Field data collection for land cover mapping
One of the main objective of the SAMIS in January is to collect ground data for the production of a land cover map of crop types. The interpretation of specific crop types, however, requires a detail knowledge of the farming practises, crop calendars and the collection of sufficient ground observations on individual crops, to train a supervised classification model based on the input of multi-temporal and multi-sensor satellite images.
A mission took place from the 11th of January to the 1st of February 2019 for the ground-data collection team using a mobile application for data collection, train the SAMIS team on post-processing of the data collected and plan next steps. GIS expert from FAO headquarters, SAMIS team and DALaM experts composed the ground-data collection team.
Prior to the fieldwork, a number of districts were selected to sample different landscape in areas with large agricultural area. The districts were selected in two provinces, Luang Namtha and Salavan, corresponding to the two main agro-ecological zones of the country: the upland region in the north and the lowland areas in the south. During the data collections, meetings with the Provincial Agriculture and Forestry Office (PAFO) and District Agriculture and Forestry Office (DAFO) were organized to explain the purpose of the mission and receive permissions to visit the districts.
The whole team was split into two groups and used distinct road itinerary. The windshield technique was adopted for sampling, the cars stopped when agricultural land was encountered and the team of enumerators collected the data in the form. In the evening, the forms were submitted in the server from the team’s devices (tablets or mobile phones). The data were collected using the Kobo Toolbox application (https://www.kobotoolbox.org/). In total around 1000 individual valid data points were submitted into the server.
On return, the team post-processes all the data to allow the correct use of the samples for the supervised classifications. The data will be used from February for training machine learning algorithms to produce a detailed crop types land cover map.
Field data collection for land cover mapping