Monitoring Snow Cover in Himachal Pradesh, India using Remote Sensing
- Arpit Shah
- Oct 17, 2021
- 9 min read
Updated: Mar 28
HYPERLINKS TO SECTIONS
BACKGROUND
The Cryosphere constitutes everything on the earth's surface that is frozen, be it glaciers, sea ice, polar caps or snow. While the inhospitable temperatures deter economic activity in these regions, they still are of vital importance to mankind as they maintain the planet's climate systems by reflecting away solar radiation. Reduction in ice composition in the polar regions is often treated as a leading indicator of widespread Global Warming. Therefore, monitoring change in the Cryosphere is of much scientific interest to researchers working in the Earth Observation sector.
The topic of this post is to monitor Snow Cover in Himachal Pradesh, India from 2018-2021. Mountains dot the state of Himachal Pradesh from its northwest to southeast and these high altitudes receive precipitation in the form of Snow. Snowmelt is the primary source of water for some of the major rivers of India. Himachal Pradesh's Snow feeds Beas, Chenab, Ravi, Sutlej and Yamuna rivers. The linkage is straightforward: less Snowfall → less Snow Cover → less Snowmelt → less Water flows to these vital, livelihood-supporting rivers.
What drove me to write on this topic was a report published last month (August 2021) which stated that the Snow Cover in Himachal Pradesh reduced by 18% between October 2020 and May 2021 - the researchers at HIMCOSTE and SAC Ahmedabad had analyzed Satellite Imagery acquired by AWIFS (Advanced Wide Field Sensor jointly developed by ISROÂ and USGS) instrument onboard ISRO's ResourceSat-2 satellite.
For my study, I have processed Satellite Imagery captured by Ocean and Land Color Instrument (OLCI) onboard ESA's Sentinel-3 satellite. I have analysed two sets of satellite imagery separately-
Four Images acquired between Winter to Summer in 2021
One Image acquired on a Summer day in each of 2018-2021 i.e. Four Images in all
The objective behind analysing the former was to observe the rate of Snow Melt. And for the latter, the objective was to see how the extent of Snow Cover is changing on a year-on-year basis.
Depicting the Satellite Imagery acquisition timeline in a calendar-view below-
METHODOLOGY
For starters, this is how Sentinel-3 OLCI Multispectral Satellite Imagery looks like in RGB mode-
The methodology that I have used to map Snow Cover is called Normalized Difference Snow Index (NDSI). Clouds can be easily differentiated from Snow in Short Wave Infrared (SWIR) wavelengths within the Electromagnetic Spectrum - SWIR is present in Solar Radiation, the medium through which Multispectral Imagery is acquired passively. However, as Sentinel-3's OLCI Instrument doesn't acquire SWIR reflection, I utilized an NDSI adaptation suited for OLCI acquisitions developed by Kokhanovsky et. al, (2019) instead - this modified algorithm delineates Snow using Near Infrared reflections (which OLCI does capture).
Depicted in Figure 4 below is the processing chain to convert raw Satellite Imagery to Snow Cover Map using SNAP software. The adapted NDSI formula is plugged in the Band Maths function of the processing chain.
Sharing a short description of the other processing steps below-
Rayleigh Correction - OLCI in Sentinel-3 captures reflection information across 21 spectral bands in Top of Atmosphere (TOA) mode and is a Passive Sensor i.e. the source of energy is not onboard the satellite - it is solar radiation. I believe you may have heard the term Rayleigh Scattering before - fine particles in the atmosphere scattering sunlight particularly in the shorter, blue wavelength which is what causes the sky to appear blue. I have to subtract this effect from the OLCI dataset in order to get an accurate representation of the solar reflectances at Surface-level i.e. remove the signal distortion due to scattering in the upper atmosphere. Converting TOA to rBRR (Bottom of Rayleigh Reflectances) is the function of this Rayleigh Correction algorithm.
Subset - Subsetting an Imagery Dataset can involve a) clipping the Geographic Extent of the Imagery and/or b) removing Spectral Bands which are not deeded and/or c) removing the Metadata that comes bundled with the Imagery package. Performing a Subset helps to reduce the size of the Imagery dataset considerably, making the processing quicker. For this Snow Mapping workflow, I've utilized just the Geographic Extent Subset.
Reproject - The purpose of applying Reprojection is to set a defined Coordinate Reference System (CRS) to the output. To understand Reprojection, one must be aware about Map Projection which involves the transfer of information captured from a 3-dimensional surface (LWH) onto the 2-dimensions of a standard Map (XY). The question is where does the reflection reading as captured by the satellite through its sensor belong to? By default, it belongs to nowhere - it is just a coordinate in an arbitrary space. The purpose of Reprojection is to to make the arbitrary XY coordinates correspond to coordinates on Earth i.e. belong to an Earth Reference System.
In case you'd like to watch a step-by-step video demonstration of the processing steps on another study area, explore this tutorial.
OUTPUTS AND INTERPRETATION
Upon processing a single Satellite Imagery dataset using the adapted NDSI for OLCI on SNAP, the output that was generated is depicted in Figure 5 below-
Depicted in Figure 6 below are the Snow Cover outputs for Himachal Pradesh across the four Imagery datasets from 2021, stacked adjacent to each other.
A marked reduction in Snow Cover can be clearly observed in the fourth image (June 2021) as compared to the previous months. Himachal Pradesh has had unseasonal snowfall in April 2021 (onset of Summer) - the third map is from this month and evidently, it shows a marked increase in Snow Cover from the previous months. Do note that this type of unseasonal Snowfall is less dense and thus, prone to quicker melting causing Avalanches.
