We continue with our assessment of digital elevation models (DEM) from last week, this time comparing the four DEMs to 115 million highly accurate IceSat-2 terrain heights measured along flat, non-sloping areas across Australia and New Zealand. We used Global Surface Water and Copernicus LC100 to assign each comparison to one of six land cover types: forest, short vegetation, wetland, barren, urban or open water (excluding ocean).
In general, New Zealand is mountainous and Australia, flat. So we analyzed each country separately to see how each DEM reacted.
Tables 1 - 6 capture Australia across the six land cover types:
The DEMs performed a little better across Australia than North America. For AW3D30, this is likely due in part to more scenes being captured across Australia than almost anywhere else, as illustrated on Figure 2 of this report.
Tables 7 - 12 capture New Zealand across the six land cover types:
Did NASADEM beat SRTM down under? For water heights, it's a wash! For land heights, NASADEM is better, some times significantly. Also, NASADEM performed better in Australia than New Zealand, very much so in forested, short vegetation and barren areas. Simply put, across Australia & New Zealand, NASADEM is absolutely better than SRTM when measuring terrain heights across flat land surfaces.
ASTGTM v003, AW3D30 v3.1 and NASADEM_HGT v001 are new & improved digital elevation models (DEM). We wanted to know how they compare to SRTMGL1 v003, which is used widely across industry and powers many products including our Find Best Sites service. (These two references provide an overview of these and other related DEM products.)
Our assessment compared these four DEMs to 252 million highly accurate IceSat-2 terrain heights measured along flat, non-sloping areas across North America, from the southern tip of Mexico to 400km north of Alaska. We also wanted to see how each DEM responded to different land cover types. So, we used the North American Land Change Monitoring System to assign each comparison to one of six land cover types: forest, short vegetation, wetland, barren, urban or open water (excluding ocean).
Tables 1 - 6 capture the results, for each land cover type. Figures 1 - 6 compare SRTM with NASADEM and Figures 7 - 12 compare AW3D30 with NASADEM; horizontal axes measure DEM height minus IceSat-2 terrain height.
The graphs in Figure 1 extend far to the right of the center line, and indicate that the DEM captured heights within the tree canopy instead of the terrain surface below, and is a well known property of C-band synthetic aperture radar. NASADEM reduced this bias as shown by its graph's slightly more balanced distribution on either side of the vertical line, and higher peak. Figures 2 and 4 represent land cover types that give the radar more exposure to the terrain surface.
We were also interested in how NASADEM compared with AW3D30 v3.1. The results are captured below in Figures 7 - 12:
AW3D30 more accurately represented terrain heights in areas with little to no vegetation (Figures 8 & 10). Another advantage of AW3D30 is coverage, which extends 2,500km north of where NASADEM SAR coverage ends at 60.25°N.
So, is NASADEM better than SRTM? Yes, for some land cover types, and no for others. Also remember that this analysis applies to flat terrain only; DEMs can behave differently in sloping or rugged terrain.
NASADEM_SHHP v001 is a related NASADEM product, offering more precision at the cost of voids and other anomalies. When corrected (it's easy to do) it also offers slightly better performance than NASADEM itself.
The IceSat-2 mission began in Oct 2018 and continues to collect extremely accurate terrain heights that can help assess the quality of DEM products in ways that were never before possible. Analysis should group results by land cover type, to identify where each DEM is strong and where it is weak. All DEMs have their place; it's just a matter of discovering where. The GEDI mission started collecting similar terrain heights in January 2019 and can serve similar purposes.
(Neuenschwander, A. L., K. L. Pitts, B. P. Jelley, J. Robbins, B. Klotz, S. C. Popescu, R. F. Nelson, D. Harding, D. Pederson, and R. Sheridan. 2020. ATLAS/ICESat-2 L3A Land and Vegetation Height, Version 3. [14N - 75N latitude; 52W - 170W longitude]. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. doi: https://doi.org/10.5067/ATLAS/ATL08.003. [Date Accessed 2020-Jul-19].)
This month saw no changes for Telus, 200 fewer channels for Bell and 53 more sites for Rogers, including 18 with 5G for a total of 490.
The graphs below show that channel counts, occupied spectrum and site counts have remained stable across all seven carriers for the past few months. This is good news. The only exception is a small matter, and relates to the SMS Spectrum Licences Site Data file itself, which has about 220,000 identical records.
This 90 second tutorial explains the importance of Line-of-sight and the Fresnel Zone to wireless link planning.
Loxcel Cellular Services provides simple yet powerful tools to help you with all this, and more.
Please contact us for more information.
You can now close any pop-up window by clicking the shaded background, identfied by the green checkmarks:
This method also works with the Account & Subscription Info pop-up.
