Digital Terrain Analysis for Root River Watershed

Transcription

1 Digital Terrain Analysis for Root River Watershed Prepared by: Tisha Hooks, Winona State University Consulting Center March 28, 2014 Executive Summary: The probability of erosion tended to be higher for sites in this dataset with higher Stream Power Index (SPI) percentile values; this effect, however, was not statistically significant. Also, though erosion was more likely to occur at BMP sites than at non-bmp sites in this dataset, this effect was also not statistically significant.

2 Introduction At each of 214 sites in the Root River watershed, the following variables were measured: SPI_perc - Stream Power Index (SPI) percentile (measured on a scale from 0 to 1) Erosion? - Erosion vs. No Erosion BMP Present? - BMP vs. No BMP The objective of this project was to determine: (1) whether the probability of erosion was related to Stream Power Index (SPI) percentile, and (2) whether the probability of erosion differed between sites at which Best Management Practices (BMPs) were being implemented by the landowners. The data are summarized graphically in Figure 1. Figure 1. Scatterplot of erosion vs. SPI percentile values for both BMP and non-bmp sites. Examining the Relationship between SPI_Perc and Erosion A logistic regression analysis was used to model the probability of erosion occurring using the SPI_perc value as a predictor. SPI_perc was not found to be a significant predictor of erosion (p = ). The odds ratios and their associated confidence intervals are provided in Table 1. Table 1. Odds ratios from model using SPI_perc to predict erosion. Odds Ratio Estimate 95% CI for Odds Ratio Unit Odds Ratio (0.9972, ) Range Odds Ratio (0.7538, )

3 These odds ratio estimates can be interpreted as follows: For these data, the odds of erosion occurring increased by 0.75% (just under 1%) for every oneunit increase in SPI_perc (this increase, however, was not found to be statistically significant). The odds of erosion occurring are 2.1 times as large when SPI_Perc = (the maximum value in the data set) than when SPI_perc = 0.60 (the minimum value). Once again, this was not statistically significant. Examining the Relationship between BMP and Erosion The relationship between the presence of BMP and erosion is displayed in Figure 2. Of the 40 sites where BMP was present, 16/40 = 40% experienced erosion. Of the 174 sites where BMP was not present, 55/174 = 31.6% experienced erosion. Figure 2. Mosaic plot describing the relationship between the presence of BMP and erosion. Fisher s exact test was used to test for a difference in the probability of erosion between groups. There was no statistically significant difference found between groups (p =.3530). Considering the Effects of SPI_Perc and BMP Simultaneously A final logistic regression model was fit to predict the probability of erosion using both predictors. Neither BMP_Present (p=.2903) nor SPI_Perc (p=.1433) were found to be significant predictors of erosion. The odds ratios and their associated confidence intervals from this model are provided in Table 2. Table 2. Odds ratios from model using SPI_perc and the presence of BMP to predict erosion. Odds Ratio Estimate 95% CI for Odds Ratio Unit Odds Ratio for SPI_Perc (0.9974, ) Range Odds Ratio for SPI_Perc (0.7716, ) Odds Ratio for BMP_Present (0.7130, )

4 These odds ratios can be interpreted as follows: After accounting for the presence (or absence) of BMP, the odds of erosion occurring increased by 0.77% (just under 1%) for every one-unit increase in SPI_Perc. This increase, however, was not found to be statistically significant. After accounting for the presence (or absence) of BMP, the odds of erosion occurring are 2.15 times higher when SPI_Perc = (the maximum value in the data set) than when SPI_Perc = 0.60 (the minimum value) After accounting for the SPI_Perc, the odds of erosion occurring were about 1.5 times higher when BMP was present. This effect, however, was not found to be statistically significant. Finally, Figure 3 shows the predicted probability of erosion for both BMP present/absent and for various values of SPI_perc. The y-axis shows the probability of erosion occurring. We see that in general, as SPI_Perc increases, so does the probability of erosion (recall, however, that this effect was not statistically significant). Also, though it appears that in this data set erosion was more likely to occur at BMP sites than at non-bmp sites, this effect was not found to be statistically significant. Figure 3. Predicted probability of erosion based on SPI_perc and the presence of BMP.

5 Appendix JMP Output Output from logistic regression model predicting the probability of erosion based on SPI_perc: Output from Fisher s Exact test: Output from logistic regression model predicting the probability of erosion based on SPI_perc and BMP: