Stratification is the process of dividing an area to be monitored up into units to increase the efficiency of monitoring. There are statistical, logistic, and management reasons for stratifying a landscape.


Because rangelands are among the most diverse ecosystems in the world, it is impossible to design a monitoring system that perfectly reflects changes in all landscape units. However, the accuracy and precision of any monitoring system can be improved by carefully dividing the area into relatively uniform units for monitoring. These monitoring units are called strata. Strata are areas located in a particular part of the landscape (e.g., flood basin or hill summit), within which vegetation, soil type, management and current status are relatively similar. All areas classified by the same stratum are expected to respond similarly to changes in management and to catastrophic disturbances, such as a combination of drought and fire. Multiple monitoring units of the same type (e.g., hill backslope) often repeat across the landscape, geographically separated from one another by other monitoring units. Differences in land uses, disturbance, or management can also be used to divide an area into strata (e.g., grazed vs. ungrazed).

Why Stratify?

From a statistical standpoint, you should consider stratifying when you have reason to believe that either (a) the average value of key indicators is not consistent across different definable units or areas (e.g. ecological sites) in your study area, or (b) within-stratum variability in these indicators is different among these units. However, as discussed above, there may be other reasons for stratifying.

  1. Reduce the number of plots that need to be monitored – The primary reason for stratifying a sample is to increase the efficiency (i.e., power) of the sampling by grouping similar sampling units together. Because variability is minimized within strata, stratification improves the precision of estimates and is a more efficient sampling technique than simple random selection.The number of locations sampled within each stratum can be different and can be related to the within-stratum variability. For example, strata that are very homogeneous may get only a few samples, whereas more heterogeneous strata would have more samples taken from them.
  2. Ensure that small areas get monitored – Some areas of interest for monitoring (e.g., riparian areas) may only occupy a small portion of the overall study area and could be missed (or under-sampled) without stratification. Assigning these areas into separate strata from the rest of the study area allows you to make sure that these areas are adequately sampled in your monitoring program.
  3. Focus monitoring effort on priority areas – Additionally, some portions of a study area may be more sensitive to disturbance or expected to show change earlier than other areas. Through stratification you can assign extra monitoring effort to these areas – increasing the precision of your indicator estimates and your ability to detect changes. Alternatively, stratification can allow us to deemphasize or omit very stable areas from a monitoring program. For example, highly stable types of monitoring units (such as bedrock) might not be included in a monitoring program if the primary objective is to monitor for degradation risk or recovery.
  4. Aid in interpretation of results – Strata can be defined to include types of land that respond similarly to disturbance or management (e.g., ecological sites). When this is done, estimates of indicator values or change can be interpreted the same across all plots in stratum.
  5. Report on units meaningful to management – Often, it is necessary to be able to report monitoring results for several different management units within a study area. These management unit boundaries can become part of the stratum definition to help ensure adequate sampling within each unit.

The Case for Simple Stratifications

Stratification is a powerful way to deal with variability in monitoring data that comes from known landscape properties like differences in soils, climate, or management. It is tempting to include many of the different features you know would influence the indicator values measured in a study area. However, for several reasons, it is best to err on the side of simplicity when defining strata for monitoring.

First, in statistical sampling, an adequate number of locations must be sampled in each stratum. Therefore, the more factors you add to your strata, the larger the sample sizes become to meet your monitoring objectives. For example, a management area with 10 different soil types, 2 aspects (north- and south-facing) and 2 management regimes could have as many as 40 strata. If the goal was to have 5 samples per stratum, then 200 sample locations would be required. If, however, it is possible to combine soils and aspects into, say 4 groups based on responses to management, then the number of samples required drops by 80% (40 samples required).

The second argument for simple stratifications is that it is often difficult to know ahead of time which landscape processes and patterns will have a significant influence on your monitoring results. In this case it is better to develop strata that ensure monitoring across a wide range of different conditions and to rely on analyses of the monitoring data and statistical techniques like post-stratification for interpretation.

Stratification and Data Analysis

The choice to use stratification in a monitoring design affects how the data are analyzed.

Comments are closed.