March 2002

Introduction to the Nuts and Bolts of Directed Crop Scouting With Remotely Sensed Data

Keywords: remote sensing, directed sampling, crop scouting

Update on Directed Crop Scouting: Issue 1 of 2

G.K. Blumhoff, SSMC Information Systems Manager

Most commercially successful site-specific crop management has focused on preplant decisions. Yet many of the most important crop management decisions must be made in-season, including weed and insect control. One of the key factors that has limited the growth of site-specific management for in-season pest control is the difficulty and cost of obtaining timely information on spatial patterns of pest-crop conditions. In many cases it has been cheaper and more profitable to treat the whole field than to find out where pests are located. Profitability and environmental stewardship could be improved if less expensive and more effective methods were available to gather spatial data on pests. This article outlines the development of a “How To Guide” for the use of remote sensing to guide scouting, reducing the cost and increasing the effectiveness of scouting.

Preplant decisions have had the advantage of allowing time for data processing and analysis, and of making use of relatively low cost data. The best example is the use of yield monitor data in making hybrid and variety choices for the next season. In most cases the producer and his advisor have at least a couple of months (e.g. October to December) to analyze the yield data before seed orders must be placed. Once a producer has a yield monitor on his combine and a global positioning system (GPS), the marginal cost of collecting data is cents per acre. In contrast, in-season pest management depends on scouting which is usually several dollars per acre and which must be repeated at regular intervals, and it requires timely decisions. The delay of one or two days may mean the difference between a minor problem and a disaster.

Pest sampling is essential for the accurate delineation of spatial patterns in crop conditions. Yet, tradeoffs exist between increases in the density of sampling, labor and the expense to collect and analyze samples, and the resulting accuracy of maps to represent the spatial patterns in crop conditions observed. The accuracy of these spatial patterns depends heavily upon the initial sampling strategies that are used to map pest/crop conditions. The primary sampling techniques for crop scouting of pests and/or crop conditions include random, grided, or targeted sampling.

Recent availability of high resolution aerial and satellite remote sensing technology has allowed detailed information at a 1m to 10m spatial resolution to be collected across a field. This information from bare soil or vegetation allows spatial patterns based on reflectance to be accurately identified. However, note that this technology does not directly indicate the quantitative levels of yield influencing factors to be determined. For this reason, it is important to identify whether these factors have high correlation to crop field conditions measured on the ground (targeted sampling). This can be determined by sampling field conditions and comparing to the aerial or satellite data acquired. If the imagery and ground data are related to one another, they can be pooled to produce a continuous map of spatial variability related to the sampled crop field conditions.

In addition to testing the feasibility of existing sampling methods (random sampling or zigzag scouting), the use of targeted sampling based on secondary information such as yield maps, soil types, and/or remote sensing has been proposed. Remote sensing offers the potential for identifying small-scale spatial patterns in crop conditions, crop yield across the field, and optimizing crop scouting methods to quantify those patterns.

A step-by-step or "cook-book" procedure was developed to incorporate remotely sensed satellite and/or aerial data for use in directed crop scouting. Figure 1 shows a flow chart of the directed method data processing and analysis. This summary outlines the steps to consider from the time image data is received to diagnostic sampling in the field. It is important to note that some data providers perform image subsetting and rectification prior to sending the data. The image classification step can be valuable when several differences are occurring in the original image data. Sometimes classifying the data into 3 - 6 classes helps when trying to figure out what major factors are causing poor crop growth conditions in certain areas of the field. Also, these classes may in turn help during Variable Rate Map production by providing 2 or 3 classes for herbicide application rate decisions. Lastly, once GIS (Geographic Information System) layers have been produced, including field boundaries and sample locations, the data can be transferred to a handheld computer linked to GPS.

Figure 2 shows boundary layers, scouting map from remote sensing, and the original remotely sensed data. The original raw data is an IKONOS satellite image acquired on 24 May 2000. The raw data acquired is displayed as a single band, panchromatic (black/white) image at 1-meter resolution. Notice the yellow points in the classified scouting maps. These are the directed sample locations to be collected in the field. Also, Figure 3 shows an example software display on a PDA (handheld) computer that a crop scout would use to navigate and assess field crop and pest conditions.

Zigzag points Directed points (classified data) Directed points (raw data)

Figure 2. Display of zigzag vs. directed sample locations. Zigzag samples are randomly selected for field observation. The map of directed points used for verification of classified and raw remotely sensed data are included.

Figure 3. Software display with scouting map, directed points, and boundary layers. This display was used to navigate and collect data in the field.

Please check out upcoming newsletters related to spatial data and the directed scouting method. This document is the first in a series of two newsletters outlining the processes and comparison of traditional versus directed sampling approaches used to assess spatial variability. Upcoming newsletters will address estimation algorithms used to produce variability maps and comparisons made between various sampling methods used.

For more information:

Blumhoff, Greg K., 2002. Spatial Variability Assessment With Remote Sensing and Directed Sampling. M.S. thesis. Department of Agronomy, Purdue University, West Lafayette, IN.