The new Export KML feature works best with accurate channel antenna information. To that end, we cleaned up most of the antenna model information provided by Canadian licensees to ISED:
(Channel counts for Bell and Telus include many duplicates.)
This cleanup offered additional benefits:
We'd like to welcome these three Canadian licensees to Loxcel Cellular / Broadband Services:
|TekSavvy SkyFi||Broadband||South-west Ontario||20|
SaskTel fusion and TekSavvy SkyFi are accessible to our Broadband service subscribers. Please contact us for information.
Loxcel Cellular Services can now export multiple cellular and fixed wireless sites to one Google Earth KML file. The video at the right captures a flyby through one of these KML files. It shows
Each site appears as a simple monopole, which simplifies the depiction of its antenna layout.
The Places panel below provides detailed technical information about each site, as well as the means to "fly" to a site or zoom in on its antenna panels.
The hollow arrow points at the monopole, and shows structure height (16m), total channel count (66), tenants (Rogers, Telus) and site code (W5064). Double click this row to view the entire site in the 3D viewer. The green arrow points at a Telus panel 16m above ground, 30° azimuth, no tilt and Tx 743-2150MHz. Double click this row to fly to its panel. The red arrow points at a Rogers panel 14m above ground, 110° azimuth, 6° tilt and Tx 735-1972MHz.
We created the video above by clicking on selected monopoles and antenna panels.
To create your own KML file, expand the control panel, center the map over the sites of interest, and click Export KML. The right arrow points to a counter of how many sites will be exported. Table 1 lists how many sites you can export.
|Subscription||Detailed 3-D||Simple pushpin|
|24 Hour||2 sites||8 sites|
|3 Month||5 sites||20 sites|
|6 Month||10 sites||40 sites|
|12 Month||20 sites||80 sites|
|Google Earth Pro on desktop||Yes||Disable 3D Buildings|
|Google Earth on mobile||Yes||Tested on iPhone 8 and iPad Pro|
|Google Earth on web||No||Version lacks key features|
|Google Earth VR||No||Version cannot load a KML file|
Google Earth's 3D Buildings layer adds perspective and photorealism to the 3D viewer. However, it also will clutter and obscure your view of the monopoles and antenna panels.
To remove this clutter, toggle the 3D Buildings checkbox from inside the Layers panel at the bottom-left corner of the Google Earth window, below the Places panel.
As the video above shows, disabling 3D Buildings makes the monopoles and antenna panels more visible and much easier to see.
The KML file will open in Google Earth as long as your subscription has not expired. The KML file will continue to work after your subscription has expired, until you close Google Earth.
A KML file becomes "stale" whenever our wireless site database is updated. This happens every few days. Google Earth can open a stale KML file and populate the Places panel with site addresses and technical information, but will not update the 3D viewer. If this happens, sign in to Loxcel Cellular Services and export another KML file, which will include the most recent updates.
The Export KML button can export a site with detailed 3D antenna panels (as shown in the video) or as a simple pushpin. To choose one, click the blue Account button to display the Account & Subscription Info pop-up. Then, make your selection next to Export KML Type. Your subscription quota shows on the line below. A warning will appear if you attempt to export beyond this quota.
For the Simple pushpin export type, when a pushpin is clicked, a window opens to show structure height and channel count. The Simple pushpin export type receives a higher export quota than Detailed 3-D antenna panels, as outlined in Table 1 above.
The following maps show the current distribution of Rogers 5G sites across Canada:
The typical SMS snapshot contain 600,000 records, of which 10 - 15% are duplicates. This month's snapshot contains 982,408 records, the excess coming primarily from 364,058 additional records for Bell Mobility Inc. Most of these additional records are duplicates across every field, which we deleted before continuing with our analysis below.
Compared to last month, April brings
Although not shown below, from March to April, Xplornet increased both
Canada, Australia and New Zealand Cellular Services use data sourced from ISED (Canada), ACMA (Australia) and NZRSM (New Zealand). We manage the quality of this data by closely tracking site and channel counts, and apply corrections as needed.
Starting this month, we're tracking a third metric: Total Bandwidth, the sum of all channel bandwidths with unique transmit frequency and horizontal azimuth. This metric is more accurate than channel counts at measuring the capacity of a wireless network, as it is not bias by the presence of duplicate channel records, common to varying degrees among most carriers.
For an example, Figure 1 shows a large spike of 221,056 Telus channels on March 2018. However, Figure 2 remained flat on March 2018, suggesting this spike represents duplicate channel records in the SMS snapshot — and not additional network capacity.
Table 1 shows how we calculate Total Bandwidth. The six channels were taken verbatim from the SMS snapshot dated 2020-03-04. All have the same 130° horizontal azimuth. All report the same 15MHz bandwidth (Tx High - Tx Low). Channels are grouped by horizontal azimuth. For each group, channels with identical Tx frequency range (# 2, 3, 5 & 6) are discarded, and channels with overlapping Tx frequency range (# 4) have their overlapping portion removed.
