that the 29 cities of the Baltic region appropri-

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 81 ate natural resources from ecosystems that cover 200 times the total area of the cities themselves. Several statistics illustrate cogently the need for national assessment of our impact on the land. First, less than 1 percent of U.S. grassland remains in anything like its natural state. Second, only 4.7 percent of the nation's forest land is unmanaged, even though most formerly forested land still has trees growing on it. Third, human activities, primarily agri- cultural drainage, resulted in the loss by 1970 of more than half of the wetlands present in the United States at the time of European settlement (NRC 1995b). In some states, for example, Ohio and Iowa, the loss exceeds 95 percent. Substantial losses have continued since then. Not all conversions of land to our own direct use exterminate the species that live there, although few wetlands species can survive drain- age of their habitat. The magnitude of the effect depends on the wisdom exercised in managing lands. Therefore, indicators are needed that mea- sure the amount of land converted to different purposes and how well the converted land remains productive and supports biological diversity. Description of the Indicator. As in total species diversity, species-area power laws supply the standard to evaluate the human impact. Native species diversity uses the same power-law relationships that were mea- sured in natural reserves for total species diversity (Sn = cAZ), but thereaf- ter its construction differs. To understand the difference, consider again the example of scoring northwestern timberland on the basis of its native wildflower diversity. To calculate total species diversity values of c and z, and obtain the natural number of species, Sn, for a square kilometer, all species were counted. To calculate native species diversity, exotic species are excluded from S~mbr. The remainder is the native wildflower diversity, Sn~mbr. The score for human impact is the proportion of that standard achieved in the timberland: Gtmbr Sn,tmbr/Sn Because Sn`~mbr cannot exceed S~mbr, G~mbr is a proportion, with values ranging from 0 to 1. On pristine land, G = 1; a land cover category with no native species receives a score of 0. People manage forestlands and other lands with a variety of strate- gies, some designed to protect the environment, others not. To incorpo- rate management variation into its value, G is measured separately for lands managed by different strategies and then multiplied by the propor- tion of the land type subject to that strategy. The sum of these products over all strategies is the score for the land category. It, too, is a propor- tion, with values ranging from 0 to 1.

82 ECOLOGICAL INDICATORS FOR THE NATION To illustrate the calculation of G. consider Cody's (1975) determination of the species-area power relationship for birds in California chaparral. The natural power relationship is Sn = 45A° 125, To evaluate G for residential areas around a city, breeding bird lists are obtained and the average number of native species (Si) enumerated for a set of square kilometer samples of residential land in what would otherwise have been chaparral. The power law tells us to expect 45 species in each square kilometer, so G = Si/45. Because of the species-area power law, one is not restricted to areas of 1 km2; data from a variety of sample areas can be used. (For an example from Tucson, Arizona, see the discus- sion of biodiversity indicators in Chapter 5.) A similar measure commonly used as an indicator of aquatic eco- system condition is based on concepts used by Karr et al. (1986) in devel- oping the Index of Biotic Integrity (IBI). The IBI is an additive index that has been locally calibrated; users need to have only sufficient knowledge to be able to identify local species. Multifactorial indicators such as IBI have been developed around a set of measures of the distributions and relative abundances of selected taxa. Each factor is assigned a numerical value (an integer between O and 6) based on the professional judgment of the evaluator. The assignment of the appropriate integer value is based on the distribution of the data for each factor from a number of reference sites. The final indicator is calculated as the sum of the individual factor scores (usually 10 to 12), which typically generates a score between 0 and 60. Indicator developers then set management goals for a particular water body based on a predetermined ranking score for the indicator. Although IBIs have been helpful in the development and evaluation of management policies in many regions of the country, the subjective nature of the judgments in assigning values to each factor, as well as problems associated with calibrating multifactorial indices (Reynoldson et al. 1997), lead the committee to recommend native species diversity rather than IBI as the best national-level indicator of human impact on species diversity. However, IBIs have considerable value at local and regional scales and we encourage their continued use. Changes observed in native species diversity come from several sources. Native species diversity decreases if more land is shifted to human use and if human use is intensified without regard to the ecologi- cal consequences. Native species diversity increases if management strat- egies on some lands are changed from those with heavy impacts to those with lighter impacts, if management strategies improve, and if land shifts to a less ecologically damaging use.

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 83 An important case of an improved land use strategy is the Safe Harbor program, an example of cooperation between private landowners, a non- governmental conservation organization (the Environmental Defense Fund), and the federal government. In return for landowners' managing their land to encourage its colonization by endangered species, land- owners become exempt from the restrictions that the U.S. Fish and Wild- life Service would otherwise place on their property were an endangered species to move in on its own. Land in the Safe Harbor program should be separated from other similar land into a special land use category in the land cover indicator. It would have an entry in the land cover indicator and special measure- ments would be made of its contribution to total species diversity and native species diversity. Native species diversity provides an indication of the success of conservation over the entire slate of land uses and man- agement strategies. Although lack of adequate information on many taxa will make devel- oping these indicators of species diversity difficult, the work should be started now. Doing so will provide an incentive to learn about taxa that are not well known at present. There is enough information to be useful already and the rate at which our environment is changing makes it urgent to implement such indicators. INDICATORS OF ECOLOGICAL CAPITAL: ABIOTIC RAW MATERIALS Nutrient Runoff Water quality and ecological conditions in U.S. coastal waters have been subjects of growing national concern for more than two decades. Several major types of pollutants oxygen-demanding organic matter, microbial pathogens, heavy metals, synthetic organic compounds that bioaccumulate to potentially toxic levels, and excessive levels of nutri- ents all have taken their toll on these ecosystems. Massive efforts sparked by the Clean Water Act to clean up municipal and industrial sewage effluents have yielded substantial reductions in point-source pollution and improved water quality near such sources. Efforts to con- trol potentially toxic heavy metals and synthetic organic chemicals have resulted in major declines in loadings of these substances to coastal waters, but the legacy of many years of inadequate controls is still seen in the polluted sediments and high body burdens of these substances in the marine organisms of many coastal areas. Among the major potential pollutants affecting coastal environments, nonpoint sources of nutrients- N and P. in particular have received relatively little regulatory attention.

84 ECOLOGICAL INDICATORS FOR THE NATION The results of excessive nutrient loadings are seen in reduced water clarity; nuisance algal blooms, including species with toxic forms like Pfiesteria piscicida and Gymnodinium spp. (red tide); and hypoxic (low oxygen) bottom waters. Outbreaks of toxic and other algae have been correlated with nutrient enrichments and appear to be increasing in estuarine and coastal waters (Burkholder 1998~. Hypoxia, generally defined as persis- tent oxygen concentrations of less than 2 mg/L, affects areas of Long Island Sound, Chesapeake Bay, the near-shore Gulf of Mexico near the mouth of the Mississippi (e.g., Rabalais et al.1996), and many other coastal areas around the world. In the Gulf of Mexico, the affected area has grown from about 9,000 km2 in 1985 to approximately 18,000 km2 at present. A large increase in the affected area occurred in 1993, apparently in response to the large spring flood, which brought a corresponding increase in nutrients into the Gulf from the Mississippi River. Current evidence links hypoxia primarily to increased inputs of nitrogen forms (mainly nitrate) to coastal waters, inputs that stimulate algal growth. Nitrogen generally limits plant growth in coastal and ocean waters, whereas phosphorus is usually the limiting nutrient in fresh waters. Although growing algae produce oxygen in surface waters, decomposi- tion of dead algae in bottom waters consumes oxygen, leading to loss of habitat for fish and other forms of aquatic life. Human alterations of the biogeochemical cycles of major nutrient elements have reached global proportions. For example, human contri- butions to the cycling of nitrogen forms equal the contributions from all natural processes (Ayres et al. 1994~. Although nearly all human addi- tions to nutrient cycles occur in terrestrial ecosystems, these systems are leaky, which is why nutrient loadings to coastal zones are elevated sub- stantially above background levels continent-wide. Quantitative data on these increases are sparse, however, and insufficient to document tempo- ral trends in nutrient losses from the continental United States to coastal and marine systems. Evidence is also lacking on the geographic extent of impacts of elevated nutrient loading on the open oceans, although it is generally assumed that impacts of human-induced nutrient inputs on the productivity of the oceans as a whole are still negligible. Because of widespread concern about the impacts of nutrient load- ings on coastal waters, and the lack of quantitative information to develop related national policies, the committee recommends the development of national- and regional-scale indicators for N and P runoff from the land to coastal areas. Data are already being collected to produce such statistics. The USGS monitors flow rates of major rivers, and state and federal agencies gather water-quality data routinely at stations near the mouths of most rivers. Obtaining accurate loading estimates from routine monitoring data is

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 85 difficult. Flow rates and nutrient concentrations vary substantially in space and time; measurements of flow and nutrient concentrations are often not made simultaneously; concentrations are usually measured infrequently relative to the time scale of their variability; and sampling networks have spatial biases because of the need to target sampling toward specific pollution sources (Smith et al. 1997~. However, models can be used to estimate nutrient fluxes from such data (e.g., Smith et al. 1997~. Current data appear adequate to characterize the yields of nutri- ents from the major hydrological cataloging units in the country (Figure 4.1) and to estimate total runoff from the conterminous United States to coastal waters. An even larger effort has estimated N and P fluxes to the North Atlantic Ocean from rivers in 14 regions of North and South America, Europe, and Africa (Howarth et al. 1996~. According to these authors, nonpoint sources of nitrogen dominate riverine fluxes to coastal waters in all regions. On an areal basis, the largest N fluxes are from watersheds in northwestern Europe and the northeastern United States (Figure 4.2~; but on a mass basis, the Mississippi River drainage basin is by far the largest FIGURE 4.1 Classification of predicted local total nitrogen yield in hydrologic cataloging units of the conterminous United States. Local yield refers to trans- port per unit area at the outflow of the unit due to nitrogen sources within the unit, independent of upstream sources. Source: Smith et al. 1997. Reprinted with permission from the American Geophysical Union.