Depicted in Figure 7Â below are the Snow Cover outputs for Himachal Pradesh across the four Imagery acquisitions from a summer day in each of the years 2018 to 2021, stacked side-by-side-
It is evident that the extent of Snow Cover in the fourth image (June 2021) is much less than the preceding two years - May 2019 & May 2020 respectively. The first image (May 2018) appears to have the least Snow Cover of them all - and one can validate this by tallying it to the findings of a comprehensive scientific study on Snow Cover conducted in Himachal Pradesh during that time.
Similarly, in Figure 9 below, I have combined all the outputs from 2018-21 (Figure 7) in a single Map-view-
While I have interpreted the output only visually thus far, it is also possible to obtain quantitative insights from within SNAP Software itself. Figure 10 below depicts the Snow Cover in terms of Geographic Extent for the four outputs from 2021-
To summarize my findings, a ~36% decrease in Snow Cover has been observed between February 2021 and June 2021. To reach a conclusive interpretation isn't straightforward as February is a Winter month while May is a Summer month i.e. it is normal for a portion of the Snow Cover to melt away in this time period. Just how much is normal is something that I'm not familiar with. However, if you recollect the findings of the scientific study I had referred to earlier in this post - the 18% decrease in Snow Cover that was observed between October 2020 and May 2021 was deemed to be unusual and was attributed to Climate Change - so I can deduce that the normal Snowmelt percentage is well below 18%. Hence, the 36% decrease in Snow Cover observed in my study is certainly alarming!
Interestingly, I would like to pose a question to the researchers in that scientific study as to why they chose October 2020 as the base month to compare May 2021 Snow Cover output to. October marks the end of monsoon - the Snow Cover is expected to be at its lowest during this time in a normal scenario and comparing it to peak Summer output where again the Snow Cover is less due to extensive melting doesn't seem like a natural comparison to me. While I can give the researchers the benefit of doubt - they are likely to have a compelling reason behind doing so, nonetheless, as informed citizens, we must not take research findings published in news portals at face value - the headline and content typically focuses on an eye-popping finding but most articles do not reveal or elaborate the methodology and rationale of the parameters utilized in the study.
That being said, that I was unfamiliar about the standard Snowmelt percentage was the precise reason why I chose to compare Snow Cover on a year-on-year basis during roughly the same time (a day during Peak Summer). This output would complement my 2021 month-on-month output and would validate whether the Snow Cover is indeed reducing, confirming the presence of Climate Change.
Figure 11Â below depicts the Snow Cover in terms of Geographic Extent for the four outputs from 2018 to 2021-
To summarize my findings, a 42% decrease in Snow Cover can be observed in the June 2021 output upon comparing it to the May 2019 output (a 34% decrease if I compare June 2021 to May 2020). The severe brunt of Climate Change is evident, and it validates what was observed in the 2021 outputs as well.
That being said, a 25% increase in Snow Cover was also observed when comparing June 2021 output to that of May 2018. While this goes against the Climate Change narrative - one can argue, and rightfully so, that the impacts of Climate Change are not linear and unidirectional, rather they are characterized by an increase in volatility of weather conditions.
Besides, two other potential reasons which could have contributed to this anomaly are-
a small section of Himachal Pradesh is not present in the Satellite Imagery from May 2018 - as can be observed on the right and on the bottom of Figure 12 below - I was late to realize that the image that I had selected did not contain the whole study area. The missing sections would reduce the Snow Cover extent in the base image, however, I don't think it would be very large - the Snow Cover of May 2018 would likely increase by only a single digit percentage.
Figure 12: May 2018 Imagery does not completely cover the violet-shaded extent of Himachal Pradesh More importantly - May 2018, where 9096 sq. km. of Snow Cover was estimated, was a very warm month. News reports validate that this region was under the spell of a record Heatwave during that time. Besides, from Figure 10 & 11, you will observe that in all the remaining 7 of the 8 derived outputs, the Snow Cover is above 10,000 sq. km - so May 2018 could very well be an anomaly by itself exaggerating the 'increase' in Snow Cover in 2021.
PARTING THOUGHTS
Himachal Pradesh isn't alone. Anthropogenic Global Warming-induced Climate Change is impacting the Cryosphere - the rates of Snowmelt have increased substantially across various locations around the world. For example, in the Sierra Nevada mountain range in USA, I've observed that the Snow Cover in May 2021 was 90% lower than that of February 2021!
Remote Sensing is invaluable for Earth Observation Workflows such as this. It helps scientists and researchers to monitor phenomena at a regional and continental scale, benefiting from frequent observations due to the high revisit time (temporal resolution) and sufficiently-high spatial resolution of the satellites. The insights derived from processing Satellite Imagery enables governments and institutions to take risk-mitigating measures in a bid to combat and counter Nature's fury.
Feel free to share your thoughts.
ABOUT US
Intelloc Mapping Services, Kolkata | Mapmyops.com offers Mapping services that can be integrated with Operations Planning, Design and Audit workflows. These include but are not limited to Drone Services, Subsurface Mapping Services, Location Analytics & App Development, Supply Chain Services, Remote Sensing Services and Wastewater Treatment. The services can be rendered pan-India and will aid your organization to meet its stated objectives pertaining to Operational Excellence, Sustainability and Growth.
Broadly, the firm's area of expertise can be split into two categories - Geographic Mapping and Operations Mapping. The Infographic below highlights our capabilities-
Our Mapping for Operations-themed workflow demonstrations can be accessed from the firm's Website / YouTube Channel and an overview can be obtained from this brochure. Happy to address queries and respond to documented requirements. Custom Demonstration, Training & Trials are facilitated only on a paid-basis. Looking forward to being of service.
Regards,