Extending Churchill’s shipping season using GIS based modeling

1

Extending Churchill’s shipping season using GIS based modeling Eric Séguin1, M. Sawada1, and K. Wilson2

1Laboratory for Applied Geomatics and GIS Science (LAGGISS), Department of Geography, University of Ottawa, 60 University St. Ottawa, ON K1N 6N9, Tel: 613-5685800 x2293, Eml: [email protected] Web: www.geomatics.uottawa.ca 2Canadian Ice Service, 373 Sussex Drive Ottawa

ABSTRACT Twenty-five percent of the grain production in Canada is located closer to the Port of Churchill than any other port. Churchill provides unique opportunities for the export of manufactured, mining, agricultural and forest products, as well as the importation of minerals, steel, building materials, fertilizer, and petroleum products for distribution in Central and Western Canada. This research aims to help extend the port of Churchill’s shipping season by determining shipping routes using geographic information systems, remote sensing and long-range ice forecasting. The need to extend its shipping season is attributable to the harshness of the climate in the Hudson Bay area. Extensive ice coverage throughout the Hudson Bay diminishes the shipping season to approximately 4 months of the year (June 23rd to November 12th). This short shipping season calls for long-range ice forecasting for shippers and the port authority to plan the large. The Canadian Ice Service (CIS) currently provides these forecasts through analog methods. In order to improve such forecasting techniques, the CIS is embarking on creating statistical and spatial models by comparing historical sea-ice with global atmospheric and oceanographic patterns. These modeling efforts will provide a forecast for the entire Hudson’s Bay. This work will feature a suitability model with a spatial-temporal analysis that predicts the path through seasonal sea-ice. Once completed, a least cost path analysis shall be conducted using the suitability model to determine the best viable routes for ships to navigate to and from the port of Churchill. This paper will demonstrate a few concepts in sea-ice prediction with GIS and will introduce the fundamental components for a thorough analysis. Key words: GIS modeling, sea-ice forecasting, sea-ice mapping 1. Introduction Geographical Information Systems (GIS) originated from the influx of microcomputers in the early 1960’s. Today, technological enhancements in the world of computers have permitted GIS to be a powerful spatial analysis tool. As such, GIS is not only a digital visualization tool for spatial and geographical data, but a science aiming to resolve, capture, model, and analyze specific issues (Worboys, 1995). Moreover, GIS has been tightly nit with remote sensing technologies, which permits the user to have practically insurmountable quantities of high resolution data. Chapman and Thornes (2003) indicate that analysis within GIS is achieved across data layers in an object-orientated programming environment allowing spatial variables to be statistically compared and thus producing new spatial datasets beneficial to a range of applications. Since this present study requires the analysis of several parameters which are spatially quantifiable, GIS is the favourable tool for forecasting the break-up of sea-ice within the Hudson Bay region. As such, GIS uses all available datasets which can be found to represent spatial variables. Conceptually, the combination of several climatology, hydrology and sea-ice datasets can be queried effectively within a GIS environment and used to predict the sea-ice distribution for a given time. Therefore, the analysis will feature the weekly regional ice charts from the Canadian Ice Service to determine the sea-ice extent during break up season over a 30 year period.

2

1.1 Study Area Hudson Bay is considered to be a large inland sea which exceeds 1 million km2 in surface area. It is connected to the Canadian Archipelago through the Foxe Basin. It can be considered shallow with an average depth of less than 150m. As such, Hudson Bay is characterized as a coastal region (Ingram & Alt, 1998). Although Hudson Bay and James Bay are to be considered separate, the scientific community usually pair them up because of their similarities. Hudson Bay is connected to the Foxe Basin by its northwest passage and also connects to Hudson Strait and Ungava Bay by its two northeast passages. The Hudson Bay water mass circulation is not well known (Prinsenberg, 1986a, c). However, the water characteristics are attributable to exchanges through the Foxe Basin and Hudson Strait. Ingram & Prinsenberg (1998) also suggest a large influx of freshwater from both rivers surrounding the bay as well as from sea-ice melt in spring and summer. They also found that the basic cyclonic circulation is maintained by inflow-outflow forcing at the northern end. Previous studies suggest that incoming tides from Hudson Strait are mainly semidiurnal with large amplitudes of approximately 3m. As in other Arctic regimes, this tide was found to be a Kelvin wave that propagates cyclonically around the bay. During winter, the sea-ice cover alters the tidal effects, making the amplitudes smaller and tidal phase advance (tide arrives sooner) (Ingram, & Alt, 1998).