86 ECOLOGICAL INDICATORS FOR THE NATION ~3 ~476 ~5 <v 3: 2 _ ~ FIGURE 4.2 Total nitrogen (TN) runoff in rivers in kg N km~2 yr -I (top) and total phosphorus (TP) runoff in rivers in kg P km-2 yr -I (bottom). See text for expla- nation of estimates. Source: Howarth et al. 1997. Reprinted with permission from Kluwer Academic Publishers.

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 87 single contributor of both N and P to the Atlantic Ocean from North America (1.82 of 4.61 teragrams [10~2 g] per year [Tg/year] for N. and 0.107 out of 0.241 Tg/year for P). A strong linear correlation exists between river fluxes of total N and the sum of anthropogenic N inputs to temperate regions, but on average, regional fluxes in rivers are only about 25 percent of the computed anthropogenic inputs. Based on data from relatively pristine areas, Howarth et al. (1996) estimate that riverine nitro- gen fluxes in many temperate regions have increased 2- to 20-fold from preindustrial times. Like the previous study, this study did not produce runoff-rate statistics for any particular year. Instead, both used informa- tion gathered over multiple years to produce annualized runoff rates gen- erally representative of "current" conditions. The committee's recommended indicator of total runoff of N and P can take values ranging from 0 (no discharge) to thousands of kg km-2 yin, with lower values being more desirable for most purposes. The status and trends in the total runoff of N and P should be computed and reported annually (e.g., in Tg/year). Annual rates for major river systems such as the Mississippi River should be reported together with aggregate num- bers. Because of the strong influence of short-term weather variations (drought or flood conditions) on regional runoff rates, the annual output runoff data (both national and regional) should also be normalized by dividing the runoff rate (Tg/yr) by the ratio of actual rainfall over the region to the long-term average rainfall for the region. Normalized data provide insights into the sensitivity of river basins to short-term weather variability. On a time scale of a few decades (or less), the national indicators of N and P runoff to coastal waters can be evaluated and used in determining the effects of national policies to control point and diffuse sources of nutrient pollution on the net loss of nutrients from the country. On regional scales, the statistics can play similar roles in evaluating the effec- tiveness of specific regional management practices. Comparison of trends in runoff with trends in fertilizer use, crop production, and coastal water quality can give further insights into the need for changes in policies, regulations, and management practices. Soil Organic Matter Soils promote the growth of vegetation, including crops; control the flow paths of precipitation as it becomes surface and groundwater; and serve as a filter for potentially harmful substances that would otherwise enter this water and the atmosphere (Larson and Pierce 1991, Parr et al. 1992, Johnson and Lindberg 1992, NRC 1993~. The ability of soils to perform those functions depends on their physical, biological, and chemi-

88 ECOLOGICAL INDICATORS FOR THE NATION TABLE 4.2 Reference and Measured Values of Minimum Data Set for a Hypothetical Typical Soil Type (Hapludoll) from North-Central United States Horizon and Characteristic Reference Value Measured Value Surface Horizon Phosphorus (mg/kg) 30 15 Potassium (mg/kg) 300 300 Organic carbon (percent) Total 2 1.5 Labile 0.2 0.15 Bulk density (mg/m3) 1.3 1.5 pH 6.0 5.5 Electrical conductivity (S/m) 0.10 1.0 Texture (percent clay) 30 32 Subsoil horizon Texture (percent clay) 35 35 Depth of root zone (m) 1.0 0.95 Bulk density (mg/m3) 1.5 1.5 pH 5.5 5.5 Electrical conductivity (S/m) 0.10 0.10 Source: NRC 1993. cat properties, which are easily measurable and whose relationships to these functions are well known. The NRC (1993) reviewed the potential for soil properties to serve as indicators of the soil's ability to promote plant growth, regulate water flow, and filter or retain chemicals of con- cern. Its summary of a minimum data set is given in Table 4.2. The NRC study (1993) identified soil organic matter (SOM) as "perhaps the single most important indicator of soil quality and productivity," and work by Bauer and Black (1994) and Reeves (1997) also supports the use of that indicator. Therefore, the committee recommends soil organic matter con- tent as the best currently available indicator of the state of soil quality. SOM is an indicator of ecological condition (soil condition, relationship to erosion) and of ecological functioning (soil productivity). The Indicator. Soil organic matter strongly influences several biologi- cal, physical, and chemical characteristics of soils. SOM is a nutrient and energy source for soil biota; it improves soil structure by strengthening soil aggregates, increases water retention and available water capacity, reduces the sealing of soil surfaces thereby promoting infiltration and reducing erosion, increases cation exchange capacity, chelates metals, and influences the fate of pesticides. SOM responds to tillage and fertilization

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 89 and it is sensitive to erosion, as the uppermost portion of the soil normally has the highest organic matter concentration. Soils high in organic matter are generally productive, and the carbon content of soils (approximately half of the SOM weight) is an important component of national and global carbon budgets. Perhaps the biggest disadvantage of SOM as an indica- tor is that it is likely to change slowly, and real trends in SOM are difficult to detect because SOM values are highly variable at both local and regional scales. Numerical Values of the Indicator. The five major factors that affect soil formation climate, organisms, topographic relief, parent material, and time control the natural distribution of SOM. Regional differences in the amount and distribution of SOM with soil depth are pronounced, as are differences in those constituents across landscapes with strong gradi- ents in soil-drainage class (e.g., depth of the water table). Concentrations of SOM in agricultural soils generally range from 1 to 10 percent in sur- face horizons, to less than 1 percent at greater depths. Long-term tillage of agricultural soils has reduced SOM content (lenny 1941, Paustian et al. 1997), probably as the result of erosion, increased soil temperature, and reduced organic matter input. SOM can recover or be maintained through careful management (NRC 1993, Reeves 1997~. Data Requirements. Data are not currently available to compute wide- spread, accurate baseline estimates of SOM. Therefore, for the moment, the SOM content of a system should be expressed as a net or percent change over time. Three methods are used to express SOM quantities: gravimetric, volumetric, and equivalent mass. The gravimetric basis, the ratio of SOM mass to soil mass, does not account for soil density. The volumetric basis, the ratio of SOM mass to soil volume, does not account for the total amount of soil present. The equivalent mass basis, the mass of SOM in a standard mass of soil, corrects the volumetric basis to account for total soil mass. The volumetric and equivalent mass bases require soil density estimates. Soil density, particularly of surface soils, varies over short periods, but equivalent mass measures are not sensitive to soil- density changes. Therefore, at this time, the SOM content of a soil is best expressed on an equivalent mass basis. SOM data can be obtained from the soil samples routinely sent to agricultural soil-testing laboratories using well-established, inexpensive measurements or through a specially devised random sampling program. Protocols for SOM measurements exist (ASA Methods Manual), and the measurements are easy to make. Currently, samples taken typically include only the top 20 cm, or the area subject to plowing, the zone most likely to respond to management practices. With bulk density samples of

So ECOLOGICAL INDICATORS FOR THE NATION the top 20 cm, SOM mass per unit area can be calculated to a 20 cm depth. Nevertheless, in many soils, such as conventionally tilled soils, there are significant amounts of SOM below 20 cm. In a program with sufficient personnel, SOM mass per unit area determined to a depth of 50 cm would be preferable. INDICATORS OF THE PERFORMANCE OF THE NATION'S ECOSYSTEMS The performance of the nation's ecosystems, or their ecological func- tioning, can be measured by their productivity, the rate at which they use (mainly solar) energy to fix atmospheric carbon dioxide. The carbon economy of ecosystems can measure the productivity of the nation's eco- systems and estimate their overall carbon budget, that is, whether the ecosystems are losing or accumulating carbon. The committee recom- mends four indicators of this carbon economy and a related one, dis- solved oxygen, that is also related to productivity. The first of these indicators related to productivity, total chlorophyll per unit area (g/m2), provides a direct measure of the production capacity of terrestrial eco- systems (the equivalent measure in lakes is chlorophyll per unit volume). The more chlorophyll per unit area (or volume) in an ecosystem, the greater its capacity to capture sunlight. The second indicator, net primary production (NPP) (g carbon/m2/year; g/m3/year in aquatic systems), is a direct measure of the amount of energy and carbon brought into an eco- system. It is also a measure of productivity in the sense commonly used in agriculture and forestry: the amount of plant material produced in an area per year. The third indicator, carbon storage is measured by net eco- system production (NEP), and is a direct measure of the amount of carbon sequestered or released by ecosystems per year. Carbon storage is par- ticularly important in view of recent concerns about greenhouse gas emis- sions, because the total amount of carbon in the form of CO2 emitted by a region equals the region's fossil fuel emissions minus its NEP. The fourth indicator, trophic status, characterizes primary production in lakes. The trophic status indicator is derived by combining measures of Secchi-disk transparency, total phosphorus, and chlorophyll a concentrations (Carlson 1977~. The committee recommends use of dissolved oxygen as an indicator of the performance of flowing-water ecosystems. In addition to the above five indicators that are directly related to productivity, soil organic matter and land use, discussed elsewhere in this chapter, are also related to ecosystem functioning. Finally, the com- mittee recommends indicators of agricultural nutrient-use efficiency and overall nutrient balance as important indicators of ecological functioning.

