#### DMCA

## Biodiversity and ecosystem services: A multi-scale empirical study of the relationship between species richness and net primary production (2007)

Venue: | Ecological Economics |

Citations: | 16 - 0 self |

### BibTeX

@ARTICLE{Costanza07biodiversityand,

author = {Robert Costanza and Brendan Fisher and Kenneth Mulder and Shuang Liu and Treg Christopher},

title = {Biodiversity and ecosystem services: A multi-scale empirical study of the relationship between species richness and net primary production},

journal = {Ecological Economics},

year = {2007},

pages = {478--491}

}

### OpenURL

### Abstract

Biodiversity (BD) and Net Primary Productivity (NPP) are intricately linked in complex ecosystems such that a change in the state of one of these variables can be expected to have an impact on the other. Using multiple regression analysis at the site and ecoregion scales in North America, we estimated relationships between BD (using plant species richness as a proxy) and NPP (as a proxy for ecosystem services). At the site scale, we found that 57% of the variation in NPP was correlated with variation in BD after effects of temperature and precipitation were accounted for. At the ecoregion scale, 3 temperature ranges were found to be important. At low temperatures (−2.1°C average) BD was negatively correlated with NPP. At mid-temperatures (5.3°C average) there was no correlation. At high temperatures (13°C average) BD was positively correlated with NPP, accounting for approximately 26% of the variation in NPP after effects of temperature and precipitation were accounted for. The general conclusion of positive links between BD and ecosystem functioning from earlier experimental results in micro and mesocosms was qualified by our results, and strengthened at high temperature ranges. Our results can also be linked to estimates of the total value of ecosystem services to derive an estimate of the value of the biodiversity contribution to these services. We tentatively conclude from this that a 1% change in BD in the high temperature range (which includes most of the world's BD) corresponds to approximately a 1/2% change in the value of ecosystem services. Introduction Biodiversity is the variability among living organisms from all sources. This includes diversity within species, between species and of ecosystems In 1972 Robert May, using linear stability analysis on models based on randomly constructed communities with randomly assigned interaction strengths, found that in general diversity tends to destabilize community dynamics Recent studies have attempted to understand the effects of diversity on ecosystem functioning using experimental ecosystems, including microcosms The debate continues. Recent experimental studies have claimed various relationships such as increases in biodiversity positively affecting productivity but decreasing stability Part of the fuel for the ongoing debate on the subject, is the fact that biodiversity is both a cause of ecosystem functioning and a response to changing conditions In this paper we try to address the BDEF relationship while leaving the 'prime mover' discussion aside. Our investigation specifically looks at the relationship between NPP and vascular plant diversity (hereon biodiversity or BD). This relationship is likely characterized by the following simultaneous causal links: • NPP responding to temperature, precipitation, soil characteristics and other abiotic factors • BD responding to temperature, precipitation, soil characteristics and other abiotic factors • NPP responding to BD • BD responding to NPP The very nature of ecological systems forces us to consider these multiple relationships between NPP and BD. Assuming temperature and precipitation (as well as other determinants of system productivity) are positive antecedents of both BD and NPP, the relationship between BD and NPP can be characterized as one of the following ( In Case 1, the positive relationship between BD and NPP is amplified by the anteceding influence of temperature and precipitation. If this were the case, we would predict that the bivariate coefficient of variation between NPP and BD should be greater (in absolute value) than the partial correlation coefficient, controlling for temperature and precipitation. In Case 2, the negative relationship between BD and NPP is suppressed by the abiotic influences. In this case, the partial correlation coefficient would be more (in absolute value) than the bivariate coefficient between NPP and BD. Note that nothing in this analysis assumes causality. The arrow between BD and NPP could also go in the other direction. In order to address this relationship we synthesized empirical data at the site and eco-region scales. Recent advances in the availability of biodiversity and NPP data have made this synthesis possible. Methods Biodiversity takes many forms (e.g. genetic, functional, and landscape diversity) in addition to simple species richness For the "site" scale of analysis (Scale 1) we performed an extensive literature search using the ISI Web of Knowledge and other tools (i.e. library-based bibliographic search engines) and were able to obtain approximately 200 observations on NPP from a total of 52 spatial locations globally. However, we found no observational studies that directly measured both NPP and total plant diversity