Assessment of impact of climatic change on groundwater quality around igbokod...
Poster_2
1. 1. Collect observations
Monthly sampling trips to nine
National Reference stations (NRS)
took measurements of:
• Chlorophyll-a
• Salinity
• Total Alkalinity
• Temperature
Data for up to four years and
methods of collection are
available through the IMOS portal[8].
Development of Remote Sensing for the monitoring of Ocean Acidification
around the Australian Coastline: A proposal
Baldry, Kimberlee1,2,3
Supervisors: Hardman-Mountford3, Nick; Greenwood, Jim3
1University of Western Australia
2Recipient of SGS Brian Doran Scholarship
3CSIRO Oceans and Atmosphere, Floreat
Abstract
The Australian Coastline is over 36,000 km long and comprises of diverse and unique marine life. It is important to understand climatic effects such as ocean acidification
which can impact this marine life. Remote sensing is a new area of data collection which uses satellite data to model parameters of interest. Recently, sea surface salinity
(SSS) measurements from space have become available through measurements of microwave emissions from the Earths surface. These are collected using the Soil Moisture
and Ocean Salinity (SMOS) instrument[1]. Together with sea surface temperature (SST) and chlorophyll-a (Chl-a) measurements, an algorithm can be used for modelling the
total alkalinity (TA) of Australian coastal waters[2]. Once this has been achieved, the constructed model can then be used to estimate pH levels around the Australian
Coastline. The challenge in this method lies within the development of such an algorithm, which is robust enough to predict pH with acceptable error.
Total Alkalinity (TA) is a measure of the buffering capacity of a water mass[3]. It is related to pH, pCO2 and dissolved inorganic carbon (DIC) through ocean carbonate
chemistry (Fig. 1). Calculations for pH can be performed using the program CO2SYS[4].
CO2 (g)
CO2 (aq) +H2O (l) ↔ H2CO3 (aq)↔HCO3
- + H+
TA
Figure 1: A simplistic representation of the ocean carbonate system. CO2 is dissolved into
the ocean which introduces H+ ions into the system. If TA is reduced, the ocean has less
buffering capacity to counteract the change in H+ ions, increasing pH
TA is related to salinity (SSS) linearly through the convective mixings of a water
mass;
𝑇𝐴 = 𝑎𝑆𝑆𝑆 + 𝑏
This relationship is referred to as the conservative relationship as it is does not
change within one water mass[5].
In reality other processes effect TA (calcification, primary productivity and the
mixing of different water masses)[6,7]. Adjusting the previous relationship to account
for this yields a non-conservative relationship;
𝑇𝐴 = 𝑎𝑆𝑆𝑆 + 𝑏𝑆𝑆𝑇 + 𝑐𝐶𝐻𝐿 + 𝑑
Figure 2: National Reference Station Locations
2. Models[9]
Neural Networks (NN)
• Offers a good fit with low residual
standard error (RSE)
• Poor extrapolation
NRS CONSERVATIVE NON-CONSERVATIVE
n RSE n RSE
COASTAL 1064 10.86 644 8.17
DARWIN 58 9.48 27 6.35
ESPERANCE 48 6.02 36 5.51
KANGAROO ISLAND 98 5.37 15 3.37
MARIA ISLAND 187 4.01 120 3.46
NINGALOO 29 5.82 18 2.36
NORTH STRADBROKE
ISLAND
168 4.50 133 4.17
PORT HACKING BAY 167 2.76 96 2.55
ROTTNEST ISLAND 119 4.79 67 3.53
YONGALA 191 8.63 132 8.10
3. Conservative or Non-Conservative?
The non-conservative relationship gives the better fit. In general, a good model
can predict TA with RSE of 4-10μmolkg-1 [10]..
5. Modelling pH Non-Conservatively
• The conservative model calculates pH levels with RSE below 0.01 for the regional
models.
• The coastal model offers a model with a bounded error (0.04 approx.) which shows
variable fit across locations.
6. What’s Next?
The problems with the proposed model include:
• Not suitable for short term monitoring (<10years); pH has been shown to increase
by 0.0017 units/year[2]
• Not all NRS are modelled by the coastal model so only a discrete model exists
• Models are only proven to hold for the dataset
The next steps for improvement involve:
1. Cross-validataing TA calculations with observations outside the data set
2. Using cruise data from the Southern Surveyor to investigate the continuity of the
model around the coastline
3. Applying the continuous model to satellite data and cross-validataing with
corresponding in-situ data
Figure 4: Representation of the difference between pH calculated from observed TA and
calculated TA for regional (red) and coastal (blue) models. The box range represents the RSE and
the extrema represents the maximum and minimum values.
Figure 3: How data is fitted by MLR (left) and NN (right)
Table1: The RSE and number of observations (n) for each NRS, including a combined NRS
model (coastal), for the different relationships
Multi-Linear Regression (MLR)
• Good extrapolation
7. References
1. Font, J., et al., SMOS: The challenging sea surface salinity measurement from space. Proceedings of the IEEE,
2010. 98(5): p. 649-665.
2. Sun, Q., D. Tang, and S. Wang, Remote-sensing observations relevant to ocean acidification. International
Journal of Remote Sensing, 2012. 33(23): p. 7542-7558.
3. Dickson, A.G., AN EXACT DEFINITION OF TOTAL ALKALINITY AND A PROCEDURE FOR THE ESTIMATION OF
ALKALINITY AND TOTAL INORGANIC CARBON FROM TITRATION DATA. Deep-Sea Research Part a-
Oceanographic Research Papers, 1981. 28(6): p. 609-623.
4. Lewis, E., D. Wallace, and L.J. Allison, Program developed for CO2 system calculations. 1998: Carbon Dioxide
Information Analysis Center, managed by Lockheed Martin Energy Research Corporation for the US
Department of Energy Tennessee.
5. Lee, K., et al., Global relationships of total alkalinity with salinity and temperature in surface waters of the
world's oceans. Geophysical Research Letters, 2006. 33(19): p. 5.
6. Cai, W.J., et al., Alkalinity distribution in the western North Atlantic Ocean margins. Journal of Geophysical
Research-Oceans, 2010. 115: p. 15.
7. Jiang, Z.P., et al., Variability of alkalinity and the alkalinity-salinity relationship in the tropical and subtropical
surface ocean. Global Biogeochemical Cycles, 2014. 28(7): p. 729-742.
8. Proctor, R., K. Roberts, and B. Ward, A data delivery system for IMOS, the Australian Integrated Marine
Observing System. Advances in Geosciences, 2010. 28(28): p. 11-16
9. Moussa, H., et al., A comparison of Multiple Non-linear regression and neural network techniques for sea
surface salinity estimation in the tropical Atlantic ocean based on satellite data. ESAIM: Proceedings and
Surveys, 2015. 49: p. 65-77.
10. Riebesell, U., et al., Guide to best practices for ocean acidification research and data reporting. 2010:
Publications Office of the European Union Luxembourg.