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS Indicators of Terrestrial Productivity 91 Except for a negligible contribution from chemoautotrophs, all energy that flows through ecosystems is supplied by green plants during photo- synthesis. Energy in the form of light is captured by chlorophyll and converted to chemical energy in the form of organic carbon. Therefore, the amount of chlorophyll present determines the fundamental capacity of an ecosystem to capture energy, its production capacity. Although soil water, nutrients, and many other factors may limit the rate of energy capture, plants regulate their chlorophyll concentrations in response to these other limits (Hurtt and Armstrong 1996, and references they cited). Thus, the abundance of chlorophyll is an excellent indicator, because it is strongly correlated with an ecosystem's actual capacity to capture energy, not just its potential capacity. Terrestrial plants obtain carbon exclusively from atmospheric CO2. The energy captured by an ecosystem in units of carbon gained is referred to as gross primary production (GPP). After capture by chlorophyll, energy is stored in organic molecules. Plant respiration consumes about 40 per- cent of GPP worldwide (Schlesinger 1997), most of it respiration by leaves, some by fine roots, and still less by stems. Roughly 25 percent of the energy gained is used to construct plant tissues (construction respiration, Foley et al. 1996~. When all sources of plant respiration are subtracted from GPP, the remainder is net primary production (NPP). Because NPP, the amount of organic carbon actually made available to other organisms in an ecosystem, fuels all ecosystem functioning, it is a useful indicator of the carbon budgets of ecosystems. In many natural terrestrial ecosystems of North America, only 4 to 10 percent of NPP is typically consumed by animals (Whittaker and Likens 1973~. Animals consume a much larger fraction of NPP in rangelands, grasslands, savannas, and in freshwater and marine ecosystems (up to 67 percent), environments in which primary production is used to synthe- size easily digested tissues, than in forested ecosystems, where much NPP is allocated to the production of wood. NPP that is not consumed by animals eventually enters the soil or aquatic sediments as undecomposed organic matter, where it is metabolized by detritivores and returned to the atmosphere as CO2. The rate of decomposition is affected by many biological and physical factors, including temperature, pH, soil moisture, nitrogen availability, and the lignin content of the detritus (Parson et al. 1988~. The difference between the sum of all nonplant respiration in an eco- system (all the CO2 carbon produced by detritivores and animals) and NPP is carbon storage or net ecosystem production (NEP). NEP is the change in the total amount of carbon in an ecosystem. If NPP is less than

92 ECOLOGICAL INDICATORS FOR THE NATION total nonplant respiration, NEP is negative and the ecosystem loses car- bon, most likely to the atmosphere as the greenhouse gases CO2 or CH4. In a balanced ecosystem, NEP is equal to zero. If NPP is greater than the sum of nonplant respiration, NEP is positive, and the ecosystem gains carbon. Over the long term in most ecosystems, NEP is zero; otherwise organic matter would disappear (NEP < 0) or massively accumulate (NEP > 0), as it does in peatlands. Supporting Ecosystem Models. The models of ecosystem energetics and carbon economy that support the use of chlorophyll a as an indicator are among the most mature and highly developed in ecology. This is one of the few areas in ecology where mechanistic predictive models already exist. Predictive phenomenological models of NPP were first developed in the 1960s and 1970s. For example, Leith's (1972) Miami Model showed that simple regressions of NPP against precipitation and temperature could predict most of the global variance of NPP. Phenomenological models continue to be important today because imagery collected for the entire Earth since 1972 by Landsat satellites permits the calculation of the Normalized Difference Vegetation Index (NDVI), which, as pointed out in Chapter 2, is a measure of chlorophyll per unit area. NPP is approxi- mately proportional to NDVI, but a different proportionality constant is necessary for each vegetation type (Dai and Fung 1993~. Mechanistic models of NPP have also been developed by physiologi- cal ecologists during the past two decades. Farquhar et al. (1980) devel- oped a model in which the rate of photosynthesis is the minimum of two different reactions light-limited energy capture and carbon-limited CO2 fixation. This model subsequently was extended and refined to include C4 and C3 photosynthesis (e.g., Collatz et al. 1992~. It predicts the rate of carbon gain (moles/square meter of leaf area/unit time) as a function of leaf temperature, light level, and the internal concentration of CO2. The model of Ball et al. (1986) predicts stomata! conductance as a function of net photosynthetic rates, CO2 levels, and the humidity gradient across the plants' stomates. When coupled with simple physical equations govern- ing the energy balance of leaves and the diffusion of gases through stomates, one can solve the equations directly for the photosynthetic rate and rate of transpiration as a function of temperature, humidity, and light level (see Foley et al. 1996, for a clear example). To predict NPP, also needed are a model of plant respiration, a model of nutrient and soil moisture limitation, and a physical description or model of the vegetation (e.g., number of leaf layers and allocation). Models that predict NPP mechanistically are described by Sellers et al. (1997), Haxeltine and Prentice (1996), Raich et al. (1991), and Neilson (1995~. To predict nutrient limitation, it is necessary to be able to predict

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 93 decomposition of organic matter. The most general model of decomposi- tion is the CENTURY model of Parton et al. (1988~. Because mechanistic models of NPP predict decomposition, they necessarily also predict NEP. Mechanistic models of NPP and NEP have been compared in a formal exercise (VEMAP 1995~. They are as accurate as phenomenological models in current climates, but have the advantage that they predict NPP under novel conditions with some theoretical justification. However, when forced with novel climates, the models in the VEMAP comparison diverged from one another. This finding indicates that predictions for novel conditions need to be interpreted with caution and that the mecha- nistic understanding underpinning the models is incomplete in important ways. Methods of estimating NEP account well for above-ground carbon storage, but they do not account for storage of carbon in soils. Therefore, data on soil organic matter need to be added to the outputs of these models. Numerical Characteristics of the Indicators. Chlorophyll per unit area ranges from 2.8 g/m2 in tropical forests and 2.6 g/m2 in temperate forests to 0.5 g/m2 in tundra and desert ecosystems. Values of NPP in natural ecosystems range from 1,400 g/m2/year in tropical rain forests (half that amount in temperate forests) to 50 g/m2/year in American deserts (Whittaker and Likens 1973~. NPP is markedly affected by land use, with values as high as 6,000 g/m2/year in some agricultural systems and as low as zero in urban centers (VEMAP 1995~. NEP for North America is thought to be currently 0.3 x 10~5 g/year (IPCC 1995~. It is not zero primarily because of the regrowth of forests cleared in the last century and early this century. Regrowing temperate forests typically have NEPs of 200-500 g/m2/year (Wofsy et al. 1993~. In contrast, tropical forests are thought to have a negative NEP currently because of deforestation (a total of roughly-l.O x 10~5 g/year [IPCC 1995~. To obtain the aggregated national-level annual indicators of produc- tion capacity (total chlorophyll), NPP, and carbon storage, the values com- puted for each land cover category are summed. Annual changes in these numbers, rather than their absolute values, are the variables of interest because they reveal whether the productivity of the nation's ecosystems is being maintained and whether the total amount of carbon being stored . . . . . IS 1ncreasmg or aecreasmg. Temporal Variability. Three types of temporal variability must be taken into account in developing the NPP and carbon storage indicators: 1. Daily Variation. NPP and NEP vary daily because of variation in light and, to a much lesser extent, because of variation in temperature and

94 ECOLOGICAL INDICATORS FOR THE NATION humidity. Obviously, NPP and NEP are negative at night and positive during the day. Chlorophyll density is stable diurnally. 2. Seasonal Variation. NPP, NEP, and chlorophyll density show dramatic seasonal cycles, with high values during summer and rainy seasons and low values during winter and dry seasons, when many plants shed their leaves. For calculating these indicators vearlv high values are the appropriate numbers. , ., ., ~ 3. Interannual Variation. All three indicators vary dramatically interannually. Natural disturbance or land use change obviously alters NEP, NPP, and chlorophyll density. For example, after a blow-down, a patch of forest has negative NEP for more than a decade, followed by a long period of positive NEP. NPP and chlorophyll density typically recover more rapidly than NEP (Schlesinger 1997~. Even without distur- bance or land use change, interannual variation in weather and climate can have dramatic impacts on these indicators. For example, Earth's terrestrial NEP currently fluctuates by at least 3 x 10~5 g/year from one year to the next (Sarmiento et al. 1995), partly in response to E1 Nino events. NDVI measurements show that, over the past decade, the length of the growing season (period of high chlorophyll density) has increased in the North Temperate Zone by 10 days (Myneni et al. 1997~. Spatial Variability. Three types of spatial variability must also be considered in developing NPP and NEP: 1. Small Scale (individual plant to stand). The most important cause of small-scale variation in NPP, NEP, and chlorophyll density is disturbance. An individual tree fall creates a patch of low chlorophyll density and negative NEP, which is subsequently converted to a patch of high chlorophyll density and positive NEP. Thus, stand-level indicators represent the average of a wide range of values at smaller scales. 2. Medium Scale (stand to region). Within the same climatic region, the primary sources of spatial variation are changes in land use and variation in topography and soils. Each of these may cause stand-level variation in NEP, NPP, and chlorophyll density as large as those caused by disturbances or changes in climate. 3. Large Scale (region to globe). Large-scale differences are driven by climate and associated land use and disturbance. Regional changes in chlorophyll, NPP, and NEP can be computed whenever it is desirable to do so, provided that the data are archived in a disaggregated form. Data Requirements. NOAA and NASA satellites routinely acquire the imagery necessary to evaluate chlorophyll density at scales as small as 100 m2 or less (see Chapter 2~. Simple phenomenological models and the

