3. Data Analysis

Data analysis is the second major step in assessing goose damage on crop yields (Figure IIA-25). The goal of data analysis is to derive as much information from the given data as possible to assist in separating goose damage from other factors. GIS provides the tools for data analysis. Establishing relationships between the data and the yield, comparing data, or calculating a statistical summary are examples of GIS data analysis techniques that can be applied.

Figure IIA-25: Required steps for assessing goose damage on crop yields.

What is GIS?
Geographic information systems (GIS) provide a systematic approach to managing large amounts of data, along with the tools necessary for analysis and interpretation. This guideline reviews some example data sets used to characterize a field, showing how GIS can help organize and manage the data so as to be used more effectively in various management decisions .

What is a query?
The query function is an integral part of any database. It is used to select a group of records from the tables within a database. Several applications use queries to select yield features by their value or location. Queries can be performed on the yield data to determine whether a relationship exists between certain soil types and crop yield.

Importance of a query
Selecting records from a database or map allows further analysis of the retrieved items. A small area of exceptional yield might be explained by soil conditions found in the same location. Within the GIS environment, locating areas of high or low yield can be done by running a query. Following the query on yield levels, one can define a query that identifies specific soil conditions. Running these two queries on the yield data is the first step to seeking a relationship between yield amounts and soil types.

There are several other reasons to query your GIS data. Among these reasons are:

For example, it is possible to generate summary statistics on the yield points that were selected previously through a query. After the statistics are compiled, it is possible to select the soils that intersect the selected yield points. No matter which results are desired, most spatial analysis starts with some form of query.

Sorting Yield Data

Whether using method 1,Query by theme, or Method 2, Query using edit mode, it is possible to perform the following:

1. Remove confounding factors

Several factors may affect yield other than goose grazing. The extent of impact from these factors should be carefully delineated and their yield separated before any further analysis.

For instance, water inundation (ponded area) has a negative impact on yield. Areas impacted by water should be clipped out and not considered in the evaluation.

Step 2 : Add the Yield Data Layer

Step 3 : Clip (delete) Yield in Water Impacted Areas

One of the most basic sources of potential yield variability is the difference in soil types within the field. The data files associated with the soil types provide a detailed database of soil characteristics that can be related to other observations and measurements.

Depending on field characteristics for each exclosure location per soil type, there are two ways to test for soil differences:

Perform a t-test

For fields having more than two soil types, perform an ANOVA using the yield from each exclosure. In many cases, it may not be possible to perform an ANOVA based on the exclosure yield. However, it is still possible to obtain reliable information about the differences in yield based on soil types by collecting data from a representative subset or sample. Sampling is simply the technique used to choose representative yield data points for analysis from a larger population. In a GIS environment, it is possible to randomly select a sub-sample from the population using bootstrapping.

What is Bootstrapping?

Bootstrapping is a resampling procedure that pulls random observations from an entire data set. This procedure generally is applied to large datasets with many observations. Electronic, automatic sampling such as yield mapping often generates this type of data. Bootstrap methods are computer-intensive processes of obtaining a subset of the data that meets statistical assumptions and can be used to calculate standard errors, confidence intervals, and significance tests. Each subsample is a random sample with replacement from the full sample.

How it is done in Arc/View?

Step 1: From the movement menu select create bootstrap file

Step 2: Insert input data

A bootstrap file creator window will open.

Specify the theme name, the number of iteration to be performed, and the sample size.

Step 3: Name and save the output bootstrap feature.

No soil Difference

In case there is no significant yield difference due to soil type, we can combine the yield data from all exclosures for further analysis.

3. Sort data by intensity of grazing

Since most of the necessary data has been collected in the field or subsequently generated, it is possible to use the data to determine whether relationships exist between goose grazing and yields.

This step can be performed using a query or a clip command. Figure IIA-27 illustrates the process of using the clip command. (a) Polygon theme (areas grazed in March) used to clip yield data point. (b) Output map representing yield data within a particular class of interest (in this case, areas grazed in March).

Figure IIA-27: Map showing March grazing (left) and yield data clipped for the area grazed in March (right).

Using GIS to extract yield data in a series of steps from within zones of impact, we can determine yields from areas grazed in April, areas grazed in March but not April, areas grazed in January but not later, areas in exclosures, and other nongrazed areas (Figure IIA-28). This procedure allows us to develop a complete picture of goose grazing impact on yields and to evaluate the effects of seasonal grazing.

Figure IIA-28: Summary table comparing the impact of timing of grazing on yield.

4. Obtain estimate of yield using paired plot comparisons

The same exclosure (control) can be used for more than just one-pair comparisons.

Step 1 : Overlay Exclosure and Paired Plots on Top of the Grazing Map

Both maps showing paired plot location in relation to respective exclosure.

Using the data gathered from the previous section perform appropriate statistical analysis (Figures IIA-29 and IIA-30).

Figure IIA- 29: Using a paired t-test to compare yield inside exclosures to paired plots grazed in March.

Figure IIA- 30: Using a paired t-test to compare yield inside exclosures to paired plots grazed through April.