simultaneously at specific locations. For the most part, the studies we encountered were species-specific, linking limited groups of species to NPP. Therefore, we were forced to search for data on biodiversity, environmental variables, and NPP separately, with spatial location as the key link among these data. LongTerm Ecological Research (LTER) and Forest Service research sites in North America were the only sites for which the required data were available (Knapp and Smith, 2001). Although limited in number, these sites span a wide range geographically and biophysically from temperate forests, to tundra to high mountain meadows. For NPP data in our Scale 2 (ecoregion) analysis we used recent global NPP satellite derived estimates, as explained below. Biodiversity data were the main variable of interest for the study and also the most difficult to standardize across sites. Our search revealed numerous gaps in the literature for biodiversity counts in spite of the increasing effort within the field to develop more accurate biodiversity figures. For our Scale 1 analysis, a few sites had biodiversity counts for the site, but not necessarily from the exact plots where the NPP data was derived. While this is a limitation, it is a bias that applies to all sites equally. The sites for which some information for both NPP and biodiversity was available was limited to 11 usable sites. Obtaining better biodiversity data for additional sites for which NPP measurements are ongoing could greatly expand the number of usable data points. For Scale 2, we used the work on North American Ecoregions of In addition to biodiversity, several physical environmental factors are important in explaining variations in ecosystem functions and services across sites. Temperature, precipitation, and soil organic matter content are three such factors we were able to include in this analysis. Temperature and precipitation have long been known to explain much of the basic global pattern of NPP We determined the soil type at each site using the FAO Digital Soil Map of the World (1995) and the latitudes and longitudes of the study sites. The FAO map yielded two useful figures for organic carbon content; the percent organic carbon of the topsoil and the percent organic content of the subsoil. The first thirty centimeters of soil was considered topsoil, while 30 to 100 cm was considered to be subsoil. Weighted averages were calculated when different horizons were present. 3. Scale 1: site level analysis Step-wise regression was used to determine the most S 6 1 ( 2 0 0 7 ) 4 7 8 -4 9 1 significant determinants of NPP over the entire data set. BD was incorporated untransformed and log-transformed. Stepwise regression yielded the following as the best model: NPP Aboveground Net Primary Production BD vascular plant species number P growing season precipitation Temperature, and organic carbon content proved not to be significant explanatory variables at this scale. All predictors were tested for suitably normal distributions using Q-normal plots. Tolerances were calculated for each of the predictor variables to test for collinearity. Tolerance for the biodiversity terms was only 0.09 suggesting a high level of collinearity. However, neither term was significant alone implying a nonlinear relationship. We recalculated the coefficients using a generalized linear model that showed the coefficient estimates to not be biased. For 8 out of 12 sites, this yields a negative correlation between marginal NPP and marginal BD, with influence becoming increasingly negative with lower diversity. This equation implies that the marginal rate of change of NPP with BD increases with increasing BD. Scale 2: North American eco-region analysis Ecoregions are defined as a physical area having similar environmental/geophysical conditions as well as a similar assemblage of natural communities and ecosystem dynamics. North America has been divided into 116 eco-regions for which data has been assembled for several types of biological diversity (including vascular plant, tree species, snails, butterflies, birds, and mammals), geophysical characteristics, and habitat threats (1 km resolution) and was classified into 19 land cover categories. NPP values that were labeled crop, urban, barren, ice or water, were removed from the analysis. NPP values for agricultural areas were removed from the analysis because it was expected that high fertilizer and irrigation inputs to these lands would boost NPP estimates but have a negative effect on biodiversity, thus reducing the relationship between NPP and biodiversity for intensively managed or altered lands. Therefore the aggregate area included in the analysis is loosely defined as 'natural area.' The remaining NPP values were then aggregated by eco-region to produce estimates of the average annual aboveground NPP for North American eco-regions for the year 2001. From this combination of sources we obtained data for 102 ecoregions for the following parameters: Number of Vascular Plants per 10,000 km 2 (hereafter BD for biodiversity), Net Primary Production (NPP), Mean Annual Precipitation (P), and Mean Annual Temperature (T). These data are listed in Step-wise regression was used to determine the most significant determinants of NPP over the entire data set. Precipitation was log-transformed and BD was incorporated untransformed