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 95 CASA model (see below) can convert these images to estimates of NPP and NEP. In addition, four national programs are already in place that acquire the field data needed to estimate NPP and NEP by inventory methods. The U.S. Forest Service's Forest Inventory and Analysis (FIA) pro- gram remeasures and censuses trees on more than 13,000 small plots (0.1 to 1 ha) every 10 years (see Chapter 2, Ground-Based Measurements). These data can be used to calculate forest NPP, but the locations of the plots and the data are difficult to obtain. If this program were augmented to include measures of soil carbon, it would yield direct estimates of forest NEP. The Forest Service's Forest Health Monitoring Program collects infor- mation on forested ecosystem condition and production on a national network of 4,000 1-ha plots. The measurements again provide useful estimates of NPP, but because there are no measurements of below- ground carbon, these data cannot provide direct estimates of NEP. How- ever, the stand-structure data can be used to constrain predictive forest- gap models such as LINKAGES (Post and Pastor 1996) or the coming generation of physiologically grounded models (Hurtt et al. 1998) to pro- vide estimates of NEP. The USDA's National Resources Inventory provides a comprehen- sive assessment of the state and performance of natural and agricultural ecosystems on private lands every five years. Inventory measurements from 800,000 sites and regression models are used to estimate productiv- ity, carbon storage, biomass, land use, vegetation cover, species ranges, and characteristics of soils. Finally, the Long Term Ecological Research (LTER) network of 21 sites provides in-depth information on NPP, NEP, nutrient cycling, organic-matter dynamics, and disturbance (see Chapter 2~. The spatial coverage of this network is obviously limited, however, representing only 19 sites in the United States and Puerto Rico. In addition to sites providing direct inventories of carbon, the new Ameriflux network provides measurements of NEP and NPP using the new technique of eddy correlation. Each of the 24 sites in the Ameriflux network contains an eddy-correlation tower with a vertical array of sonic anemometers and CO2 sensors that continuously measure the vertical gradient of CO2 and the rate of vertical air flow. These data allow NEP to be computed from the amount of CO2 that is removed from or added to the air by the ecosystem. One note of caution is that, for the two sites for which both eddy-correlation and inventory estimates of NEP have been calculated (Oak Ridge and Harvard Forest), the two methods differed by 50 to 100 percent (S. Wofsy, Harvard University, and W. Post, Oak Ridge National Laboratory, personal communications 1998~.

96 ECOLOGICAL INDICATORS FOR THE NATION At least two kinds of additional data are likely to significantly enhance our ability to estimate NPP and NEP in the near future. First, high- resolution satellites and the new canopy VCL satellite (see Chapter 2) will provide powerful additional data. Second, aircraft measurements of the vertical profiles of CO2 in the atmosphere above an array of locations will allow NEP to be estimated by tracer-transport inversion. This technology represents a large-scale analog of the method used in eddy-correlation towers. If wind speed and direction across the parcel are known from weather data, and the amount of fossil fuel consumed in the region is also known, then the CO2 gradient in the atmosphere over the parcel can be estimated, assuming a balanced ecosystem (NEP = 0~. If the actual gradi- ent is less steep than the expected gradient, then the ecosystem within the parcel is consuming CO2, and NEP is positive. Tracer-transport inversion currently allows resolution only at continental scales, but planned arrays of measurements will allow tracer-transport inversion estimates of NEP for separate regions within the United States. Two other groups of models can provide useful estimates of eco- system performance. First, gap models, which were first developed in the 1980s (Shugart 1984 and the references it cites), are now being constructed with the same mechanistic underpinnings as the NPP-NEP models (Hurtt et al. 1998~. These new models should predict population dynamics with the accuracy of gap models, and ecosystem functioning with the accuracy of mechanistic NPP-NEP models. Additionally, Potter et al. (1993) have developed a model that is intermediate between mechanistic and satellite models of NPP-NEP. This model, called CASA after the participating institutions (Carnegie, Stanford, and NASA-Ames), uses satellite NDVI to constrain estimates of production. However, it also includes mechanistic models of respiration, allocation, and decomposition that allow the com- putation of NEP. Although the CASA model cannot predict the future (because it requires temporal sequences of satellite imagery as a driving variable), its mix of mechanism and reliance on data make it ideal for relating chlorophyll density, NPP, and NEP. An Indicator of Aquatic Productivity Trophic Status of Lakes People tend to settle around water and to discharge wastes, treated or not, into lakes, rivers, bays, and estuaries. As a result, the structure and functioning of aquatic ecosystems have been highly modified in most parts of the United States. The status of some of the nation's lakes is currently monitored by EPA and other agencies, but a suitable national indicator of lake status has not yet been devised. Such an indicator can be developed from the fact that a few fundamental characteristics determine the functional properties of lakes and their ability to provide the many

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 97 goods and services valued by society. The key characteristics are them- selves closely interrelated: nutrient status, rate of biological production, and net biological production. Together, these characteristics define a lake's trophic state. Together with basic physical conditions, such as morphology and hydrology, primary production in lakes is determined by inputs of energy and chemicals most importantly nutrients, inorganic ions, and natural organic matter. Lakes generally act as traps or sinks for substances, some of which are exported, but many of which are deposited in sediments. These sediments serve as repositories of information on the history of watershed conditions and the ways a lake has responded to inputs from its watershed. Lakes that have low concentrations of nutrients, low rates of primary production, and generally low standing crops of plants, are called olig- otrophic. Conversely, lakes with high nutrient levels, high plant produc- tion rates, and an abundance of plant life are called eutrophic. Lakes intermediate in these characteristic are called mesotrophic, and extremes on the ends of the trophic state continuum are termed ultra-oligotrophic and hyper-eutrophic respectively. The most appropriate trophic state condition of a lake varies accord- ing to the goods and services expected of it. For drinking water, the more oligotrophic a lake is, the better, because such waters are easier and less expensive to treat. Recreational values also decline as lakes become more eutrophic, because nuisance algal blooms, odor problems, and fish kills occur with increasing frequency, and water clarity decreases to the extent that swimmers no longer find the lake attractive. On the other hand, higher concentrations of nutrients, and thus higher rates of primary pro- duction, generally yield higher rates of fish production. However, as primary productivity increases, the nature of the fish community changes, and eventually less desirable fish species become dominant. The optimal trophic state for fishing thus depends on the type of fish desired. Lakes across the nation suffer from many stresses caused by human activities. Atmospheric deposition of acidity affects many softwater lakes in the Northeast and Upper Midwest. High concentrations of mercury in fish, also derived from atmospheric deposition, are also widespread. Sus- pended sediment from soil erosion causes poor water clarity in many reservoirs of the Southeast and Great Plains states. However, the most widespread, human-induced problem for lakes in the United States is excessive nutrient enrichment or eutrophication: it accounted for 43 per- cent of the 6.7 million acres of lakes found to be impaired in the most recent survey of national water quality (U.S. EPA 1998~. Eutrophication affects lakes of all types and sizes and in all geographic regions, including backwater areas and impoundments on many rivers and in coastal and

98 ECOLOGICAL INDICATORS FOR THE NATION estuarine waters. It is closely associated with intensive human distur- bance of watersheds through urban development and agriculture. Of the many biological, chemical, and physical characteristics of lakes that vary along the trophic continuum, three are commonly used as indicators of the trophic status of lakes: Secchi-disk transparency, total phosphorus, and chlorophyll a. These indicators respectively measure physical, chemical, and biological characteristics of lake status. Secchi- disk transparency measures water clarity in terms of the depth at which a white disk (20 cm in diameter) is just visible in the water column. In oligotrophic lakes, Secchi-disk transparency generally is greater than 3 meters; in ultra-oligotrophic lakes, such as Lake Tahoe, California, trans- parency values can exceed 20 meters. Transparency values during summer generally are less than 2 to 3 meters in eutrophic lakes and less than 1 meter in hyper-eutrophic lakes. In extreme cases, transparency is reduced to a few centimeters. Although Secchi-disk transparency is a technologically crude measurement, it has great appeal as an indicator of lake water quality because it measures a condition directly related to human perception of lake quality. Furthermore, because decreased trans- parency in most lakes is caused by algal growth, transparency is also a good measure of lake status. Phosphorus generally is the most important nutrient limiting plant growth in lakes, and total phosphorus (TP) is the most commonly used measure of trophic state. Oligotrophic lakes generally have TP concentra- tions of less than 10 ,ug/L, and eutrophic lakes TP concentrations greater than 20 to 30 ,ug/L (the criterion separating trophic state classes varies somewhat by geographic region). Hyper-eutrophic lakes may have TP concentrations as high as several hundred ,ug/L. The concentration of chlorophyll a is a direct quantitative measure of algal densities in a lake. Average chlorophyll a concentrations during the growing season in oligotrophic lakes, which have low phytoplankton den- sities, usually are less than 5 ,ug/L. Surface waters in eutrophic lakes have average summer chlorophyll a concentrations greater than about 10 ~/L; peak concentrations during algal blooms may exceed 50 ,ug/L. r-~ ~ —, Values of the three indicators vary seasonally in lakes. Although many limnologists argue that at least 5-6 measurements are needed over the growing season to accurately assess the trophic state of a given lake (e.g. Heiskary and Walker 1988; Brown et al. 1998), less-frequent sampling is sufficient for large regional surveys. One set of measurements during an index period should be adequate if measurements are done consis- tently from year to year on a well-defined (and large) sample of lakes. The critical period of recreational use for north temperate lakes is approxi- mately the beginning of fuly to the end of August. Serendipitously, the variance in trophic state variables appears to be lowest during this period,

