Assessing 40 Years of LULC Changes: Need Help with Accuracy Evaluation

[VikramjitKakati]

Hello,

I am conducting a Land Use Land Cover (LULC) change detection study spanning 40 years, with analysis intervals of ten years. I am using ArcGIS Pro for this project, covering the period from 1990 to 2023.

I need guidance on how to perform an accuracy assessment for the Landsat images from 1990, 2000, and 2010. Given the significant changes in land use patterns over time, the current base map information is not applicable for historical accuracy assessment.

Additionally, the available survey maps are not accurate.

Could you please suggest the best approach to proceed with the accuracy assessment?

Thank you,

Prof. Kakati

[nikhilgiseng]

Their are 2 parts to my answer in regards to the challenge that you have described above:

1. Tools & Techniques to perfrom Accuracy assessment in ArcGIS Pro

ArcGIS Pro offers Accuracy Assessment tool that uses a reference dataset to determine the accuracy of a classified result. The values of the reference dataset must match the schema. Reference data can be in the following formats: (Link ArcGIS Pro Accuracy Assessment Tool )

• A raster dataset that is a classified image.
• A polygon feature class or a shapefile. The format of the feature class attribute table must match the training samples. To ensure this, you can create the reference dataset using the Training Samples Manager page to read and write the dataset.
• A point feature class or a shapefile. The format must match the output of the Create Accuracy Assessment Points tool. If you are using an existing file and want to convert it to the appropriate format, use the Create Accuracy Assessment Points tool.

2. How to get the reference data forAccuracy assessment of old Imagery for year 1990, 2000 & 2010

This is a fairly common problem in LULC mapping. It is always possible to validate a recent map, since the land cover/land use can be checked for accuracy through field visit. However, for past LULC maps, all you can do is to use the maximum trusted data source you might have at your disposal which can be checked against your past remotely sensed maps. It can either be aerial photographs, reference LULC maps, ancillary data.

A common approach people often use is sampling from Google Earth satellite images. They are free but you cannot download them, but rather just browse them. However, since Google is providing a timeline feature for free, aiming at your area of interest, you can freely access old picture and sample data for each class of interest. Lots of published papers use this approach for validation when ancillary sources/aerial photographs or reference datasets for the past years are not available, yet it has proven to be a realiable method (see http://adsabs.harvard.edu/abs/2015SPIE.9644E..1NR ).

Also find attached to this answer the FAO Map Accuracy Assessment and Area Estimation practical guide advocating the same method for sampling data from Google Earth for map validation, along with formulas and a link to an interesting plugin developed to this end, CollectEarth (http://www.openforis.org/tools/collect-earth.html).

Hope this helps.