Summary

Between 2018 and 2020 NJFW’s continuous stream temperature monitoring network, that was used to identify brook trout resilient habitat to climate change in NJ, was analyzed for presence of water temperature (T_w) surges as defined as a 2℃ increase of water temperature, or greater, in a 30-minute period following methodology from Nelson and Palmer (2007). Of the 227 sites analyzed between 2018 and 2020, 28 individual sites (8.1%), all within headwater streams, had at least one T_w surge and a total of 134 events identified. We were interested in the relationship between anthropogenic stressors of deforestation, paved roadways, rooftops, and agriculture on T_w surge events and how these impacts were related to the absence/presence of wild trout.

Modeling was done to understand the impact of land use types on stormwater events. The following results were documented:

• Mixed effects models were attempted to determine if the random effect of year caused significant variation in models due to variability in the amount of rainstorms by year. Variation by year was very low (<12%) and mixed models were not used.
  
• Multiple linear regression, using a stepwise approach (both ways), revealed a significant linear relationship (adjusted R² of 0.3559; p < 0.00158) between dependent variable MeanSurge and percent agricultural land (positive correlate) and percent impervious cover (positive correlate).
• Multiple linear regression, using a stepwise approach (both ways), revealed a significant linear relationship (adjusted R² of 0.1162; p < 0.04251) between dependent variable Events, and percent urban land cover (positive correlate).
• Multiple linear regression, using a stepwise approach (both ways), revealed a significant linear relationship (adjusted R² of 0.2876; p < 0.005509) between dependent variable MaxSurge and percent agricultural land (positive correlate) and percent impervious cover (positive correlate).

Modeling was done to understand the presence or absence of trout species due to land use, stormwater impacts, and stream temperature. The following results were documented:

• Multiple or simple logistic regression found no significant relationship between the predictor variables (i.e. land use variables, storm event variables, or temperature) and the presence or absence of brook trout.
• Multiple logistic regression using stepwise (both ways) selection of variables found that increase in Urban land cover was associated with a decrease in the odds of wild trout presence (p-value < 0.02791). A 1 unit increase in Urban Land Cover decreases the odds of trout presence by an odds ratio of 0.914203.

Methods

Stream temperature data was logged at 30-minute time intervals and all 227 sites were examined for increases of >2℃ within a single 30-minute interval. Events were verified with local weather station data precipitation events. Spatial analysis of landscape and land use variables were calculated in R Studio using the NLCD 2019 raster and the subshed draining to each site location. The following T_w surge metrics were extracted from the water temperature data following the same methodology as Nelson and Palmer (2007):

• Maximum surge temperature (MeanSurge) was identified as the difference between the pre-surge temperature and highest temperature attained.
  
• Duration of the T_w surge (Duration) was recorded for each event as the time until the temperature fell within 2℃ of the pre-surge temperature.
• Frequency (Events) was the number of days in the summer time period (June 15th to August 31st) that a T_w surge event occurred.
• The maximum surge temperature (MaxSurge) was taken from all T_w surge events at each site.

Multiple-variable regression models were used to test for relationships between landscape-level predictors (%impervious cover, %urban, %agriculture, %forested, %Altered Land (%Urban + % Agriculture), and %wetlands) and the several measures of summer temperature surges listed above.

The following pearson correlation table below shows the relationship between land use variables and temperature surge metrics:

There appears to be some strong positive correlations:

• Altered Land and MeanSurge
• Altered Land and MaxSurge
• Urban and Events

And strong negative correlations:

• Forest and MeanSurge
• Forest and Events

Results

Between 2018 and 2020, out of 227 sites continuously monitored, 28 (8.1%) had detected 134 temperature surges (>2℃ increase during one 30-min interval). The most surge events were recorded in 2019 (n = 73) likely do to precipitation totals in northern NJ being higher in 2019 than in 2020, and a low number of sites (n = 56) that were monitored in 2018. Sites ranged from having just one temperature surge to 16 temperature surges, a mean temperature jump of 3.26℃, and the maximum temperature surge was 11.953℃. The maximum temperature surge can be seen on the thermograph below, which took place on August 2nd, 2020:

The figure above shows stream temperature from July through August 15th of 2020 at the site with the most Altered Land (80.71%), which is the combined land cover of urban land and agricultural land. The drainage area above this site is under development and is impacted by a golf course, multiple stormwater basins, ponds, roads, and a large townhome community. Interestingly, despite the very high T_w surges, sensitive coldwater fish (i.e. brook trout and slimy sculpin) were still documented directly downstream of the impact in 2021. The primary reason is that this stream is strongly influenced by groundwater, which explains it’s unresponsive nature to increasing summertime air temperatures.

## 
## Call:
## lm(formula = MeanSurge ~ Impervious + Agriculture, data = Surge)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.0551 -0.6396 -0.1449  0.3276  2.2849 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  2.01845    0.35728   5.649 7.02e-06 ***
## Impervious   0.07536    0.03492   2.158 0.040715 *  
## Agriculture  0.04130    0.01040   3.970 0.000534 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.8489 on 25 degrees of freedom
## Multiple R-squared:  0.4031, Adjusted R-squared:  0.3554 
## F-statistic: 8.442 on 2 and 25 DF,  p-value: 0.00158

## [1] 0.8021403

## 
## Call:
## lm(formula = Events ~ Urban, data = Surge)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.3842 -2.0676 -1.2863  0.5394 10.6587 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)  
## (Intercept)  1.55197    1.50325   1.032   0.3114  
## Urban        0.09648    0.04523   2.133   0.0425 *
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.633 on 26 degrees of freedom
## Multiple R-squared:  0.149,  Adjusted R-squared:  0.1162 
## F-statistic: 4.551 on 1 and 26 DF,  p-value: 0.04251

## [1] 3.500681

## 
## Call:
## lm(formula = MaxSurge ~ Agriculture + Impervious, data = Surge)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -2.5167 -0.9429 -0.5119  0.8971  5.2893 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)   
## (Intercept)  1.92584    0.77413   2.488  0.01988 * 
## Agriculture  0.07741    0.02254   3.435  0.00208 **
## Impervious   0.15097    0.07565   1.996  0.05699 . 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.839 on 25 degrees of freedom
## Multiple R-squared:  0.3404, Adjusted R-squared:  0.2876 
## F-statistic: 6.451 on 2 and 25 DF,  p-value: 0.005509

## [1] 1.738017

## 
## Call:
## glm(formula = Trout ~ Urban, family = binomial(link = "logit"), 
##     data = Surge)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -2.2014   0.2535   0.3930   0.6298   1.3454  
## 
## Coefficients:
##             Estimate Std. Error z value Pr(>|z|)   
## (Intercept)   4.3481     1.5863   2.741  0.00612 **
## Urban        -0.0897     0.0408  -2.199  0.02791 * 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 29.096  on 27  degrees of freedom
## Residual deviance: 22.003  on 26  degrees of freedom
## AIC: 26.003
## 
## Number of Fisher Scoring iterations: 5

## (Intercept)       Urban 
##   77.333537    0.914203

Map of Frequency of T_w surge Events

In 2021, stream temperature monitoring continued and was targeted at identifying the source of stormwater impacts on streams with documented impacts between sampling from 2018-2020. The following map depicts the efforts in 2021 in an attempt to identify the primary source of the T_w surges.