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 99 and maximum chlorophyll a levels and minimum Secchi-disk transpar- ency values typically occur during this period (Kloiber et al. in press). Moderate interannual variations occur in the indicators in response to variations in climatic and hydrologic of factors. Response times for the indicators to changes in external nutrient loadings generally is rapid (few years or less) provided that the maintenance of eutrophic conditions in a lake is not controlled by internal recycling of phosphorus from nutrient- rich bottom sediments. Consequently, in many cases it is possible to detect a trend in the trophic status of a given lake from annual data collected over only a few years. Current Data Collection Efforts. Because Secchi transparency is easy and inexpensive to measure much information is available on lakes, and many states have citizen monitoring programs that rely primarily on transparency measurements. Abundant data also are available on chloro- phyll a and TP, but there currently is no organized program to obtain, store, and analyze lake trophic-state data on a national basis. Instead, data are gathered by a wide range of governmental units, primarily state natural resource and environmental protection agencies, municipalities, park boards, and watershed districts, as well as private lake associations, consulting firms, and university researchers. The Clean Water Act (Section 305b) requires states to report every two years on the quality of their lakes to the U.S. EPA, which in turn submits a report to Congress (e.g., U.S. EPA 1998~. However, there are no prescribed standards for data collection and reporting, and the sequence of reports cannot be used to gauge temporal trends because sampling units and reporting methods vary inconsistently over time. Indeed, it is questionable whether the 305b reports represent an accurate assessment even of lake status at a given time because sampling programs generally are not designed to allow data to be extrapolated to the population of lakes in a state as a whole. The three primary measures of lake status are strongly correlated: TP and chlorophyll a are related in a positive log-linear fashion and Secchi- disk transparency is related in a negative hyperbolic fashion to chloro- phyll a. The three indicators have been transformed into simple indices that express one concept of trophic state quantitatively (Carlson 1977~. Some states use these indices to classify their lakes according to trophic state; for example, Minnesota uses the following Trophic State Index (TSI) ranges to categorize lakes: less than 40, oligotrophic; 40-50, mesotrophic; 50-70, eutrophic; and more than 70, hyper-eutrophic (Heiskary and Walker 1988~. There has been considerable development and use of the TSI (Carlson and Simpson 1996) and related methods for collecting data. Where possible, we recommend calculating the TSI from measurements

100 ECOLOGICAL INDICATORS FOR THE NATION of all three factors: TP, Secchi depth, and chlorophyll a. TSI should include observed values of chlorophyll a where possible, since chlorophyll a pro- vides the most direct measure of biological activity. Deviations from the expected relationships among chlorophyll a, total phosphorus, and Secchi- depth transparency signal regional variations in water color, factors other than primary production that limit transparency, or limitation of produc- tion by some nutrient other than phosphorus. TSI can be aggregated nationally by computing a frequency distribu- tion of trophic states across lakes. The frequency distribution of trophic states (but not an average of TSI values) should be used as a national-level indicator, because changes in this distribution provide the most useful information. The number of lakes that become more eutrophic over time would clearly indicate mismanagement of fertilizers, sewage, and other sources of nutrients. A national increase in numbers of oligotrophic lakes would be an indication of better management of nutrients. Threshold values can be used as an additional way to direct attention to important trends in lake status. For example, increases in trophic state are not always undesirable, but hyper-eutrophic lakes have fewer valuable properties than other lakes. Therefore, the national indicator should also record the number of lakes that are hyper-eutrophic, in addition to reporting changes in frequency distributions of lakes and trophic states. Example of the Use of the Indicator. Volunteer monitoring programs are well established in many states. These programs have provided a useful baseline to assess the biological condition of lakes. TSI has been incorpo- rated into many volunteer lake-monitoring programs, because of the low cost and simplicity of collecting Secchi-disk transparency readings. Although Secchi-disk transparency does not provide a complete measure of trophic state, it provides a repeatable estimate of trophic state across a range of conditions. Some volunteer monitoring programs include chloro- phyll a and total phosphorus measurements to the array of samples col- lected by volunteers. The distribution of lakes' trophic states should be computed using an unbiased sample of lakes. However, lakes assessed by volunteers are generally biased toward recreationally accessible lakes. Agencies and municipalities generally assess lakes that provide an important local or regional resource such as drinking water. Conservation organizations and park services generally have an interest in assessing more pristine waters. Therefore, careful attention needs to be paid to the statistically reliable selection of lakes to be represented in the national index of lake status, as has been done in the selection of lakes for regional studies of the effects of acid deposition (Linthurst et. al. 1986~. Information from the large number of lakes already sampled can be

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 101 assembled to develop a baseline frequency distribution of lake trophic state. Because the Great Lakes contain so much more of the fresh water in the United States than all other lakes together, the data on these lakes should be computed and reported separately, using the many stations that are already used to record conditions in different portions of the basin under different land influences. The trophic state of each of the Great Lakes should be presented as the distribution of values among that lake's sampling stations as well as their average value, because changes in TSI values at those stations indicate changes in management of portions of each of the Great Lakes. TSI can provide sufficient resolution to detect changes in average trophic state of a few percent when as few as 50 lakes are sampled (NRC 1994~. However, given the low cost and skill required to collect these data, it is feasible to monitor thousands of lakes nationwide. A nation- wide collection of TSI values on several thousand lakes on an annual basis would provide an accurate and precise measure of trends over periods of less than 10 years. The distribution of trophic states needs to be reported and interpreted in two ways: as overall number of lakes in various trophic states, and as numbers of lakes that have changed and the direction of that change. The distribution of trophic states of lakes is a useful national indicator, but the distribution of trophic states can also be compiled regionally or locally relatively simply. How Will Technology Developments Affect This Indicator? The use of remotely sensed data could provide some measurements to replace in situ sampling. Remote sensing is likely to be most useful for measuring inaccessible lakes and providing quality control for volunteer monitoring efforts. Chlorophyll a extraction methods have recently been simplified so that field filtration and extraction can be done with only minimal train- ing. Samples would still need to be measured spectrophotometrically or fluorometrically in the laboratory. In situ fluorometry is technologically possible, but it provides a more qualitative than quantitative estimate of chlorophyll a. In situ probes with fluorometry probes would be useful to measure within-lake variance in chlorophyll, but would require confirming calibration samples to provide quantitative between-lake comparisons. Previous studies (e.g., Brown et al. 1977a, and b; Lillesand et al. 1983) have found good correlations between satellite reflectance data and water clarity and chlorophyll a concentrations. Recent improvements in satel- lite technology and especially in the software to process images quickly and efficiently, coupled with much lower costs for the images themselves, indicate great potential for use of satellite imagery to gather trophic state information on a broad regional or nationwide basis (Kloiber et al. in press). Satellite data cannot replace ground-based (in-lake) measurements

102 ECOLOGICAL INDICATORS FOR THE NATION altogether; there is a continuing need for the latter information to cali- brate reflectance data from satellites. However, satellite imagery has the potential for extending trophic state assessments virtually to every lake in a region in a very cost-effective manner. Both Secchi-disk depth and phosphorus measurements are well devel- oped and standardized. They are unlikely to be replaced by remote sensing or other field techniques. Various electronic devices can measure turbidity and transparency, but not more cost-effectively than a Secchi disk used by a volunteer. Improvements to personal computers and the Internet will enhance the effectiveness of the collection of data by a combination of government and volunteer monitors. Advances in personal computing will allow data collection and functions to be standardized and stored in a central data- base that can receive data from a large number of volunteers. Software can also standardize quality control and data verification. An Indicator of Trophic Status of Streams Stream Oxygen Indicators of the status of streams could be based on models of flowing-water ecosystems, such as the River Continuum Concept (RCC), which is a broad, integrative framework for conceptualizing stream- riparian systems (see Chapter 2~. However, the ability of the RCC to explain or predict primary production and respiration is limited, because of the complex relationships among terrestrial production, river size and flow rate, and inputs to instream processes. An indicator that better captures the balance between instream primary production and respira- tion is dissolved oxygen (DO) concentration, or stream oxygen, whose values can be directly related to NPP and respiration (R). If NPP is much greater than R. excessive plant growth (algal blooms, stream channels clogged by macrophytes) usually results; if NPP is much less than R. the result is low DO concentrations, often followed by fish kills and undesir- able odors. The concentration of DO in a river at any time is a complicated func- tion of temperature and various sources and sinks for oxygen. Oxygen solubility (S[O2~) in water decreases in a nonlinear fashion with increas- ing temperature (thus, S[O2~) = 14.5 mg/L at 0°C and 9.2 mg/L at 20°C). The main sources of oxygen in water are photosynthesis and atmospheric re-aeration (which occurs only when the oxygen concentration in the water is less than its solubility at a given temperature and pressure). The main sinks of oxygen are microbial respiration, Vitrification (bacterial oxidation of ammonium to nitrate), plant respiration, sediment oxygen demand, and degassing to the atmosphere (when the water is saturated with oxygen). Ambient concentrations much greater than the saturation

