Weed-Mapping

 
 
Subscribe to our free E-Newsletter, "Agri-News" (formerly RTW This Week)Agri-News
This Week
 
 
 
 Spatial distribution of weeds is characterized using weed density data collected at locations in a field. Different weed-mapping techniques may be used to build different maps of weed density; however the technique chosen must be appropriate to meet the objective(s). This study shows five different methods of mapping the same weed data and discusses merits and objectives of each technique.

Objectives for weed density maps vary and may include:

  1. Showing large-scale trends to provide a visual of a weed distribution on a field basis.
  2. Integrating multiple sources of data by merging a weed density with elevation or soil nutrient data to construct a spatial model.
  3. Judging the influence of secondary data to reveal spatial trends at different scales.
  4. Quantifying weed density over an entire field for a global average.
  5. Visualizing variability to understand local variations.
  6. Assessing uncertainty to allow an expert to evaluate the risk involved in any decision making.
  7. Deciding on locally variable treatment procedures such as variable rate herbicide application.
Weed density was sampled in a 87 ac field near Saskatoon, Saskatchewan in 1995. The total number of weeds for all species was counted and mapped using hand contouring, gridding with inverse distance (IDW) or nearest neighbor, kriging, and simulation to illustrate the appropriateness of each technique.

Contour maps of weeds are a valuable tool for displaying and representing weed densities in Figure 1. This hand contoured map show a series of lines drawn to represent constant data values which gives a smooth representation of the data for understanding large scale trends. The map can mask finer details due to smoothing and is subjective, depending on a user’s experience and judgement for integration of expert knowledge.

Contouring can also be accomplished using commercial software packages, which offer flexibility for data analysis and interpolation. Computer contouring begins by gridding the variable using techniques like splines, kriging or triangulation. Next, the contours are traced through gridded values to reveal weed density trends but they do not replicate short scale variability. Computer contouring is subjective and tends to rely heavily on a user’s choice of settings that can mask anomalies and smooth erratic phenomena (data not shown).

Weed density can be gridded by inverse distance type methods, which weight data according to a mathematical relationship that accounts for distance to nearby data. Two methods that assign all or a portion of the weight to the nearest data include nearest neighbor or inverse distance weighting (IDW), respectively. These methods are illustrated in Figure 2. IDW has a tendency to generate "bull's-eye" patterns of concentric contours around the weed density sampling location. When data are sparely sampled, nearest neighbor is ineffective in extrapolating the data where there are none (Figure 2).

Kriging and simulation are mapping techniques that require a variogram in deriving weights for mapping. Directional variograms for the weed distribution are shown in Figure 3. The maximum direction of continuity has an effective range of approximately 150 m (black line) while the minimum direction of continuity has a range of 80 m (red line).

Kriging relies on least squares to produce unbiased estimates that minimize the difference squared between the estimated and true value. Figure 3 illustrates the appropriateness of kriging for visualizing trends. Weed density sampling locations are widely spaced, hence the smoothing of the map. The northeast corner of the field had one high weed density value, which resulted in the red area on the map.

Variability is expected in biological data and unique or smooth maps are unrealistic. Simulation captures the uncertainty of weed densities and Figure 4 provides 3 alternative numerical models or realizations that mimic this uncertainty. An average of 101 realizations of this weed distribution is shown in the lower right corner in Figure 4. This realization is very similar to the kriged map in Figure 4. These different realizations are used for risk-quantified decision making in herbicide application rate maps.

Table 1. Suitability of mapping techniques in meeting different objectives.
Large scale trends
Integrate Secondary data
Influence of secondary data
Quantify weed density
Visualize Variability
Assess Uncertainty
Decision
making
Hand Contouring
Green
Green
Yellow
Green
Red
Red
Yellow
Computer Contouring
Green
Yellow
Red
Green
Red
Red
Red
IDW
Yellow
Yellow
Red
Green
Red
Red
Red
Kriging
Yellow
Green
Green
Green
Red
Red
Yellow
Simulation
Red
Green
Green
Yellow
Green
Green
Green
Green-appropriate;
Yellow-could be used but not ideal;
Red-inappropriate


Rarely are there enough data to provide reliable statistics; however, geostatistical techniques allow characterization of spatial variability and can be used to facilitate decision making. Table 1 indicates the suitability of mapping techniques in meeting those objectives. Cost, timeliness and convenience of measuring weed data may limit what is measured for analysis. Secondary information such as elevation from remote sensing data may mitigate these issues and help infer spatial statistics since they are more densely sampled. Kriging and simulation allow the incorporation of secondary data where local knowledge is limited.

*L. Hall, Alberta Agriculture; A. G. Thomas, Agriculture and Agri-Food Canada; K. Norrena, C. Deutsch, University of Alberta.


Figure 1. Mapping technique using hand contouring. Weed density is indicated on the contour lines.


Figure 2. Mapping technique using inverse distance weighting (IDW) (left). Mapping technique using nearest neighbor (right). Weed density is indicated by the colors with red being high and blue low.


Figure 3. Variogram model used in kriging and simulation mapping techniques (left). The solid line represents the fitted model whereas the dashed lines with the dots is the experimental model. Weed densities mapped using kriging (right). Weed density is indicated by the colors with red being high and blue low.


Figure 4. Mapping technique using simulation. These maps represent 3 different simulated maps along with an average of 101 maps in the lower right corner. Weed density is indicated by the colors with red being high and blue low.

 
 
 
 
Share via AddThis.com
For more information about the content of this document, contact Roger Bryan.
This document is maintained by Laura Thygesen.
This information published to the web on January 15, 2003.
Last Reviewed/Revised on January 22, 2018.