EX2 - Land use and Land Cover/ Accuracy Assessment

Question 1

Land use:

what’s the difference? Which is easier to map using remote sensing?

Answer 1

Use of land surface by humans; economic context
•Agricultural
•Residential
•Commercial

Question 2

Why map land use and land cover

Answer 2

Identifying agricultural practices and environmental impacts

Question 3

Large scale vs. small scale LULC maps

Answer 3

Small scale LCLU map = large area = less detail
used for regional planning, where loss of resolution and detail and resulting integration and simplification of
information may actually be advantageous.

Large scale LCLU map = small area = more detail
Used by local governments who need very detailed information for local planning

Question 4

Anderson classification scheme (no need to memorize the categories; just remember the name “Anderson” and salient points about this classification)

Answer 4

1) nominal
2) mutually exclusive classification scheme,
3) most wisely used classification scheme

4) lends itself to use with images of varied scales and resolutions

Level I categories
broad scale, coarse resolution imagery obtained from broad scale satellite imagery or high altitude aerial photography

More detailed Level III and Level IV categories can be defined by analysts

Question 5

Some challenges of land use land cover mapping

Answer 5

1) Issue of multiple use: A forested area may simultaneously serve as a source of timber, a recreational area for hunters and hikers, and as a source of runoff that supplies water for an urban region
2) Decide minimal threshold size for parcels to be represented on the final map
3) The issue of mixed categories. At small scales, there may be unavoidable inclusion of other categories. How to resolve these?

Question 6

Historical land cover interpretation for environmental analysis

Answer 6

Archives of aerial photography in US go back to 1930s

A valuable resource to establish historical pattern of LULC change

Help understand sequence of events and assessment of risks to environment and nearby populations

Question 7

General purpose LULC classification

Answer 7

classifications that serves many purposes, but not specifically tailored for any specific application

Most common / widely used
Example: Anderson classification

Question 8

SPOT vs. AVHRR vs. MODIS

Answer 8

different variable for LULC maps:
coverage
scale
resolution

Question 9

Precision

Answer 9

precision describes the variation you see when you measure the same part repeatedly with the same device

Sharpness (or certainty) of a measurement.
•Higher variability leads to poor precision
•Low variability creates high precision
•It is therefore closeness of a set of measurements to one another

High precision = low variability of estimates

Question 10

Accuracy assessment

Measurement of map accuracy

Answer 10

involves comparing 2 sources of information:

A reference map / reference data based on a different source of information and assumed to be accurate

If reference data itself is erroneous, then accuracy assessment is erroneous

The LULC map and reference data must be:

• co-registered
• use same or compatible classification schemes
• based on information collected during the same time of the year.

Question 11

non-site specific accuracy

Answer 11

Non-site specific accuracy:

• does not consider agreement b/w LULC map and reference data at specific locations
• only the overall percentages for each category on the two maps. Can be misleading

Question 12

Error matrix / confusion matrix

Answer 12

Error matrix (aka confusion matrix) identifies overall errors for each category as well as misclassifications between categories by category

SEE PP

Question 13

Example of how an error matrix is compiled

Answer 13

• 20 sample points
• 1 reference data
• 2 classification schemes, resulting in 2 LULC maps:
  Map 1
  Map 2
• Classification types:
  1 = Non-forest
  2 = Forest
• results in a table with 4 colums and 20 rows
• a count then proceeds
• a new table is made with classification symbols as columns and rows
• Reference data is columns, maps data is rows

Question 14

Errors of commission:

error/row total

Answer 14

Assignment of “non-forest” pixels on the ground to “forest” on the map.

Here, the analyst has actively committed an error by assigning a region of forest to a wrong category

Question 15

User’s accuracy and Producer’s Accuracy

Answer 15

User’s accuracy:
# of correct cat'd map pixels/ row total for CAT

Producer’s accuracy
# of correct cat'd map pixels/ column total for CAT

Question 16

Overall Accuracy

Answer 16

Overall accuracy is the overall proportion of correctly classified pixels for all classification types

add diagonal correct values/ absolute total

Question 17

Overall Accuracy without the error matrix

Answer 17

Overall accuracy is a simple measure of accuracy

By itself, it suggests the relative effectiveness of a classification

However, without the error matrix, it cannot provide convincing evidence of the classification’s accuracy

Question 18

Quantitative Assessment of Error Matrix

Answer 18

Error matrix reveals the overall nature of errors present.

objective assessment needed if we want to know if two maps (LULC map and reference map) are in agreement (statistical agreement)

should test if agreement between two maps is merely chance or a truly successful classification

kappa = Observed - Expected/ 1 - expected

Question 19

Land Cover:

Answer 19

The visible features of the Earth’s surface
•Wetlands
•Barren
•Urban
•Grassland

Question 20

Special purpose LULC classification

Answer 20

Designed to address a specific classification issue, with no attempt to provide comprehensive scope

Specialized alternatives to general purpose strategy or levels III and IV within a hierarchical system such as
the Anderson classification

Example: Wetland classification of Cowardin(1979)

Question 21

Accuracy:

Answer 21

Correctness / Exactness

accuracy describes the difference between the measurement and the part’s actual value,

Question 22

Site specific accuracy:

Answer 22

based on detailed assessment of agreement
b/w the LULC map and reference data at specific
locations which are selected using different sampling
techniques

Question 23

Errors of omission:

Error/ column total

Answer 23

Assignment of “forest” pixels on the ground to “non-forest” on the map.

That is, an area of “real” forest on the ground has been omitted from the map

Question 24

Quantitative Assessment of Error Matrix

Kappa

Answer 24

К (Kappa) is a measure of the difference between the observed agreement between two maps (as reported by the diagonal entries in the error matrix) and the agreement that might be attained solely by chance matching of the two maps.

К attempts to provide a measure of agreement that is
adjusted for chance agreement.

kappa = Observed - Expected/ 1 - expected

Observed = overall accuracy reported in the error matrix (sum of the diagonal entries divided by the total number of pixels)

Expected = an estimate of the contribution of chance
agreement to the observed overall accuracy. This is
calculated using the row and column totals