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 103 value at a given temperature and pressure indicate a preponderance of photosynthetic activity and the likelihood of high nutrient concentrations, algal blooms, and/or excessive growths of macrophytes. Ambient con- centrations much below the saturation value indicate a preponderance of respiration and the likelihood of organic enrichment from wastewater or from a high rate of plant production upstream caused by high nutrient levels. Many states set minimum standards for DO at values that gener- ally protect fish and other aquatic life. Typical minimum standards are around 5 mg/L, but in streams with particularly sensitive species such as trout, higher minimum standards are used. Spatial and Temporal Variability. Because stream oxygen can change rapidly in surface waters, care must be taken to obtain data that can be interpreted and compared with data from other sites. Because oxygen concentrations vary significantly during the course of a 24 hour period, monitoring should be done at the same time of day or, preferably, con- tinually over a 24 hour period. To aid in the interpretation of the oxygen data, data collection should include a few other basic water chemistry variables. Electrical conductivity (a measure of the total ionic content of the water), pH, and turbidity, which are all easy to measure, constitute the minimum set of chemical measurements that should accompany mea- surements of oxygen in a national stream-monitoring program. Current Data-Collection Efforts. Most states monitor DO as part of ambient monitoring programs. Additionally, federal agencies, such as the Environmental Protection Agency, the Army Corps of Engineers, and the Geological Survey, monitor DO at selected sites. Furthermore, per- mitted dischargers sometimes monitor DO in receiving streams. Example of the Use of the Indicator. Stream oxygen can be used in two ways to indicate the state of flowing-water ecosystems. DO is an impor- tant environmental requirement of fish and other aquatic life. Where DO concentrations are low, 3 to 4 mg/L or less, fish reproduction can be severely affected, and invertebrates that process organic matter can decline or be extirpated. Very high DO concentrations (supersaturation) can result from excess oxygen production either as a result of eutrophica- tion or as a result of elevation of total dissolved gases. Elevation of total dissolved gases has been associated with gas-bubble trauma and increased mortality of fishes below large dams. Oregon bases part of its water-quality index on the unweighted har- monic mean square of various subindexes, including one for stream DO (Cude 1996, Doljido et al. 1994~. The stream DO subindex is based on the oxygen concentration for measurements that are at or below 100 percent

104 ECOLOGICAL INDICATORS FOR THE NATION saturation and on the percent saturation if oxygen measures are above 100 percent saturation. This index value is scaled from 0 to 100 and is monotonic and positive for concentrations from 3.3 to 10 mg/L. For supersaturation, the index value decreases as saturation increases from 100 to 275 percent. How Will New Technology Affect this Indicator? One recent technologi- cal development that is likely to affect the use of stream oxygen measure- ments is the increasing availability of reliable probes and devices for telemetering data, which allow continuous monitoring, data storage, and remote access to data. As the cost of these technologies drops, the cost of data acquisition will decline, so that the principal obstacle will be the time needed to change the devices' batteries. Improved technology should allow more sophisticated and varied use of stream oxygen as an indicator. For example, natural patterns of primary production and respiration can produce day-night differences in stream oxygen. The absence of such differences should be cause for concern during much of the year. The new technologies described above should make the detection of those differences or their absence much easier for most sites by reducing the current need for extraordinary sampling effort. INDICATORS OF NUTRIENT-USE EFFICIENCY AND NUTRIENT BALANCES IN AGROECOSYSTEMS An agroecosystem is an ecosystem managed intensively for the pro- duction of food or fiber. Cropland, pasture, and range make up more than 55 percent of the total land area of the contiguous United States (USDA 1997~. Agroecosystems are typically studied at field, whole-farm, and regional scales (FAO 1994~; methods being developed in precision agriculture require studies at the subfield scale. In developed countries, agroecosystems are managed for high production and are characterized by intensive nutrient and pesticide inputs, fast growth-harvest cycles, and low genetic diversity of crops and animals (Odum 1984, Matson et al. 1997~. These intensive agricultural management practices can have ad- verse impacts on soil quality, and they generate offsite effects on surface water, groundwater, and the atmosphere. Agroecosystems in the United States also contain parcels of unmanaged or lightly managed areas such as woodlots, fencerows, and riparian areas. The soils of agroecosystems have long been the focus of national con- servation efforts to reduce losses by erosion. More recently, concerns have expanded beyond erosion to include salinization, compaction, loss of soil organic matter and attendant desirable physical properties, and the accumulation of trace elements and toxic substances (NRC 1993~. Over

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 105 the past five decades, increased demand, trends in dietary preferences, and the development of new technologies have led to greatly expanded use of chemical fertilizers, pesticides, and water for irrigation and live- stock production. Research to find ways to reduce the deleterious effects of these practices on surface and groundwater quality and on soils is leading to policies to protect soil and water while sustaining the profit- able production of agricultural goods, but considerable improvements can still be made (NRC 1993~. Nutrient cycles differ dramatically between agricultural ecosystems and the natural ecosystems they replaced. The most significant recent change involves massive movements of nutrients across landscapes. For example, fertilizers are transported to crop-producing areas in the spring, grain is transported to animal-producing areas in the fall, and animal manures become wastes or excess fertilizer because the rate of production exceeds local needs and the cost of transport make redistribution eco- nomically infeasible (Magdoff et al. 1997~. More land, fertilizer, pesticides, and irrigation water are needed to support animal production, and the environmental impacts are greater than if dietary choices demanded less animal protein. The importance of animal protein in human diets, which is consumer-driven, is an important factor in agriculture's impacts on the biosphere. Nations, such as the United States, with extensive concentrated animal production facilities generate large amounts of excess nutrients because nutrient use in animal production is much less efficient than in producing crops (van der Ploeg et al. 1997~. The NRC (1993) analyzed agricultural practices and impacts to iden- tify opportunities that held the most promise for "improving the environ- mental performance of farming systems while maintaining profitability." The broad recommendations of that report were the following: 1. Conserve and enhance soil quality as a fundamental first step to environmental improvement. systems. 2. Increase nutrient, pesticide, and irrigation efficiencies in farming 3. Increase the resistance of farming systems to erosion and runoff. 4. Make greater use of field and landscape buffer zones. Following this framework, the committee evaluated and recommends national-level indicators of nutrient-use efficiency and balance. The high productivity of most modern agriculture depends on added nutrients. In most agroecosystems, more nutrients are added to the sys- tem than are extracted from it in harvested products; and the imbalance is

106 ECOLOGICAL INDICATORS FOR THE NATION far greater for animal production than for crop production. These excess nutrients find their way into the soil, the atmosphere, and water. We define the proportion of added nutrients removed in products as nutrient-use efficiency. Because the efficiency with which nutrients espe- cially N and P are used in the production of crops and animal products is of great economic and environmental significance, it is important to monitor changes in inputs and outputs from agricultural lands. Losses of agricultural chemicals account for a major share of nonpoint-source N and P pollution of ground and surface waters (NRC 1993~. Because point- source control of N and P inputs to surface and groundwaters has been easier to achieve, nonpoint sources account for an increasing share of the total inputs (Sharpley and Meyer 1994~. The increasing demand for agri- cultural products will generate powerful pressures for increased agricul- tural chemical use. N export from agroecosystems is known to adversely affect drinking- water supplies. Nitrate (NO3) in drinking water can be acutely toxic, and it can cause methemoglobinemia in infants (Spaulding and Exner 1993~. The maximum contaminant level has been set at 10 mg NO3-N LO for drinking water. Regions of irrigated agriculture such as the wheat belt and California's Central Valley have the highest incidence of elevated NO3 levels, but many other such areas are scattered about the United States (Spaulding and Exner 1993, Kolpin 1997, Lichtenberg and Shapiro 1997~. N also contributes to eutrophication of aquatic and estuarine systems (Spaulding and Exner 1993~. Local, regional, and national annual N budgets for cropland and agri- cultural watersheds indicate that more N is added to croplands (manure N plus fertilizer N plus N fixed by legumes) than is removed in crops. The amount of N not transferred to crops varies widely as a function of site conditions and management practices. Site conditions that control N transformations vary among regions, along topographic gradients, and with different cropping practices. Also gaseous losses from and additions to the soil are difficult to measure; and cropping practices other than fertilization can release stored soil N at substantial rates (Keeney and DeLuca 1993, David et al. 1997~. Therefore it is difficult to rigorously determine the fate of excess fertilizer and manure N. and difficult to relate the excesses directly to elevated ground- and surface-water N levels (Keeney and DeLuca 1993, David et al. 1997, Kolpin 1997~. Gaseous losses of N (as N2O and NH3) have other influences. N2O is a significant greenhouse gas. NH3, volatilized from fertilizers and animal manure, is eventually deposited as NH4. After being taken up by plants, atmospherically deposited NH4 acidifies the soil and, in some regions (e.g., the montane watersheds of the northeastern United States), contrib-