The result: 15MHz @ 2142.5 - 2157.5MHz plus 7.5MHz @ 2157.5 - 2165.0MHz equals 22.5MHz total bandwidth.
Table 2 shows the difference in Total Bandwidth if duplicate and overlapping channels, as described above in Table 1, were not removed. Telus' 32.9% is caused by many duplicate channel records in the SMS snapshot.
|Total Bandwidth (GHz)|
|Carrier||With Dups||No Dups||Difference||Percent|
The chaotic graphs in Figures 1 - 6 and the over-reporting of total bandwidth in Table 2 raise this question: is Total Bandwidth a valid metric? To answer, examine Figures 7 - 9, which are the equivalent graphs for Australia, a country of comparable size to Canada:
And Figures 10 - 12 show the same for New Zealand:
Table 3 shows minimal over-reporting of Total Bandwidth for carriers across Australia and New Zealand:
|Total Bandwidth (GHz)|
|Carrier||With Dups||No Dups||Difference||Percent|
Figures 8 and 11 show steady and accurate growth of Total Bandwidth across all carriers in Australia and New Zealand, and Table 3 reports minimal over-reporting — suggesting Total Bandwidth is a relevant and helpful metric to measure the capacity and growth of a nation's wireless network. The chaos pictured in Figures 1 - 6 and over reporting captured in Table 2 suggest data issues with Canada's SMS snapshots, especially for Telus and Freedom.
This table shows the monthly growth of Telstra's 5G network across Australia, starting in May 2019:
|# Telstra 5G Sites|
You can see these 5G sites and more at Australia Cellular Services.
January's SMS snapshot contains 234 Rogers 5G sites of which all but 20 are missing from February's snapshot. The red line in the top-left graph below shows February's snapshot lost 584 Rogers sites, of which 214 are from their 5G network. We doubt Rogers is scaling back their nascent 5G network, so these lost sites are likely an error.
Meanwhile, Telus' channel and site counts remain unchanged from January, and Bell's site count increased from 7,172 to 7,234, along with small increases in channel counts across all cellular spectrum bands.
Rogers recently announced ...
... we're starting our rollout of Canada's first 5G network in downtown Vancouver, Toronto, Ottawa and Montreal
Like we did last year with Telstra in Australia, we are now tracking the growth of Rogers' 5G network across Canada:
Rogers 5G uses Block I (in green below) from the BRS (2500Mhz) Frequency Block Plan:
Rogers 5G uses the lower 20MHz of Block I, leaving its upper 5MHz Restricted Band (RB) unused. Block I uses Time Division Duplexing (TDD) to share one frequency for downlink (DL) and uplink (UL) transmissions. In contrast, paired spectrum blocks (in cyan above) use Frequency Division Duplexing (FDD) which assigns different frequencies to DL / UL (eg. 2640 / 2520MHz).
TDD can operate statically or dynamically. Dynamic TDD adjusts the ratio of time slots assigned to DL and UL, as demand warrants. For example, people at a concert or a sporting event need more UL to push videos & audio recordings to social media; conversely, on the GO train, people need more DL to consume Netflix or podcasts. Dynamic TDD can handle either scenario. We do not know if Rogers 5G uses dynamic or static TDD; but for now — to meet the hype surrounding 5G — we'll assume dynamic TDD.
The growth of Rogers' 5G network is as follows:
|Rogers 5G Sites|
Air Canada Centre
Square One, SW corner
St. Lawrence Market
165 Grange Ave
330 Gerrard St E
York University, west side
The graphs below show cellular site counts for Canada's 20 most populous Census Metropolitan Areas (CMA), ordered by population (most populous first). They provide more detail than the site graphs we have published monthly since Jan 2019.
A CMA comprises one or more adjacent municipalities and townships that are closely integrated in some way. For example, the Toronto CMA consists of the city of Toronto along with Mississauga, Brampton, Markham, Vaughan, Richmond Hill, Oakville, Ajax, Milton and 15 other smaller townships. In contrast, the Halifax CMA consists of the municipality of Halifax by itself.
We can perform custom analysis of Canada's cellular & fixed wireless sites, channels & bandwidth capacity along other census geographies, from as small as the neighborhood to as large as the nation. Please contact us for more details.
Last month, Rogers added 92 sites across Ontario and 583 sites across Quebec, which you can see in the sudden rise of the red line in the top-left graph below. But, most of these sites are not new, but only a correction to the disappearance of a similar number of sites one year earlier.
In short, SMS snapshots from Dec 2018 to Nov 2019 do not accurately represent Rogers coverage across Quebec and Ontario.