Barbara Allen-Diaz

Classification trees: an alternative non-parametric approach for predicting species distributions.

Abstract. The use of Generalized Linear Models (GLM) in vegetation analysis has been advocated to accommodate complex species response curves. This paper investigates the potential advantages of using classification and regression trees (CART), a recursive partitioning method that is free of distributional assumptions. We used multiple logistic regression (a form of GLM) and CART to predict the distribution of three major oak species in California. We compared two types of model: polynomial logistic regression models optimized to account for non-linearity and factor interactions, and simple CART-models. Each type of model was developed using learning data sets of 2085 and 410 sample cases, and assessed on test sets containing 2016 and 3691 cases respectively. The responses of the three species to environmental gradients were varied and often non-homogeneous or context dependent. We tested the methods for predictive accuracy: CART-models performed significantly better than our polynomial logistic regression models in four of the six cases considered, and as well in the two remaining cases. CART also showed a superior ability to detect factor interactions. Insight gained from CART-models then helped develop improved parametric models. Although the probabilistic form of logistic regression results is more adapted to test theories about species responses to environmental gradients, we found that CART-models are intuitive, easy to develop and interpret, and constitute a valuable tool for modeling species distributions.

Vegetation change along gradients from water sources in three grazed Mongolian ecosystems

Foliar cover of plant species; grass, forb and total herbaceous biomass; soil P, K, N and C; and percent coarse fraction of soils were sampled over two years along grazing gradients from livestock water sources in three grazed Mongolian steppe ecosystems of varying productivity. Samples within each of the three systems (mountain-steppe, and desert-steppe) were classified into plant communities using TWINSPAN and species-environment relationships in each system were examined using CCA. Community classifications were driven by the presence/absence of ruderal species and highly palatable grasses in the steppe and mountain-steppe and by the presence/absence of salt-shrub or Caragana shrub species and associated Iris species in the desert-steppe. Ordinations were largely driven by soil nutrient concentrations, particularly P and K, in all three zones. Bulk density and percent coarse fraction to 10 cm were also important in the desert-steppe. Distance from water, which we assumed to be inversely related to grazing pressure, was a significant driving factor in steppe and mountain-steppe ordinations, and was negatively correlated with P and K. We speculate that elevated nutrient concentrations near water sources result from livestock redistributing nutrients in the landscape by voiding urine and feces in the areas where they congregate. Livestock may thus influence species composition in these systems both through the direct effects of defoliation and trampling, and the indirect effects of nutrient enrichment and depletion over the broader landscape. This hypothesis deserves further testing under controlled conditions.

SAGEBRUSH-GRASS VEGETATION DYNAMICS: COMPARING CLASSICAL AND STATE-TRANSITION MODELS

The State-Transition (ST) model has been proposed as a replacement for the widely used Classical linear succession model and its derivative Range Condition (RC) model for describing and predicting rangeland community dynamics in response to man- agement. Although the ST model offers significant advantages because it accommodates nonlinear and nonequilibrium theory and is more amenable to quantitative testing of hy- potheses about community change, to date its applications have not fully utilized those advantages using empirical data. We compare the utility of the Classical, RC, and ST models in the Artemisia tridentatalPseudoroegneria spicata vegetation type, utilizing a long-term data set from southeastern Oregon. First we develop and examine the Classical and RC models for their ability to describe and predict observed vegetation changes; second we develop an ST model by classifying and quantitatively identifying states and transitions that were observed over a period of 20 yr. The Classical and RC models adequately describe most of the observed changes in vegetation through use and application of broad descriptive categories on extensive range sites. This greatly reduces the utility and predictive value of the Classical and RC approaches as guides to management and restoration. The states and transitions developed quantitatively for the ST model offer considerably more precision and predictive value than the Classical/RC models but require large, long-term, site-specific data sets, which are usually not available for rangeland systems. The specific seral stages developed for the Classical model and commonly described in the literature differ signif- icantly from the states and transitions derived quantitatively from the empirical data. States in the ST model exhibit a significant time dependency and could not have been adequately developed without long-term observations.