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 107 utes substantially to the N and H+ loads that must be assimilated by sensitive landscapes (Iohnson and Lindberg 1992, Vitousek et al. 1997~. Soil phosphorus lost from agricultural systems to watersheds acceler- ates eutrophication of lakes and streams. Productivity and algal blooms in lakes and streams are promoted by elevated inputs of dissolved inorganic P. and by labile P ("algal-available P") bound to sediments or in labile organic combinations. In general, P is less mobile than N. because it adheres strongly to soil constituents. P applied in excess of crop uptake is retained to a considerable extent in nonsandy mineral soils. Quantita- tively important leaching losses of P through the soil to ground and sur- face waters are limited to areas of sandy soils, organic soils, and cases of extreme P loading. Surface runoff and tile-drain effluent from fertilized cropland, pastures, and animal-feeding operations and the accompany- ing suspended sediment are thus the most important vectors for P deliv- ery to surface waters. The importance of each source varies according to region, soil conditions, and management practices. As concern about environmental impacts of agricultural nutrient losses has grown over the past decade, so has research on how to manage agricultural nutrient use to minimize loss, while maintaining productivity and profitability. Because nutrient loss from agricultural systems is a site- specific problem (NRC 1993, Harris et al. 1995), site- and practice-specific mitigation measures are required. Many efforts are under way to gain the understanding necessary to balance environmental and productivity needs (see reviews by NRC 1993, Harris et al. 1995, Daniel et al. 1998, Sharpley et al. 1996~. Changes in nutrient-use efficiency in agricultural systems and nutri- ent losses from these systems have been driven by the increased availabil- ity of chemical fertilizer since World War II; by the availability of trans- portation for fertilizer, feed, and agricultural products; and by increases in meat consumption, which has led to the growth of specialized animal- production systems. The trend toward greater livestock production in the Western world during the latter half of this century is a major con- tributor to overall loss rates of nutrients used in agriculture, because the nutrients in manure are much less efficiently incorporated into animal products than into crops (e.g., van der Ploeg et al. 1997~. This situation is magnified in small meat-producing countries such as the Netherlands, where animal feeds grown on five to seven times the Dutch agricultural land area are imported. This large import of nutrients is driving country- wide nutrient enrichment as the manure is applied to Dutch agricultural land and excess nutrients make their way into ground and surface waters (Van der Molen et al. 1998~.

108 ECOLOGICAL INDICATORS FOR THE NATION The Indicators. Nutrient leakage is an inherent property of current agricultural activities (see review by Magdoff et al. 1997) and will remain so for the foreseeable future. Because the demand for agricultural prod- ucts will increase as the human population and economic activity increase, the only way to reduce losses of nutrients to ground and surface waters (and to the atmosphere in the case of gaseous N losses) is to develop and implement site-specific and practice-specific management techniques that improve the efficiency of nutrient use in crop-producing areas, and that limit the leaching and runoff of nutrients from animal-producing opera- tions. Considerable agricultural research is being conducted on this matter; NRC (1993) covers useful management alternatives. If animal products become less important in people's diets, overall agricultural nutrient losses will decrease (van der Ploeg et al. 1997~. However, even if substantial improvements in nutrient-use efficiency occur, overall losses from agricultural lands will increase if demands for agricultural products outpace improvements in nutrient-use efficiency. Accordingly, it is use- ful to have indicators of both the overall efficiency of nutrient use in the production of crops and animal products and the overall nutrient bal- ance. Efficiency indicators are ratios or percentages that can increase or decrease with time. Balance indicators record the excess of nutrients applied to agricultural land over nutrients removed in harvested prod- ucts. Indicators of nutrient-use efficiency and overall balance can be created for use at virtually any scale from farm fields to countries. Nutrient-Use Efficiency. N and P use-efficiency indicators are useful for crops or industries, for counties, and for watersheds in which ground or surface waters are perceived to be adversely affected. These indicators can be used in testing trends in the effectiveness of management pro- grams locally, regionally, nationally, or on a crop-specific basis. Aggre- gated data at a national scale have been useful for detecting and under- standing trends in N-use efficiency in Germany (van der Ploeg et al.1997~. N and P budgets for cropland are often very hard to construct given the difficulties in tracking nutrients applied in excess of crop uptake, and especially in determining gaseous inputs and outputs of N. Different authors have used different indicators to represent fertilizer-use efficiency, and several assumptions are usually made in estimating N budgets for croplands (e.g., lenkinson and Smith 1988, Black 1993, NRC 1993~. We have adapted the approach and budget methods used by the NRC (1993) and van der Ploeg et al. (1997) in constructing the following indicators to monitor trends in agricultural nutrient-use efficiency at a national scale. For cropland:

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS (1) Nitrogen-use efficiency Nc = N removed in crop biomass (mass y-l) chemical fertilizer N applied + animal manure N applied + N fixed by legumes (mass ye) (2) Phosphorus-use efficiency Pc = P removed in crop biomass (mass y-l) chemical fertilizer P applied + animal manure P applied (mass ye) 109 Data inputs for N and P removed in crops are crop yields by type, dry-matter percentages, and biomass N and P content. Fertilizer sales by county (e.g., EPA 1990, Smith et al. 1997), county animal censuses (e.g., U.S. Bureau of Census 1987), and per-animal nutrient excretion rates (U.S. Soil Conservation Service 1992) can be used as estimates of the terms in the equations. Legume N fixation requires assumptions, but they are straightforward and useful if they are uniformly applied (NRC 1993~. Because the indicators are percentages, they take values ranging from O to 100 percent, higher values indicating greater efficiency. Depending on availability of data, these indicators could be calcu- lated annually and should be able to detect changes in nutrient-use effi- ciency on a decadal scale. For example, in the former West Germany, crop N-use efficiency, calculated as shown above (without N fixed by legumes), decreased from about 100 percent in 1964 to a low of 72 percent in 1984, then increased to about 81 percent in 1990 (van der Ploeg et al. 1997~. Using NRC (1993) data for 1987 (NRC's Table 6-3, medium N- fixation scenario), N-use efficiency for U.S. cropland was 59 percent and for P 31 percent. Because such large quantities of nutrients in excess of those harvested are applied to U.S. agroecosystems, the potential for improvements in nutrient-use efficiency is great (NRC 1993~. Nutrient-use efficiency indicators should provide integrated mea- sures of how well management strategies are working. Application of the indicators at smaller scales would provide opportunities to target specific geographic areas or crop-production systems. A rough but potentially useful measure of the efficiency of N and P use in agriculture on a national scale can be approximated by the equation (3) Na = N content of crops produced for human consumption + N content of animal products produced (mass N ye) Chemical N fertilizer applied to cropland + N fixed by legumes (mass N ye)

110 (4) Pa = ECOLOGICAL INDICATORS FOR THE NATION P content of crops produced for human consumption + P content of animal products produced (mass p y-l) Chemical P fertilizer applied to cropland (mass P ye) These indicators could be applied to smaller scales such as states, counties, or watersheds, but at these smaller scales the indicators would need to be modified to account for imported (or exported) nutrients in crops produced for animal feed. The large decrease in N-use efficiency for agriculture as a whole (i.e., animal and crop production) from 1951 (77 percent) to 1980 (27.2 percent) in West Germany resulted from increasing rates of applications of chemical fertilizers, the rise of livestock produc- tion (a much less efficient user of nutrients than crop production), and an attendant 10-fold increase in imported N in feed (van der Ploeg et al. 1997). Nutrient Balance. The same data can be used to compute indicators of overall nutrient balance. Changes in these indicators reflect the combined effects of changes in nutrient-use efficiency and in the quantity of agricul- tural products. Tables 4.3 and 4.4 show the mass-balance approach used by the NRC (1993) for N and P applied to cropland. Such budgets require several assumptions, but they are straightforward. For nitrogen, the indi- cator would be N fertilizer applied plus recoverable N in animal waste (manure) applied plus N fixed by legumes minus N removed in harvested crops (assuming steady-state nutrient pools in crop residues). A similar indicator could be used for P. as described for the nutrient-use efficiency indicator. These national-level indicators would provide important input for understanding changes in water quality and other aspects of nutrient cycling. The units of the indicators are Mg/year for the area of interest. Nutrient-balance indicators could also be used at local and regional scales. States and counties conduct censuses of farm animals, and methods are available for estimating per capita manure production by poultry and livestock (NRC 1993). The indicator would not be affected by variations in the distances between the animal production facilities and the crop- lands that support them. The indicator would have smaller values in areas of crop-only production and higher values where animal produc- tion is concentrated. Such indicators would help states and counties evaluate their agricultural practices in light of potential nutrient losses to surface water and groundwater.

