Closed cjacobscrisioni closed 10 months ago
Same in NZL, northern island. According to the current code, these centres are the biggest in... (Google maps estimates to larger settlements between parentheses)
Hamilton 100m (80m to Auckland) Tauranga 120m (83m to Hamilton) Rotorua 80m (53m to Tauranga)
KPH statistics from initial working set seem to capture max speeds accurately?
Speed in Australia is max 130 km/h (https://onroad.com.au/Blog/Speed-Limits-in-Australia)
ValuesMetric meter per 3.6second ValuesType Float64 Count 28493355
Formal Range km_hr [-1.79769313E 308, 1.79769313E 308) Metric Units meter per 3.6second Maximum 130 Minimum 7 Sum 1.35359602E 09 Average 47.5056737 Variance 715.492665 StdDev 26.7486951
Reverse engineering speeds from the impedance section (which includes a penalty for crossroads)
ValuesType Float32 Count 28493355
Formal Range [-3.40282E 38, 3.40282E 38) Maximum 130.000015 Minimum 1.72358705E-05 Sum 1.08357872E 09 Average 38.0291728 Variance 632.99575 StdDev 25.1594072
Conclusion: on average, speeds from the impedance attribute are 20% lower. This can explain a part of the overestimation.
Reverse engineering from the minutes attribute (which is directly imposed from the KPH attribute, with a small value if the conversion throws a null.
ValuesType Float32 Count 28493355
Formal Range [-3.40282E 38, 3.40282E 38) Maximum 130.000015 Minimum 6.99999905 Sum 1.35359602E 09 Average 47.5056737 Variance 715.492665 StdDev 26.7486951
Reverse engineering from the impedance attribute after junction penalties are switched off
ValuesType Float32 Count 28493355
Formal Range [-3.40282E 38, 3.40282E 38) Maximum 130.000015 Minimum 6.99999905 Sum 1.35359603E 09 Average 47.5056737 Variance 715.492665 StdDev 26.7486951
Conclusion: junction penalties are plausibly too large cf Google Maps. I will take out the junction penalties for now.
New results without junction penalties (Google maps between parentheses):
Hamilton 90m (80m to Auckland) Tauranga 95m (83m to Hamilton) Rotorua 65m (53m to Tauranga)
... so still an overestimation, but not as extreme as previously. Given that the speeds seem realistic I have three explanations for the remaining difference:
This calls for a broader evaluation of networks. Closed for now.
Case in point: centroids in Tauranga, Google maps vs current version
I ran the settlement hierarchization algorithms for Australia and Oceania (/Analyses/Obtain_Settlement_Hierarchies/Australia_Oceania) but the results for AUS seem to overestimate travel times between settlements (ie, two hours according to our analysis, ~1.5h according to Google maps).
Potentially due to speeds being coded in MPH while we assume KPH. However, Australia uses the metric system (since 1966).
First check: is the overestimation similar for NZL?