2. Sea-ice Modeling Sea-ice models are computer simulations of ice movement and development. Data for modeling is taken from various sources such as reconnaissance aircraft, satellite images, ship reports, floating buoys, on site observations, etc. Ice modelling is a valuable tool due to the problematic limitations of observing ice over large regions. The Canadian Ice Service (2003) affirms that sea-ice cannot be continually monitored due to cost, technological limitations and human constraints. Because marine transportation, resource exploration requires up-to-date information about ice conditions, ice modeling is one way to provide essential data for assessing ice hazards in navigational decision-making and operational planning. (Canadian Ice Service, 2003) Ice modeling displays information such as concentration, thickness, speed, direction and location. Other information may be added such as snow depth, ridges, strength, ice pressure, melting of ice and ice growth. (Canadian Ice Service, 2003)

2.1 Types and Varieties of Sea-ice Models There are many different types and varieties of sea-ice models. The reason for the multitude of models is due to the fact that the scientific community has not agreed on the most viable way of forecasting sea-ice distributions. Another reason lies in the scale at which the model is being used. The size of the study area (geographical area) sometimes needs particular parameters and computational algorithms that highly depend on scale. For instance, Weiss and Marsan (2004) speak of scale properties of sea-ice deformation and fracturing, where large scale fracturing and deformation is induced by currents and winds, while more local scales might be induced by tidal actions. Some models account for snow ice formation while others focus on the thermal inertia of the ice floes. The reason for these different models is that certain areas are more affected by certain sea-ice parameters than others. For example, the Arctic is more affected by the thermal inertia of ice floes, while in Antarctica; sea-ice is dominated by snow ice formation (Mélias, 2002). Also, the rate at which new technological enhancements and scientific contributions to the art of sea-ice forecasting improves, new methods are continually adopted. As such, sea-ice models today are improved in order to properly model sea-ice dynamic and thermodynamic processes and there interactions with the atmosphere and the ocean (Houghton & alt., 1996). Previous studies on sea-ice forecasting were one dimensional in space and focused on sea-ice thermodynamics (Maykut and Untersteiner, 1971; Semtner, 1976), introducing vertical heat transport in the floes. Similar forecasting methods were then adopted by Parkinson and Washington (1979), which introduced sea-ice thermodynamics and the added air-sea-ice flux computations and simple ice transport.

3

Hibler (1979) introduced the widely recognized viscous-plastic rheology. Mélias (2002) indicates that Hibler’s new methods greatly improved the modelling of sea-ice circulation in Central Arctic. Global models were then introduced by Fichefet and Maqueda (1997) with relatively sophisticated sea-ice thermodynamics, viscous plastic rheology for dynamics and a mixed layer that interacted between both parameters (Mélias, 2002). As such, sea-ice modelling approaches, coupled with an ocean and atmospheric model can generally forecast sea-ice with a great degree of confidence. Dynamic models are based on momentum and mass conservation equations, which determines ice drift and deformation. General parameters needed to calculate sea-ice dynamics include ice mass per unit area, ice velocity, the Coriolis parameter and its thermodynamic counterpart (Wu & Alt, 2000). Therefore, the basis of the equation would show that ice motion depends mainly on external forces, i.e., wind stress (calculated by wind velocity), water stress (calculated by current velocity) and finally by the internal ice stress (calculated from the constitutive law). Many sea-ice models consider sea-ice as a viscous-plastic material (Hibler, 1979) and accordingly use the constitutive law for calculating the internal ice stress. Thermodynamic models calculate growth rates of sea-ice, which are based on heat exchanges at the air/ice, air/water, and ice/water interfaces (Wu & Alt, 2000). The heat fluxes are caused by solar radiation, incoming and outgoing long-wave radiation, sensible and latent heat fluxes, conduction through the ice layer, absorption and emission of energy due to phase changes (Wu & Alt, 2000). Wu & Alt (2000) indicate that the rate of growth of the ice cover is determined by the sum of heat fluxes at the ice surface and ice bottom, while the growth rate on the open water is determined by the sum of heat fluxes on the open water surface and the heat of fusion of sea-ice. The last thermodynamic parameter focuses on the thermodynamic growth rate, which is calculated using Hibler’s scheme (1979) for calculating the growth rate of compactness.

