¿Cómo podemos ayudarte?
Zoning methods: differences between quantile, cluster and manual values.
The “index zones” report allows one to obtain a map of “homogeneous zones” from the green index or another index of a selected date.
he system takes the index values of each pixel and groups them according to the criteria (quantile, cluster or manual values) and number of classes defined by the user. As a result, a polygon map is obtained that can be used for multiple purposes such as: planning re-fertilization tasks and variable herbicide application, evaluating drifts, estimating future yield with field validation, planning harvesting according to crop maturity, etc.
Depending on the type of analysis to be performed and the further use of the resulting map, the user can select between three different clustering methods, as shown in the map comparative below:
Quantiles | Clusters | Manual Values |
Each method has different characteristics and purposes and it is recommended to explore them in order to find the appropriate classification for each case and utility.
Quantile
The quantile method allows to create zones based on a criteria of regular frequency intervals (statistical method). The main feature of this method is that it defines areas of similar area.
In other words, quantiles are those values of the variable (NDVI in this case) that, ordered from lowest to highest, divide the distribution into parts, so that each of them contains an equal portion of these values.
For example; if all values in a distribution are divided into 4 parts, each part contains 25% of the data. As follows: Q1 is the first quartile which leaves 25% of the data to its left; Q2 is the second quartile which leaves 50% of the data to its left; and Q3 is the third quartile which leaves 75% of the data to its left.
More information about Quantiles
It allows us to identify macro-areas, being of greater importance to group zones and thus be able to make macro decisions, for example, the logistics of Sunflower Harvest.
Quantiles – Large chain – (2019-01-28) |
Low | Medium | High |
34.73 Ha. | 32.00 Ha. | 37.51 Ha. |
33.32 % | 30.70 % | 35.98 % |
0.53 | 0.79 | 0.82 |
Cluster
Cluster Analysis, also known as Conglomerates Analysis, is a multivariate statistical technique that seeks to group elements (or variables) trying to achieve the utmost homogeneity in each group and the greatest difference between these groups. It is a multivariate statistical method of automatic data classification, in this case NDVI.
The Cluster method creates zones based on a grouping method by similarity (K Means), that is, it takes into account the creation of groups based on similarity, their complementarity or distance, among other data.
Some possible uses are:
- Re-fertilization
- Evaluating drifts
- Variable herbicide application
- Yield estimation with field validation, for harvesting by macro-environments.
Clusters – Large chain – (2019-01-28) |
Low | Medium | High |
10.13 Ha. | 12.74 Ha. | 81.37 Ha. |
9.72 % | 12.22 % | 78.06 % |
0.29 | 0.57 | 0.79 |
Manual Values
The Manual Values method allows us to choose the NDVI value that will serve as a cut-off value to separate the zones, this tool allows us to manually choose the NDVI value that will separate the classes.
It allows us to define areas of interest prior to a field validation, for example:
- Problem areas (Fire-Flood).
- Weed patches, for weed mapping.
Manual values – Large chain – (2019-01-28) |
Low | Medium | High |
3.64 Ha. | 6.51 Ha. | 94.09 Ha. |
3.50 % | 6.25 % | 90.26 % |
0.18 | 0.28 | 0.77 |