Spring-water Nitrate Increased with Removal of Livestock Grazing in a California Oak Savanna

We characterized spatial and temporal changes in nitrate concentrations of the leachate from annual grasslands and subsequently emergent spring-waters and tested the effect of livestock grazing removal on them. Nitrate patterns indicated that annual grassland soils are a likely N source to spring-fed wetlands, which appear to intercept and transform N along its hydrologic path from upland soils to spring-fed, headwater streams. Aboveground biomass and soil N extractions suggested that removal of livestock grazing from these wetlands impaired this function by allowing dead plant material to accumulate inhibiting plant production (hence, plant N demand), resulting in elevated stream-water nitrate (NO3−) concentrations. Nitrous oxide (N2O) fluxes indicated that grazing removal may increase the relative importance of this N-loss pathway. Microbial biomass varied with season but was not affected by grazing treatments suggesting that N2O losses were related to differences in NO3− availability rather than grazing effects on microbial community composition or their activity. Spring-fed wetlands provide important ecosystem services such as plant uptake and denitrification at transition zones between terrestrial and aquatic ecosystems. These N-retention and transformation functions may be enhanced through biomass harvesting by livestock.

Grazing effects on spring ecosystem vegetation of California's hardwood rangelands.

Three watersheds at the University of Californias Sierra Foothill Research and Extension Center (SFREC), Marysville, Calif. were selected to study cattle grazing effects on the vegeta - tion surrounding cold-water springs and their downslope creeks. Three spring-creek systems from each of 3 watersheds were ran- domly assigned to grazing treatments (9 total). Treatments were ungrazed, lightly grazed (1,500 kgha -1 residual dry matter), and moderately grazed (1,000 kgha -1 residual dry matter) based on degree of use in upland pastures encircling the spring-creek sys - tems. Total herbaceous cover at springs varied significantly among the 6 years only once (greater in 1994 than all others) covarying with previous years rainfall. Grazing intensity did not affect total herbaceous cover at springs. A year X grazing treat - ment interaction (P<0.05) was detected for total herbaceous cover at spring-fed creeks. Three years after grazing removal, total herbaceous cover on ungrazed creek plots surpassed cover at moderately grazed and lightly grazed plots. Moderately grazed plot herbaceous cover declined steadily throughout the first 3 years, while lightly grazed cover remained relatively sta - ble. Plant community composition and stability by year and graz - ing treatment were analyzed with TWINSPAN. With few excep- tions, stable plant communities persisted on sites regardless of grazing intensity or cover changes. Total herbaceous cover was sensitive to interannual fluctuations, especially under increased grazing intensities. This attribute renders cover a more useful gauge of ecosystem health than plant composition as the latter may not provide evidence of potentially deleterious grazing X cli - mate interactions until after soil erosion or water table charac- teristics are seriously, perhaps permanently, altered.

DETECTING CHANNEL MORPHOLOGY CHANGE IN CALIFORNIA'S HARDWOOD RANGELAND SPRING ECOSYSTEMS

Permanent channel cross-sectional transects perpendicular to flow were used to estimate changes in spring and resultant creek channel morphology. Three cattle grazing treatments (none, light, and moderate) were applied to 2-5 ha pastures containing a perennial spring and resultant creek cohort for 5 years. Grazing effects on the total change in channel morphology were not detected, nor did our method detect channel morphology change over the 5 year study period. Ungrazed springs and creeks were observed to change more than grazed springs and creeks although these differences were not statistically significant. Observed, but not significant, change over time appears related to rainfall patterns. Permanent channel cross-sections, one of the currently recommended methods for monitoring livestock grazing impacts on stream channek

How do trees affect spatio-temporal heterogeneity of nutrient cycling in mediterranean annual grasslands?

may not be adequate for detecting channel changes in low-flow spring/creek systems.

