Courtesy of TravelBlog
Although the hydrologic cycle can be simplistically defined as the movement of water in various forms across the planet, it is definitely not as easy to accurately track, measure or estimate in detail the individual processes
and constituents that form the entire cycle.
One of the critical parameters that need to be measured (or estimated) in order to be able to piece together the hydrologic cycle is the snow water equivalent (SWE)—the amount of water that is contained in a snowpack.
It is relatively easy to determine snow cover over the Earth’s surface using satellite images. However, there are several challenges in accurately determining the amount of water that would be produced when the snowpack melts, especially in mountainous terrain. First, the topography in mountainous regions is uneven and complicated. The SWE is a function of elevation, orientation, vegetative cover and weather among several other factors. Second, water levels in these regions undergo huge seasonal variations, with snowfall in winters followed by thawing and melting of the snowpack in warmer climates when the melted waters flow overland as snowmelt.
In an attempt to address these challenges in the estimation of SWE, Noah Molotch, a snow hydrologist at NASA’s Jet Propulsion Laboratory (JPL), has developed a new technique for accurately tracking snowmelt based on remotely sensed data. Molotch used a spatially distributed snowmelt model to estimate the SWE of the headwaters of the Rio Grande basin in the San Juan Mountains in Colorado. The Rio Grande basin has a minimum height of 2,400 meters (7,900 feet) and a maximum height of 4,200 meters (13,800 feet). Nearly sixty percent of the yearly precipitation in this region is in the form of snowfall during winter, with the snow cover at higher elevations persisting until May or July.
Molotch’s model is based on three parameters:
- Melting Rate of Snow: This is calculated as a function of solar radiation and air temperature.
- Depletion Rate of Snow Cover: This data is obtained from images captured by NASA’s Landsat-7 satellite fitted with an Enhanced Thematic Mapper Plus (ETM+) instrument, the Terra satellite that is equipped with a Moderate Resolution Imaging Spectroradiometer (MODIS) instrument, and the Advanced Very High Resolution Radiometer (AVHRR) in the Polar Operational Environmental Satellites (POES).
- Maximum SWE Built-Up: Estimates of snowmelt are first made over a pixel at a resolution of 100 meters. The individual predictions are integrated over the time period of snow cover observation by the satellites—in this case, the snowmelt period over the years 2001 and
2002—and added up to determine the maximum SWE built-up over the entire area before the start of the melting season.
Accuracy of the Model
Molotch’s model was used to estimate SWE based on data obtained from ETM+, MODIS and AVHRR. The performance of the model varied greatly in all three instances. For data from ETM+, the model provided the most reasonable estimates vis-à-vis field measurements. However, the performance of the model deteriorated in terms of accuracy with respect to field measurements when based on data from MODIS and AVHRR.
- Variation in Data Estimates: The maximum mean SWE obtained using data from MODIS and AVHRR were 45% and 68%, respectively, lower than that obtained based on data from ETM+. These variations arose due to differences in the snow covered area (SCA) products of the three sensors. Molotch claims that SCA products from MODIS and AVHRR provided lower estimates of snow cover detection as compared to ETM+. This, in turn, led to the model yielding lower SWE estimates while using data from MODIS and AVHRR as compared to data from ETM+.
- Mean Absolute Error: At a resolution of 100 meters, the SWE estimated using data from ETM+ had a mean absolute error of 23% as compared to measurements on the ground. However, when SCA data from MODIS and AVHRR were used, the error increased to 50% and 89%, respectively. Molotch attributes this discrepancy to the difference in the temporal and spatial resolution of the three sensors. He claims that the ETM+ provided SCA products with higher spatial resolution than did MODIS and AVHRR. This caused the model, which is a spatially distributed snowmelt tracker, to provide more accurate results using ETM+ data vis-à-vis field measurements. Molotch also claims that if SCA products from MODIS were improved upon to yield higher spatial resolution, the data could be integrated into his snowmelt model to yield more accurate results.
Nonetheless, the new model comes across as a significant improvement over previous snowmelt reconstruction techniques on several grounds:
- Ten-fold Increase in Spatial Resolution: Molotch’s model can be used to estimate SWE down to a scale of 100 meters over an area of 3,419 square kilometers, which is supposedly a ten-fold increase over the resolution obtained by previous models.
- High Precision: Based on SCA data obtained from ETM+, the model was able to provide estimates of SWE to within 23% of the actual quantity determined from field measurements. This level of accuracy was also maintained when the basic parameters of the model such as melting rate of snow and depletion rate of snow cover varied under different climatic conditions.
- No Ground Measurements Required: Relying solely on data provided by satellite images, the mathematical model does not have to depend on field measurements. This provides a significant advantage while attempting to measure SWE over large mountainous regions.
Molotch’s model is definitely a giant leap towards increasingly accurate measurement of snowmelt. Further research could very well lead to new techniques for measuring snowmelt and determining its contribution to the global hydrologic cycle on a real-time basis.