111 v) o o v) u v) o v) H bC o 5- V) V) . - O V) ~ ~U C ~ ~ O · - 5- U cr~ O .0 ·bC t~ X ~ ._' E~ U O > ~ o E~ CD O CD 5- o E~ CD O CD 5- O ~ ~ .N 50- ~ 5 - ~0 ~ . O O O O O O O O O O O O O O O O O O O O O ON ~ ~ ~ O ~ ON C~ ON ~ O p ~ ~ ~o Lr) ~ Lr) 00 ~ ~1 ~1 ~ Lr) ~ 00 ~ ~ ~ O O O O O O O O O O O O O O O O O O O O O C~ CO O O 0 00 0 00 t3N 00 CO CO ~ ~) Lr) ~ ~ O Lr) Lr) ~ Lr) ~i O O O O O O O O O O O O O 00 0 Lr) Lr) O O O O O O O O O O O O O O O O O O O O C~1 0 00 0 0 0 0 0 0 00 ~ ~ ~ ~ ' ~ ~ ~ ~ ~ O C~ CO 00 0 Lr) C~ ON 00 00 O ~ ~ C~ O O C~ Lr) oo ~ Ct) C~l~ O O O O O O O O O O O O O O O O O O O ~ ~ ~ O 00 - ` O O O O O O O O O O O O O O O O O O O ~ O ~ ~ 00 ~ ~ ~) O 00 t3N 00 ~ ~ Lr) O O O O O O O O O O O O O O O O 00 t3N ~ ~ CO ~ O O O O O O O O O O O O C~l O 00 0 0 0 0 0 0 ~ ~ ~ ~ ~ ' ~ ~ ~ ~ ~ O C~ CO 00 0 Lr) C~ ON 00 00 O ~ ~ C~ O O C~ Lr) oo ~ Ct) C~l~ ~ ~ ~ ~ o o o o ~N 00 c~i o o o o o o O O O O O O o o o o ~N 00 O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O ~ Lr) CO ~ O C~ O C~ ~ ~ O 00 t3N ~ 00 Lr) Lr) C~l 00 Lr) ~ ~ `` Lr) O~ ~ ~ 00 ~1 O O O O O O O O ~ 00 O ~ O O O O O O O O O ~ O O O ~ ~ ~ ~ Lr) Lr) oo O O O O O O O O O O O O O O O O O O O O O O O O O O O ~ ~h ~ oo o oo o o ~ Lr) ~ o oo ~ ~ ~ ~ Lr) Lr) ~ ~ `` ~ Lr) ~ Lr) ,~ cn ~ ~ o ~ 'o ~ ~ ~ ~ .= ~ ~ ~ ~ o ~ o ~ o o o o o o o ~ o o o o o o o ON O 00 ON ~N co cn u ~ ·_1 u ~ z . . o v,

112 oo V) o o V) u V) o V) H V) o V) o o V) V) o o . V) ~U 5- ¢ ~ E~ ~ U - o E~ O > ~ CD O CD 5- ~ o E~ CD O CD 5- O ~ 5- ~0 ~ . O O O O O O O O O ~ Lr) 00 ~ ~ O O O O O O O ~ 00 Lr) Lr) O O O O O O O O 0 ~1 0 0 C~ 00 00 ~ O O O O O O O O O C~l O ~ Lr) ON ~ ON ~ ~ 00 Lr) ~ O O O O O O O O O O O Ct) ~ O O O O O O Lr) Lr) - ` ON ~ ~ C~ 0 00 ~ ~ 00 ~ ~ ~ Lr) ~ ~ o o o o o o o o o o o o o o o o o o o o co Lr) oo o o ~ o ~ Lr oo o ~ oo ~ ~ ~ c~ co ~ Lr) ~ ~ ~ Lr) o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o ON Lr) O Lr) O ~ Lr) 00 ~h ~ co oo o ~ Lr) Lr) o ~ Lr) oo ~ ~ ~ Lr) ~) ~ ~ ~ ~ ~ o o o o o c~i o o o o o o o o o o o o o o o o o Lr) o o o o o o o o o o o o ~ o o o o o o Lr) o ON ~ O ~ 00 CO Lr) ~ O ON CO ~ C~ O 00 ~ ~ 00 C~ co ~ c~ o o o o o o o o o o o o o o o o o o o o o o Lr) o ~ o ~ ~ ~ o C~ C~ 00 Lr) C~ O ON CO ~ Lr) Lr) Lr) Lr) Lr) o o o o o o o o o o o o Lr) Lr ~ co o o o o o o o o o o o o o o o o o o ~ ~ o oo ~ ~ ~ ~ Lr) co co .= .= 'o ·= .= z . . o v,

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS RESEARCH NEEDS 113 Although the committee's recommended indicators are based on solid theoretical justification and extensive data, the precision and interpreta- tion of the indicators should be improved by additional research and development. Moreover, future research may suggest new indicators that may be better than existing indicators or may be usefully added to the set of regularly reported indicators. More research is needed to iden- tify organisms and biological processes that are especially sensitive to stresses and perturbations and to determine more accurately the temporal behavior of indicators. To address problems associated with spatial scale in indicator performance, it is important to test new indicators carefully in pilot programs before implementing them nationally (NRC 1995a). For the recommended indicators that are new, especially the ones that mea- sure land cover, land use, and species diversity, further work is needed on how best to operationalize the indicators and to help identify future research plans. This work, which should include one or more workshops, should include academic scientists, practitioners, agency scientists, and other interested parties. Temporal Behavior of Ecological Indicators Knowing temporal variations in indicator values is important for interpreting monitored data. Because very few monitoring efforts exceed a decade, and many for no more than a few years, other methods need to be used to provide a record long enough to capture normal variations in the proposed indicator values, as well as to investigate surprise variations. The source of much of the variation in indicators may be found at the interface between population and ecosystem processes. Population cycles are well studied, but how population cycles affect temporal patterns of species diversity, carbon storage and flow, and nutrient runoff is unknown. Experimental and theoretical investigations into the relationship between current population cycles and related ecosystem processes will provide the mechanistic understanding needed to improve predictions and inter- pret the causes of indicator behavior. In addition, research is needed on applications of the mathematical techniques of spectral analysis (Platt and Denman 1975) and wavelet analysis (Wickerhauser 1994) to time series of model outputs and to patterns in the paleorecord. Spectral analysis identifies the periods and amplitudes of fluctuations in a time series of data. An example of how this method can assist in the design of sampling systems is provided in Appendix A. Wavelet analysis identifies periods in which there is a sudden change in system behavior. These periods may correspond to

4 ECOLOGICAL INDICATORS FOR THE NATION "surprises." These mathematical techniques have been developed for analysis of digital signals and other relatively clean data sets free from stochastic effects. Ecological data, in contrast, are often incomplete be- cause of gaps in the record. In addition, in ecological research, stochastic noise often obscures the detection of fluctuations of long periods or low amplitudes, and the detection of sudden changes in the data spectrum. Research is required on adapting these mathematical techniques to the more problematic data sets typical of environmental monitoring and the paleorecord. Keystone Species Species whose removal results in a large effect on some functional property of an ecosystem called keystone species are good targets for indicators. Research is needed to develop a predictive theory of keystone species and to identify tolerant species. Traditionally, ecologists have looked for and identified keystone species by their effects on the species richness and composition of the community in which they live. Keystone species also may have major effects on primary and secondary productiv- ity and nutrient cycling. Unfortunately, although a number of keystone species have been identified, no predictive theory of keystone species yet exists. Tolerant Species and Assessing the Regional Importance of Local Sites Evaluating places using only indicators that focus on each site sepa- rately would not lead to decisions that would sustain the greatest amount of species diversity. A hypothetical example suffices to explain why. Suppose a company intends to build a factory but it does not care on which of two equal-sized natural areas it builds. For purposes of preserv- ing biological diversity, it might appear obvious that the site with the fewest species is preferable for the factory. The issue is more complicated than that because one needs to know if the diversity on the richer site is sustainable (as has been taken into account for measuring species densities), and also the extent to which the site is redundant in the system of reserves that maintains S for the region as a whole. If the richer site has virtually the same list of species as another site and the species appear to be sustainable at the other site, the value of the rich site is lower. Conversely, if the alternate site for the factory, although relatively poor in diversity, harbors species found nowhere else in the region, or nowhere else in the world, it has higher biodiversity value than the richer site. The lessons of this simple example

INDICATORS FOR NATIONAL ECOLOGICAL ASSESSMENTS 115 can be used to suggest ways to evaluate the contribution of a site, Ri, to regional diversity. To achieve this, for example, separate Si into two components. The species not found in sustainable condition elsewhere are Ri. The defini- tion of "elsewhere" may vary depending on the objective of the group doing the evaluation. In other words, elsewhere means "anywhere else in the area whose diversity one is trying to protect." In this context, a species whose range is so fragmented that its survival depends on a number of sites should be counted as a full species in the Ri of each site that forms a necessary part of its support range. To use Ri one must be able to measure sustainability. Sustainability is increased by avoiding overload in Di, but understanding sustainability fully also requires knowledge of the population dynamics, meta- population dynamics, and species interactions of the species contributing to Ri. Because obtaining all this knowledge is difficult and laborious, it cannot be done for all the species in a place. A reasonably accurate esti- mate of Ri can probably be obtained by focusing investigations on a few, charismatic taxa, such as birds, fishes, mammals, butterflies, wildflowers, and trees, instead of a random scattering of species in many taxa. How- ever, considerable research will be needed to generate the data necessary for computing measures of sustainability. As human uses take more of a region's area, many sites will shrink or disappear. As a result, the list of species restricted to a critical few sites and the average value of Ri will both climb. A place that lacks importance today may become important in the future. Thus, Ri must be reassessed periodically, probably once each decade. Ri is not an indicator but com- puting its values is an important part of an overall assessment of the performance of a system of reserves, so that biodiversity protection is achieved most efficiently.