The URL can be used to link to this page

Home My WebLink AboutBEC_Project_UpdateBIOLOGICAL - ENVIRONMENTAL CLASSIFICATION (BEC) SYSTEM AND SUPPORTING FLOW – BIOLOGY RELATIONSHIPS IN NORTH CAROLINA – PROJECT UPDATE Conducted by: RTI and USGS Funded by: Environmental Defense Fund, NC DENR, and NC WRC STREAM CLASSIFICATION • a priori classifications x drainage area • Biological attributes – Benthos –Fish A PRIORI CLASSIFICATIONS X DRAINAGE AREA • Level III Ecoregions (ERIII) • Level IV Ecoregions (ERIV) • Ecological Drainage Units (EDU) • Classifications subdivided by drainage basin size (DA): – Headwater (hd): < 100 km 2 – Creek & small rivers (sm):100 >= X <518 km 2 – Medium rivers (md): >= 518 X < 2,590 km 2 – Mainstem rivers (lg): >= 2,590 >= X < 10,000 km 2 – Large rivers (vl): >=10,000 km 2 3 BIOLOGICAL SAMPLING AND DRAINAGE SIZE 4 Basis Class Drainage size No. basins Headwater < 100 km2 857 Creek & small rivers <518 km2 329 Medium rivers >= 518 X < 2,590 km2 104 Mainstem rivers >= 2,590 >= X < 10,000 km2 29 Large rivers >=10,000 km2 10 TOTAL 1329 Basin Class Drainage size No. basins Headwater < 100 km2 701 Creek & small rivers <518 km2 154 Medium rivers >= 518 X < 2,590 km2 3 TOTAL 858 Invertebrate Sampling: Stream Fish Sampling: CLASSIFICATION BASED ON INVERTEBRATE BIOTA • Examination of a priori classifications in understanding invertebrate distributions and flow relations EVALUATE CLASSIFICATIONS • Indicator species: are there taxa that differentiate among classes (indicator species analysis, Dufrêne and Legendre, 1997)? • Invertebrate metrics: can invertebrate metrics differentiate among classes (CART analysis)? • Hydrologic variables: can PNV hydrologic metrics differentiate among classes (CART analysis)? 7 DATA • Sites with Excellent, Good, or Good-Fair conditions: 1,097 sites – Indicator species analysis – Invertebrate metrics in CART analysis • All sites with PNV hydrologic information: 1,734 sites. • Invertebrate metrics (168): – Richness, % richness, abundance, % abundance – Tolerance metrics – Functional group metrics – Flow preference (fast/slow) metrics (Vieira et al. 2006) – Large filter-feeder metrics 8 INDICATOR SPECIES ANALYSIS 9 Indicator species analysis identifies taxa that have a high affinity for a class based on occurrence and abundance NO. TAXA ASSOCIATED WITH ECOREGION III 10 Blue Ridge Piedmont SE Plains Coastal Plain Mayflies 38 8 4 4 All taxa 151 40 60 96 NO. TAXA ASSOCIATED WITH EDU CLASSES 11 Albemarle/Pamlico Cape Fear Pee Dee Coastal Plain Piedmont/ Fall Zone Coastal Plain Piedmont New River Coastal Plain Mayflies 2 0 1 3 16 4 All 28 16 8 18 41 27 Tennessee R Upper Pee Upper Upper Upper Blue Ridge Dee River Roanoke Santee Savannah Mayflies 3 1 3 4 16 All 20 3 13 8 72 INDICATOR VALUE CONCLUSIONS 12 • Level III Ecoregions: suitable • Level IV Ecoregions: unsuitable • EDU: suitable • Level III + Stream Size: unsuitable • Level IV + Stream Size: unsuitable • EUD + Stream Size: unsuitable INVERTEBRATE METRIC CART ANALYSIS Can CART analysis of invertebrate metrics predict ER-III, ER-IV, EDU, ERI-II+DA, ER-IV+DA, or EDU+DA classes? 13 14 CLASSIFICATION TREE EXAMPLE: ECOREG IV NO. CLASSES DETECTED WITH CART 15 Classifications Drainage area classes (hd, sm, md, lg, vl) Level IIILevel IV EDU Level III Level IV EDU No. classes 4 24 11 18 74 43 No. classes in CART model 4787 85 PNV FLOW CHARACTERISTICS AND CART ANALYSIS Can CART analysis of PNV flow characteristics predict ER-III, ER-IV, EDU, ERI-II+DA, ER-IV+DA, or EDU+DA classes? 16 CLASSIFICATION TREE EXAMPLE: ECOREG IV 17 NO. CLASSES DETECTED WITH CART 18 Classifications Drainage area classes (hd, sm, md, lg, vl) Level III Level IV EDU Level III Level IV EDU No. classes 4 24 11 18 74 43 No. classes in CART model 3969119 BENTHOS CONCLUSIONS 1. More complicated modeling may be necessary to understand the effects of flow alteration within the context of land-use changes. 2. Models are most likely State-wide models incorporating elements of classification (e.g, ERIII.DA). 3. Trade off between number of classes (spatial resolution) and significant biological differences. 19 CLASSIFICATION BASED ON FISH 20 • Finding predictive relationships between a priori classifications/biological metrics and stream fish community data CLASSIFICATION BASED ON FISH Methodology 1. Determine which classification systems (a priori, species-based, etc.) explain the most variability in stream fish community composition. 2. Examine systems identified in Step 1 to determine whether differences in cluster assignment are due to fundamental biological characteristics. 