User Tools

Site Tools


Report a bug, broken link, or incorrect content

Inventorying Riparian Areas Using Nested Area Frame Sampling

Quick Look

A method to perform a stratified sample inventory using remote sensing imagery.

(Ratings are given on a 1-5 scale. Click on any rating for an explanation)

Method Overview

This method utilizes Nested Area Frame Sampling (NAFS) (Koeln and Kollasch 2000) to inventory large areas or regions where fieldwork is too expensive or inaccessible. NAFS is a multistage stratified area sampling procedure that uses several different resolutions of imagery. The NAFS approach stratifies the area of interest using coarse resolution imagery, samples the strata using moderate resolution imagery, and refines the sampling using high-resolution imagery. The use of stratified image resolution is to simplify processing and reduce image acquisition costs. Higher resolution imagery can be substituted at any point. This approach is highly scalable and field-based sampling can be nested into the structure.

The NAFS approach involves four steps:

  1. Creating the strata
  2. Establishing the location, number, and size of primary sampling units (PSUs)
  3. Mapping riparian areas within the PSUs
  4. Calculating riparian estimates

Figure 1: An example NAFS process for riparian inventory (from Ruefenacht and others, 2005)

Creating the Strata

The study area is stratified based on predominate land cover. Stratification is performed to improve sample location and number, and to provide a basis for calculating estimates for the entire study area (Koeln and Kollasch, 2000).

Figure 2: NAFS schematic showing stratification and random placement of primary and secondary sampling units (from Ruefenacht and others, 2005)

Stratification can be performed using a variety of methods. The following procedure has been implemented in previous investigations and is provided as a general guide:

  1. Using coarse imagery, divide the study area into segments based on normalized difference vegetation index (NDVI) (for more information on segmentation see the CART Classification Method)
  2. Simplify the segments into strata using unsupervised classification (any classification method will work)
  3. Refine the segments based on elevation using a digital elevation model (DEM) resampled to match the imagery’s resolution
  4. Refine or simplify the strata based on additional data such as regional landcover data and the National Hydrology Dataset

This process produced the following six stratum for the state of Wyoming:

  1. Lowland rangelands and bare soils
  2. Upland rangelands and forest lands
  3. Irrigated agriculture, high order riparian, and ponderosa forest
  4. Forest lands
  5. High alpine and sparse vegetation
  6. Lakes and reservoirs

Figure 3: An example stratification (from Ruefenacht and others, 2005)

Establishing the Location, Number, and Size of PSUs

In this multistage sample design, PSUs are randomly placed within each stratum. However, prior to doing so, the appropriate number and size of the PSUs must be determined for each strata The PSU size should capture the spectral variability within the strata while remaining as small as possible. The optimal size is determined by randomly selecting a number of different areas at a variety of different sizes. Using moderate resolution imagery, the spectral information for each of these areas is extracted from the imagery, and the variance (shown here as standard error) is plotted against the sampling unit area. The PSU size is selected based on qualitative evaluation of an acceptable balance of variability and sampled areas.

Figure 4: Example PSU sampling size selection chart. In this case, a PSU size of 1000 ha was selected (from Ruefenacht and others, 2005)

The number of PSUs is established using a similar methodology. Within each stratum, select a broad range of numbers of PSUs that are the size determined in the previous step. Using moderate resolution imagery, the spectral information for each of these areas is extracted from the imagery and the variance (shown here as standard error) is plotted against number of sampling units. The number of PSUs is selected based on the minimizing variability while keeping the number to a minimum. The selection is based on a qualitative assessment of the point that strikes the best balance between these two criteria.

Once the size and number of PSUs is established, the PSUs are randomly located within the strata. Random points are selected, and then the PSUs are grown to the appropriate size using the ERADS Region Grow utility. This utility produces a PSU boundary that corresponds to the pixel edges of the imagery. If secondary sampling units (SSUs) are appropriate for the investigation, they can be located and placed within the PSUs by repeating the steps above on a finer scale. Field data can be incorporated into the study as SSUs.

Figure 5: Example number of PSUs selection. In this case, the optimal number of PSUs was determined to be 40 (from Ruefenacht and others, 2005)

Mapping Riparian Areas Within the PSUs

A valley bottom analysis is performed on each of the PSUs to identify potential riparian areas using high-resolution imagery. This analysis can be limited to the geomorphic methods described on the Valley Bottom Mapping page; however, incorporation of the methods described on the Riparian Vegetation Mapping page will improve the results. This process divides each PSU into a minimum of two classes, riparian and non-riparian. Additional classes based on vegetation type can be added to the extent that imagery and training data allow them to be resolved. The area of the riparian portions is calculated and used to estimate the total riparian area in each stratum and in the entire study area.

Calculating Riparian Estimates

The size (area) of the riparian component is calculated for the PSUs in each stratum to derive an estimate of the mean area and standard error. The total areas and standard errors are divided by the total PSU area and multiplied by the total stratum areas to derive riparian estimates for each stratum and the entire study area.

Data Inputs

  • Imagery at several different resolutions
  • Digital elevation models
  • Training data for vegetation mapping

Method Products

The method produces an inventory of riparian areas based on stratified sample selection.

Riparian Application Example

Ruefenacht, B.; Guay, A.; Finco, M.; Brewer, C.K.; Manning, M. 2005. Developing an image-based riparian inventory using a multi-stage sample: Phase I Report. Rep. No. RSAC-4022-RPT1. Salt Lake City, UT: U.S. Department of Agriculture Forest Service, Remote Sensing Applications Center. 17p.

Finco, M.; Megown, K.; Ruefenacht, B.; Manning, M.; Brewer, K. 2008. Wyoming riparian vegetation inventory using remote sensing and other geospatial data: Phase II—high resolution imagery sources. Rep. No. RSAC-4022-RPT2. Salt Lake City, UT: U.S. Department of Agriculture Forest Service, Remote Sensing Applications Center. 27 p.

Technical References

Koeln, G.T.; Kollasch, R.P. 2000. Crop area assessments using low, moderate and high resolution imagery: a GeoTools approach. Earth Satellite Corporation, Rockville, MD. 12 p.


  • Sampling density must be optimized to minimize error and time investment
  • The quality of the land cover strata has a significant impact on the overall accuracy
  • Valley bottoms derived from 10 m DEMs produce better results but are more time consuming to produce than those derived from 30 m DEMs.

Software/Hardware Requirements

There are no available software packages to implement this method. The majority of the steps discussed here can be performed using ERDAS Imagine and ESRI’s ArcGIS. Additional software and hardware requirements will depend on the valley bottom mapping method and vegetation mapping method.

Additional Information

Web Search Results



You must have an account and be logged in to post or reply to the discussion topics below. Click here to login or register for the site.

remote_sensing_methods/inventorying_riparian_areas_using_nested_area_frame_sampling.txt · Last modified: 2012/08/27 09:47 by calbury