Cattle Grazing and Yosemite Toad (Bufo canorus Camp) Breeding Habitat in Sierra Nevada Meadows

Abstract• In this study we analyzed heterogeneity in nutrient cycling induced by trees in Mediterranean annual grasslands, comparing years of higher and lower than average precipitation and analyzing the effects of two different solar radiation scenarios.• Organic matter and consequently upper soil N, K, Ca and Mg were significantly greater in those locations receiving the highest levels of solar radiation, and as expected from many other studies in the literature, there was an increase in all macronutrients (except P) as well as pH below the canopy.• Contrary to what was expected, plant nutrient concentrations did not directly reflect those found in the soil, with the exception of K. The studied grassland responded to increased nutrient availability by enhancing growth and changing botanical composition rather than by increasing plant nutrient concentrations. Hence, the total amount of accumulated nutrients in the ecosystem was larger below the tree than outside it, although this is mainly a consequence of plant growth enhancement. The levels of Ca, Mg, and Na in plants decreased during the driest year, and the N content was mostly determined by the composition of the grass.• Temporal nutrient variability, particularly within-years, explained most of the variability in plant nutrient concentration, while spatial variability induced by trees was determined to be of secondary importance. These results are significant for ecosystem nutrient modelling.Résumé• Dans cette étude, nous avons analysé l’hétérogénéité du cycle des nutriments induite par les arbres dans les prairies méditerranéennes annuelles, en comparant les années à précipitations supérieures et inférieures à la moyenne et en analysant les effets de deux scénarios de rayonnements solaires différents.• La matière organique et par conséquent N, K, Ca et Mg des horizons supérieurs du sol étaient significativement plus élevés dans les stations recevant le plus de rayonnement solaire. Comme prévu à partir de nombreuses autres études dans la littérature, une augmentation de tous les macronutriments (sauf P) et aussi du pH a été observée sous la canopée des arbres.• Contrairement à ce qui était prévu, les concentrations de nutriments dans les végétaux n’ont pas reflété directement celles trouvées dans le sol, à l’exception de K. Les prairies étudiées ont répondu à un accroissement de la disponibilité en éléments nutritifs par une augmentation de la croissance et un changement de la composition botanique plutôt que par une augmentation des concentrations de nutriments dans les plantes. Par conséquent, les quantités totales de nutriments accumulées dans l’écosystème étaient plus importantes sous les arbres qu’en plein découvert, bien que cela soit principalement une conséquence de l’amélioration de la croissance des plantes. Les niveaux de Ca, Mg, et Na dans les plantes ont diminué au cours de l’année la plus sèche et la teneur en nutriments azotés était largement déterminée par la composition de l’herbe.• La variabilité temporelle intra-annuelle des éléments nutritifs expliquait la plus forte part de la variabilité de la concentration en nutriments des plantes annuelles, tandis que la variabilité spatiale induite par les arbres était d’importance secondaire. Ces résultats sont importants pour la modélisation des éléments nutritifs dans les écosystèmes.

Grazing for Biodiversity in Californian Mediterranean Grasslands

Abstract Exclusion of cattle by fencing has been proposed to alleviate possible negative grazing impacts on hydrologic, water quality, and cover habitat conditions within Sierra Nevada meadows used by Yosemite toads (Bufo canorus Camp) for breeding. Our objectives were to: 1) determine associations between breeding pool habitat conditions and use of potential breeding pools by toads; and 2) determine how habitat conditions respond to cattle exclusion treatments on the Sierra National Forest, California. We randomly selected two toad occupied and two unoccupied breeding pools in each of nine meadows for this study (n  =  36 breeding pools). After baseline data collection in 2006, three meadow fencing treatments were implemented over the course of 3 yr. Treatments were fencing to exclude cattle from the entire meadow; fencing to exclude cattle from toad breeding and rearing areas, with grazing allowed in the remaining unfenced portion of the meadow; and cattle grazing allowed across entire meadow. We monitored hydrologic, water quality, and cover habitat variables as well as toad occupancy during the breeding seasons of 2006 through 2008. Concentrations of water quality constituents were uniformly low all years regardless of treatment. Occupied pools were shallower, warmer, and more nitrogen enriched than unoccupied breeding pools. We found no evidence of improved toad breeding pool habitat conditions following fencing compared to standard US Forest Service grazing management.

Using Digital Photography to Examine Grazing in Montane Meadows

On the Ground Californias Mediterranean climate zone supports grasslands that are biologically diverse. Livestock grazing is being increasingly used to promote native species diversity at both the pasture and landscape scales. Several federally and state-listed vertebrates and insects respond positively to grazing to improve habitat by opening and lowering grassland vegetation. More work is needed on enhancement of native plants. Research results need to be more extensively applied, tested, and monitored under variable conditions.