Professional Documents
Culture Documents
Results
Regression analysis revealed hardly no correlation between Chlorophyll a concentrations measured and light (figure 4b). Although weak, temperature and salinity had positive and negative correlations to chlorophyll a values respectively (Figures 4a and c).
Temperature v Chl a
4.5 4 3.5
Conclusion
Although salinity, temperature, and light are typically strong components that determine algal concentrations, they showed little to no correlation to the observed Chl a values present during the time the data was collected. There are some areas that the map was able to accurately predict Chl a values, but largely values appear to be under estimated. This could be due to the strong currents that indirectly affect temperature regimes in the area6 or to other known factors. Even though the parameters selected were unable to give strong predictions of algal concentrations in the area, our data underscores and re emphasizes just how important the nutrient contribution is to the algal blooms in the area4. It is our suggestion that further studies continue to focus on modeling ice loss as it controls the nutrients in this system.
Chl a v Light
4.5 4
Salinity v Chl a
y = 0.0127x + 1.584 R = 0.0006 Chl a [ug/L]
4.5 4 3.5 3 2.5 2 1.5 1 0.5 0
Chl a [ug/L]
b)
c)
In particular, light penetration, temperature, and salinity are predicted to change in this area4 and are usually key indicators of algal biomass5. The aim of our study is to try to show a pattern in which chlorophyll (a) is directly associated with light, temperature, and salinity and how these components may predict the production of Chlorophyll (a). The relationship that temperature, salinity, and light have with algae can be simplified (Figure 1).
Temperature (deg C)
10
12
29
30
31
32
33
34
35
Light (PAR)
Salinity (PSU)
Measured
200 400 km
Two predictive maps were created; one where no weights were placed on the components and one where the weights were applied. Visually, neither we able to completely predict the measured Chl a concentrations. Therefore, the two maps were layered on top of each other (figure 5). This appears to have inhanced the predictive capability.
Sources Acknowledgements
Many thanks to Dr. Marjorie Brooks for providing us with this data and Dr. Tony Oyana for his guidance!
1 Springer et al. 2007. Fisheries Oceanography 5(3-4):205 2 Grebmeier et al. 2006. Science Magazine 311:1461 3 Brown et al. 2011. Jour. Geophys. Res. 116:1 4 Stabeno et al. 2001. Fisheries Oceanography10(1):81 5 Grzebyk and Berland. 1995. Jour. Plank. Res.18(10):1837 6 Napp and Hunt. 2008. Fisheries Oceanography 10(1):61
Predicted
Prediction Map
[Export_Output].[CHL_A]
Filled Contours
0.001176 - 0.207846355 0.207846355 - 0.445553412 0.445553412 - 0.718958104 0.718958104 - 1.03342132 1.03342132 - 1.39510902 1.39510902 - 1.81111314 1.81111314 - 2.28959064
0
Fig 5 Comparison of the measured Chl a values to the predicted Chl a values
Method
Regression analyses were performed to see how well each component explained Chl a data collected (Figure 4). In ArcMap, graduated symbols were used to color code each parameter, converted to raster data. Afterwards, values were reclassified using data from regression analyses (Figure 3). The new values were then extrapolated throughout the map using Kernel Smoothing and then each layer was combined into one final predictive map using raster calculator. Chl a data was also extrapolated throughout a separate map to compare the maps predictive capabilities (Figure 2).
Graduated Symbol
Output Feat.
Kernel Smoothing
Export to Raster
Raster Calculator
Criteria used for modeling Chl a concentrations: 1.The slope from the linear regression analyses was used to break each component up into their classifications. 2.The trend line was used to determine how to classify the data (i.e. if the trend was positive, increasing values received a higher ranks). Higher values were assigned to areas more likely to have higher chlorophyll a values. 3.The R2 values from the individual regression analyses were used to weight the importance of each component in the final map
Fig. 3 Criteria used to build the predictive modle from the components temperature, salinity, and light.
Feature to Raster
Reclassify
Output Feat.
Output Feat.
Final Map
Chl A
Kernel Smoothing
Final Map