3. Methods Determining the nature of sea-ice coverage in Hudson’s Bay required using weekly regional-ice charts from the Canadian Ice Service to determine the sea ice extent during break up season over a 30 year period of digital records.

3.1 Calculating sea-ice extent from the weekly regional ice charts The weekly regional ice charts were supplied by the Canadian Ice Service (CIS), initially taken from RADARSAT images, were converted into gridded format for ArcGIS functionality by CIS forecasters. Each grid represents the sea-ice extent for a specific date. To be consistent over the data record, all dates were interpolated to the same day each year, called the historical date. As such, the weekly ice charts follow ice break up, week 26 to 47. All gridded coverage’s were converted from the Gregorian calendar to the Julian calendar in order to escape from seasonal trends. As for the breakdown of maps, it was necessary to make a histogram of the available ice charts due to the multitude of weekly charts.

Frequency of Hudson Ice Charts from 1972 - 2002

0

1

2

3

4

5

6

7

165 170 175 180 185 190 195 200 205 210 215 220 225 230 235 240 245 250 255 260 265

Julian Day

Cou

nt (F

requ

ency

)

Figure 1: Frequency of Hudson Ice Charts from 1972-2002

4

The first stage of this present analysis dealt with importing the gridded coverages in ArcGIS 9.0. Gridded coverages from July 26th 1999 to June 24th 2002 had a different coordinate systems/map projections than their counterparts. Every chart was converted into the standard projection employed at CIS as shown in Table 1 (Canadian Ice Service, 2002): Table 1: Projection of all gridded coverages

Projection Lambert Units Meters Spheroid Clarke 1866 Datum North American Datum 1927 1st Standard Parallel 49.00 2nd Standard Parallel 77.00 Central Meridian -100.00 Latitude of Projections Origin 40.00 False Easting 0.00 False Northing 0.00

Once these charts were properly projected, they were separated into their respective 5 day Julian date. This split all the available data from day 173 to day 257 in 5 day increments. The maps would then demonstrate sea-ice extent in an organized fashion, showing the general trend for sea-ice extent in the past 30 years of accumulated data throughout the break up season.

Figure 2 & 3: Representation of the classification scheme for the gridded coverage files

Figure 2 & 3 represent the weekly ice charts in gridded format with their respective attribute tables. Each map is classified into 14 separate categories, where 0 is land, 0.2 to 0.3 are water and values of 1 to 10 are ice concentrations. Once the visual verification of the classification scheme was completed, the gridded coverages were reclassified into binary maps through the spatial analyst extension within ArcGIS 9.0 into binary maps. Where a value of 0 was land or water and a value of 1 was ice. Figure 4 is reclassified; where the brown colour indicates land or water and the pink represents sea-ice extent.

5

Figure 4: Reclassified binary map of the weekly regional ice charts

Once all binary maps were completed, the analysis was undertaken to compute the running 5 day ice extent for Julian dates 173-177 using the raster calculator in the spatial analyst extension of ArcGIS 9.0.

5. Results and Discussion The final product is a series of maps indicating the 2-Dimensional sea-ice extent and its dissipation throughout the break up season.

Figure 5: Frequency of 5 day ice extents for Hudson Bay from 1972-2002 (Julian Days 173 to 217)

6

Figure 6: Frequency of 5 day ice extents for Hudson Bay from 1972-2002 (Julian Days 218 to 257)