3. Using iterative 80/20 training/test random samples, fit models and predict unused observations to evaluate predictive power of classification systems. PREDICTING ECO-REGIONS WITH FISH DATA 23 Can fish data predict any of the chosen a priori classification regions? CLASSIFICATION BASED ON FISH Bottom line: EDU ‘moderately to substantially’ predicted by data INDICATOR SPECIES ANALYSIS 25 Classification System No. of significant indicator species No. of levels represented % of levels represented EDU 92 11/11 100 EDU DA 24 8/43 19 Omernick III 114 4/4 100 Omernick III DA 13 7/18 39 Omernick IV 16 8/24 33 Omernick IV DA 28 6/74 8 Bottom line: for EDU and Omernick III, all eco-region levels represented by significant indicator species EDU Region Species Indicator Value Albemarle Pamlico Coastal Plain Anguilla rostrata 0.5504 Umbra pygmaea 0.3022 Gambusia holbrooki 0.2993 Cape Fear River Coastal Plain Micropterus punctulatus 0.0524 Cape Fear River Piedmont Lepomis cyanellus 0.4343 Fundulus rathbuni 0.2673 Notropis altipinnis 0.2458 New River Etheostoma kanawhae 0.625 Phoxinus oreas 0.5784 Campostoma anomalum 0.5354 Rhinichthys obtusus 0.5333 Pee Dee River Coastal Plain Etheostoma mariae 0.2791 Tennessee River Blue Ridge Nocomis micropogon 0.4859 Notropis leuciodus 0.3783 Upper Pee Dee River Scartomyzon sp cf lachneri 0.2509 Percina crassa 0.2212 Upper Roanoke River Luxilus cerasinus 0.8709 Lythrurus ardens 0.8666 Hypentelium roanokense 0.8493 Etheostoma nigrum 0.5884 Etheostoma vitreum 0.5805 Moxostoma erythrurum 0.5716 Upper Santee River Etheostoma thalassinum 0.3644 Notropis chlorocephalus 0.3559 Cyprinella chloristia 0.3475 Upper Savannah River Notropis spectrunculus 0.4227 Scartomyzon rupiscartes 0.3526 PREDICTING ECO-REGIONS WITH FISH METRICS 27 Can fish metrics predict any of the chosen a priori classification regions? PREDICTING ECO-REGIONS WITH FISH METRICS 28 Bottom line: No* FISH CONCLUSIONS 1. Statistical significance of classes frequently based on multivariate spread of biological data 2. Permutations based on EDU or Omernick III represent most promising options for eco-region variables 3. Additional fish metrics may need to be developed 4. Could re-examine most promising eco-regions with data of lower taxonomic resolution 5. Eco-region classifications alone unlikely to contain enough information to characterize variability of stream fish data at desired taxonomic resolution NEXT STEPS 1. Do a priori classifications improve (reduce variability, improve fit) flow-biology relations relative to analyses of all sites? 2. Do response models (inver response = (Elev + Landuse + Hydro) offer better prediction of biological responses? CLASSES AND FLOW RESPONSES 31 0 20406080100 0 2 0 4 0 6 0 8 0 All sites Eco Deficit Fa s t V e l o c i t y I n v e r t s 0 20406080 0 2 0 4 0 6 0 8 0 Ecoregion ER66 Eco Deficit Fa s t V e l o c i t y I n v e r t s 0 20406080100 02 0 4 0 6 0 8 0 Ecoregion ER45 Eco Deficit Fa s t V e l o c i t y I n v e r t s 0 5 10 15 20 25 30 0 5 1 0 1 5 2 0 2 5 Ecoregion ER63 Eco Deficit Fa s t V e l o c i t y I n v e r t s Eco Deficit Ab u n d a n c e o f r h e o p h i l i c t a x a All sites Blue Ridge Piedmont Coastal Plain Ab u n d a n c e o f r h e o p h i l i c t a x a CLASSES AND FLOW RESPONSES 32 1. Eco-region class used to improve fit of flow-biology model. 2. Classification used as grouping variable. 3. For a given biological response, magnitude of flow relationship may vary by class. 4. For a given biological response, important flow variable my vary by class. 5. Use: for a new site, identify eco-region, biological response, and important hydrologic variable. Use fitted curve to identify thresholds. Confidence interval depends on modeling method. Potential outcomes: STATE-WIDE MODELING 33 5.5 6.0 6.5 7.0 7.5 8.0 8.5 23 4 5 6 7 8 EcoDeficit Intolerant richness Fi t t e d I n t o R i c h 34567 23 4 5 6 7 8 RichTOL=EcoDeficit+ELEV+IMPERV+FOR+ERIII Intolerant richness Fi t t e d I n t o R i c h RichTOL RichTOL Fi t t e d R i c h T O L Fi t t e d R i c h T O L RichTOL=EcoDeficit RichTOL=EcoDeficit+ELEV+ IMPERV+FOR+ERIII.DA where: TVi = tolerance value of taxon i n = number of taxa STATE WIDE MODELING 34 Potential outcomes: 1. Eco-region one of many predictor variables. 2. To predict how hydrologic change will influence biology at new site, need to have information for each variable in model (elevation, eco-region class, land use, etc.) 3. Use: holding all else constant, how do changes in flow variable(s) impact biological response; confidence interval easy to calculate. NEXT STEPS –WHICH SHOULD BE DONE? 1. Do a priori classifications improve (reduce variability, improve fit) flow-biology relations relative to analyses of all sites? 2. Do response models (inver response = (Elev + Landuse + Hydro) offer better prediction of biological responses?