Download A user's guide for identifying rice paddocks using GIS and remote
Transcript
These “approximate” boundaries should be identified separately so that the ones that are classified as rice can be updated in the future from either a previous aerial photograph or by using a GPS in the field. When the difference is small (for example, due to a missing rice bay in a paddock), this defines a change that can probably be fixed from past aerial photos. When the change is too big to be identified properly from previous photos, someone will need to update the boundary using a GPS unit in the field. Remember, the date of update, who recorded the GPS data, and who updated the GIS boundary could be recorded in the GIS paddock boundary file. These three steps are outlined below. 5.2 Defining flooded areas on the imagery The purpose of this initial pass is only to assess which paddocks may need updating and requires a per-pixel classification. Since water absorbs light in the wavelengths around 1650 nm where band 5 is centred, the rice areas during this time of year have very low values. Also, non-rice paddocks are predominantly some much drier surface (e.g., soil, stubble, or green crops that are not near canopy closure), revealing very high values in ETM band 5. Therefore, it is generally very easy to classify water versus non-water (which is really what we are doing); we assume that all the flooded paddocks are rice. We will put these values in a per-paddock context later. Be careful to note whether a heavy rainfall occurred just prior to image acquisition as this might influence the results of this analysis. • First, the GRID of ETM band 5, generated in the Ground Validation section, above is classified using the threshold value determined in that same section. As determined above, the threshold between rice/non-rice from our training sets was 75.5838. • This entails using the “Map Calculator” in the Analysis/ menu. Set up the equation to define the classes based on the above threshold (see figure below): • This will result in the calculation of a GRID that has a two classes: zero’s for nonwater areas and one’s for water (see below). CSIRO Land and Water - 20 -