Greetings folks - I recently answered a question regarding the accuracy of LiDAR and how to explain it to folks. I wrote the following and thought that this might be beneficial.

**Greetings,**

LiDAR data accuracy is measured using the National Standard for Spatial Data Accuracy (NSSDA). The traditional method of determining accuracy was the National Map Accuracy Standard (NMAS). One of the primary differences between these two methods is that the NMAS was developed for paper maps at a particular scale while the NSSDA was developed to measure accuracy at a ground scale of 1:1.

For the LiDAR data the contract specifications listed a 15 cm Root mean square error (RMSE) as the maximum error as determined by independent validation. This is where the local surveyors and MnDOT came in and captured some points and we tested the vertical on those points against the vertical on the LiDAR surface.

RMSE is calculated as the sqaure root of the average of the set of squared difference between the survey and the lidar surface. This is a statistical measure that quantifies the level of error in the data.

Using NSSDA accuracy is reported in ground distances at the 95% confidence interval. This means that 95% of the vertical values in the dataset will have an error with respect to true ground position that is equal to or smaller than the reported accuracy value.

So, accuracy is defined as: Accuracy_z = 1.96 * RMSEz

In the case of Block 2 the vertical error tested out to 13.7 cm or 5.39”. Accuracy at the 95% confidence interval is then calculated as 5.39” * 1.96 = 10.56”. So the accuracy of the data set is just under a foot. In order to compare this to the NMAS (which is what most surveyors/engineers are used to) you have to multiply the value by 2 because NMAS states that the maximum allowed vertical tolerance to be one half the contour interval.

So, in the case of this data the data supports the generation of contours at the 1’8” interval at a 95% confidence interval. But for all practical purposes you can let folks know that they can create 2’ contours with a high degree of confidence.

But it’s also important to know/understand that LiDAR performs better in some environments than others. The link below points to the validation report for Block 2. If you take a look at the graph shown in this PDF you can see that the LiDAR did very well in Open, Forested and Urban environments – good enough to create 1’ contours at the 95% confidence interval. But LiDAR does not do very well in areas of heavy grass, brush, cattails etc and that is shown by the higher values of error indicated in those environments. The lesson to take away is that knowing what environments you are working in will help you understand where have to invest a little in additional survey work and where you may not. Cattails have shown to be the place where LiDAR suffers the most.

In the case of tall grass the RMSEz is 22cm - that's 8.66". Accuracy at 95% interval is then (8.66 * 1.96) * 2 or 33" - nearly three feet.

I hope this helps your understanding of accuracy and how it’s measured and evaluated for the lidar data.

tim

[url=ftp://ftp.lmic.state.mn.us/pub/data/elevation/lidar/projects/central_lakes/block_2/central_lakes_block_2_validation_report.pdf][/url]

LiDAR data accuracy is measured using the National Standard for Spatial Data Accuracy (NSSDA). The traditional method of determining accuracy was the National Map Accuracy Standard (NMAS). One of the primary differences between these two methods is that the NMAS was developed for paper maps at a particular scale while the NSSDA was developed to measure accuracy at a ground scale of 1:1.

For the LiDAR data the contract specifications listed a 15 cm Root mean square error (RMSE) as the maximum error as determined by independent validation. This is where the local surveyors and MnDOT came in and captured some points and we tested the vertical on those points against the vertical on the LiDAR surface.

RMSE is calculated as the sqaure root of the average of the set of squared difference between the survey and the lidar surface. This is a statistical measure that quantifies the level of error in the data.

Using NSSDA accuracy is reported in ground distances at the 95% confidence interval. This means that 95% of the vertical values in the dataset will have an error with respect to true ground position that is equal to or smaller than the reported accuracy value.

So, accuracy is defined as: Accuracy_z = 1.96 * RMSEz

In the case of Block 2 the vertical error tested out to 13.7 cm or 5.39”. Accuracy at the 95% confidence interval is then calculated as 5.39” * 1.96 = 10.56”. So the accuracy of the data set is just under a foot. In order to compare this to the NMAS (which is what most surveyors/engineers are used to) you have to multiply the value by 2 because NMAS states that the maximum allowed vertical tolerance to be one half the contour interval.

So, in the case of this data the data supports the generation of contours at the 1’8” interval at a 95% confidence interval. But for all practical purposes you can let folks know that they can create 2’ contours with a high degree of confidence.

But it’s also important to know/understand that LiDAR performs better in some environments than others. The link below points to the validation report for Block 2. If you take a look at the graph shown in this PDF you can see that the LiDAR did very well in Open, Forested and Urban environments – good enough to create 1’ contours at the 95% confidence interval. But LiDAR does not do very well in areas of heavy grass, brush, cattails etc and that is shown by the higher values of error indicated in those environments. The lesson to take away is that knowing what environments you are working in will help you understand where have to invest a little in additional survey work and where you may not. Cattails have shown to be the place where LiDAR suffers the most.

In the case of tall grass the RMSEz is 22cm - that's 8.66". Accuracy at 95% interval is then (8.66 * 1.96) * 2 or 33" - nearly three feet.

I hope this helps your understanding of accuracy and how it’s measured and evaluated for the lidar data.

tim

[url=ftp://ftp.lmic.state.mn.us/pub/data/elevation/lidar/projects/central_lakes/block_2/central_lakes_block_2_validation_report.pdf][/url]

Powered by mwForum 2.27.3 © 1999-2012 Markus Wichitill