As temperatures rise in May and June, thaw leads in the northwest portion of the bay become persistent (Canadian Ice Service, 2003). As depicted in Figure 5. Clearing of the pack ice continues southward from Chesterfield Inlet – Southampton Island area and westward from the Quebec side of the bay. This cyclonic ice melt is due to the general cyclonic circulation discussed in the introduction. Ice melt beyond that point is a slow process and accelerates in July with an open water shipping route to Churchill forming by the end of the month. This would suggest that Julian Day 213 to 217 in Figure 6 would be suitable for shipping. This ice melt contributes more fresh water than the accumulated river runoff from the entire basin over the May-June period (Ingram & Alt, 1998). The pack will often separate into a few large pieces before melting in mid-August (Julian Day 227, Figure 6) (Canadian Ice Service, 2003). However, it is important not to neglect other parameters that would contribute to the successful forecasting of sea-ice. This analysis merely models sea-ice extent in a 2-Dimensional form. More parameters need to be added to properly map sea-ice break-up and its consequent movement. In order to properly forecast sea-ice patterns it is therefore important to understand its properties. The properties of sea-ice depend greatly on its dynamic and thermodynamic processes (Lemke, P. and alt., 2000). 6. Conclusions A reciprocal relationship exists between the ice cover which regulates the thermodynamic processes in the atmosphere and the ocean’s hydrometeorological processes (Timokhov, 1984; Lemke and Alt., 2000). As such, the laws governing ice cover movement differ greatly. Hence, ice cover does not present itself as merely a solid phase of water but essentially as a unique geographical and geophysical medium which obeys its own laws of existence and is closely connected with the air and water in which it envelopes (Timokhov, 1984).

7

7. References Canadian Ice Service. “Education Corner”. [http://ice-glaces.ec.gc.ca/App/WsvPageDsp.cfm?ID=10340&Lang=eng]. Environment Canada, May 19th, 2003. Canadian Ice Service. MANICE: Manual Of Standard Procedures For Observing And Reporting Ice Conditions, 9th Ed., Meteorological Service of Canada. [http://ice-glaces.ec.gc.ca/App/WsvPageDsp.cfm?ID=172&LnId=23&Lang=eng]. 2002. Chapman, L. Thornes, J. E. The use of geographical information systems in climatology and meteorology. Progress in Physical Geography 27, 3. pp. 313-330. 2003. Fichefet, T., Maqueda, M.A.M. Sensitivity of a global sea ice model to the treatment of ice thermodynamics and dynamics. Journal of Physical Research. 102 (C6), 12609-12646. 1997. Hibler, W.D. A dynamic thermodynamic sea ice model. Journal of Physical Oceanography. 9, 815-846. 1979. Houghton, J., Filho, L. M., Callander, B., Harris, N., Kattenberg, A., Maskell, K., Climate Change 1995: The Science of Climate Change. Cambridge University Press, Cambridge, MA, 572 pp. 1996. Ingram, R.G., Prinsenberg, S. The Sea: Chapter 29. Coastal Oceanography of Hudson Bay and surrounding eastern Canadian Arctic Waters Coastal Segment (26, P). Volume 11. John Wiley & Sons, Inc. 1998. Lemke, P., Harder, M., Hilmer, M. The Response of Arctic Sea Ice to Global Change. Kluwer Academic Publishers. Netherlands. Climatic Change 46: 277-287, 2000. Maykut, G.A., Untersteiner, N. Some results from a time-dependent thermodynamic model of sea ice. J. Geophys. Res. 76 (6), 1550-1575. 1971. Mélias, D.S. A global coupled sea ice-ocean model. Elsevier. Ocean Modelling 4, 137-172. 2002. Prinsenberg, S.J. The circulation pattern and current structure of Hudson Bay. In Canadian Inland Seas, I. Martini, ed. Elsevier, Amsterdam, pp. 163-186. 1986a.

Prinsenberg, S.J. Salinity and temperature distribution in Hudson Bay and James Bay. In Canadian Inland Seas, I. Martini, ed. Elsevier, Amsterdam, pp. 163-186. 1986c. Semtner Jr., A.J. A model for the thermodynamic growth of sea ice in numerical investigations of climate. Journal of Physical Oceanography 6, 379-389. 1976. Timokhov, L.A. Dynamics of Ice Cover. Gidrometeoizdat Publishers. Leningrad, 1984. Weiss, J., Marsan, D. Scale properties of sea ice deformation and fracturing. Elsevier SAS. Comptes Rendus Physique 5 (2004) 735-751, 2004. Worboys, M. F. GIS: a computing perspective. London: Taylor and Francis. 1995.

8

Wu, H., Bai, S., Li, H., Zhang, Z., Wang, Z., Wang, K., Liu, Q. Modeling and Forecasting of Bohai Sea Ice. Journal of Cold Regions Engineering. pp. 68-80. June 2000.