and log-transformed. Step-wise regression yielded the following as the best model: All predictors were tested for suitably normal distributions using Q-normal plots. Tolerances were calculated for each of the predictor variables to test for collinearity. All tolerances were high except for BD, which had a tolerance of 0.28. Since the threshold of inappropriately high collinearity is generally set between 0.20 and 0.25, we retained the parameter. By including both BD and ln(BD), we are able to model a more non-linear relationship between BD and NPP, a strategy that is supported by the site-scale results above. For the vast majority of ecoregions, this yields a negative correlation between marginal NPP and marginal BD, with influence becoming increasingly negative with lower temperature However, further exploration using stepwise regression revealed a significant interaction between ln(BD) and temperature. This led us to hypothesize a variation in the relationship between NPP and BD over a temperature gradient. To assess this, we performed the following analysis. First, we ordered the ecoregions by mean annual temperature. Then using the model: We performed OLS regression using a moving window of 20 data points. We began with the 20 coldest ecoregions, and after each regression moved the window one data point in the direction of higher temperature. This yielded 83 individual regression outputs from which we took the R 2 measure of goodness of fit and the estimated coefficient for ln(BD). We also calculated the average of temperature for all twenty data points in each subset. Finally, we plotted the goodness of fit and the coefficient for ln(BD) as a function of average temperature Based on the output in Low temperature At low temperatures, the mean summer temperature (ST) explains the vast majority of variation in NPP at the ecoregional scale (R 2 ∼ 0.53). Further, neither BD nor ln(BD) were significant alone, but together they greatly improved the model. All other variables, including surprisingly precipitation, were not significant. This yielded the model: Ordinary Least Squares (OLS) regression coefficients for this model are shown in R 2 for the model was 0.65 with p < 0.0001. The squared partial correlation for the BD terms controlling for summer temperature was 0.25. Therefore in this analysis 25% of the variation in NPP corresponded to variation in biodiversity. Using the regression model, we can calculate the partial derivative of NPP with respect to BD: As with the regression over the entire data set, this is largely negative Mid-temperature Stepwise regression over data points 35-61 yielded no variables significant at the 0.10 level. Log-transformed annual precipitation was a mediocre predictor of NPP (R 2 ∼ 0.09). High temperature In the high temperature range, we could not use Summer Temperature (ST) because the tolerance was only 0.10 indicating an unacceptable level of collinearity in the predictor variables. Stepwise regression using all variables but ST yielded the following model: Ordinary Least Squares (OLS) regression coefficients for this model are shown in R 2 for the model was 0.65 with p < 0.0001. The squared partial correlation for ln(BD) was 0.26 suggesting that BD accounted for approximately 26% of the variation in NPP increases significantly (R 2 = 0.72), but regression coefficients are not much affected. Discussion: the empirical link between BD and NPP The results generate a number of discussion points. This investigation implies that the marginal rate of change of NPP with BD increases with increasing BD. While the data at Scale 1 is sparse and difficult to validate, it is worth noting a very similar model was found as at the ecoregion scale with comparable coefficient estimates. It suggests that if additional observations become available, it would be worth looking for a similar pattern of temperature dependency as was discovered at the ecoregion scale. The number of observations available for Scale 2 provided latitude for a more rigorous statistical investigation. By including both BD and ln(BD), we were able to model a more non-linear relationship between BD and NPP. Obviously the feedback effects between BD and NPP The moving window regression, with 83 model runs, suggested that it was inappropriate to fit the same model over the entire temperature gradient. Ecosystem function studies have long recognized the varying effects of temperature as a 'modulator' of ecosystem processes with various effects Further, at the low temperature end the data suggests that high biodiversity has a negative effect on NPP. For the midtemperature range we found no strong relationship in our investigations. If data were available for other abiotic factors (soil water content, soil carbon) perhaps a relationship would surface. It is also possible that at middle range temperatures the relationship between the predictor variables and NPP is not monotonic and therefore exhibits a canceling effect. In our high temperature range, we found NPP and diversity to be strongly linked. Assuming BD as independent, high biodiversity had a strong positive effect on NPP accounting for up to 26% of the variation. There were a number of factors we were unable to include in the model, like soil water and soil nitrogen content. These characteristics in natural systems can have large impacts on NPP and BD