Decision Tree Based Cropland and Crop Type Mapping Using MODIS EVI Data for Ecosystem Services Assessment, a Case of Lake Tana Sub Basin
No Thumbnail Available
Date
2014-06-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Mapping cropland and characterizing their surface vegetation cover for major crops growing
in the region over time and space is an essential input to many applications including, natural
resources assessment, food security monitoring and economics of land use practices. For such
applications, multi temporal remote sensing data and GIS techniques play important role both
as source of data and techniques. This paper took the advantage of the multi-temporal
Moderate Resolution Imaging Spectro-radiometer (MODIS) Enhanced Vegetation Index (EVI)
data to map cropland and types of crop growing in the study area and characterize their surface
cover vegetation variability. The output of the study used to assess implications with respect to
ecosystem services mainly grain production, soil erosion and sediment load to Lake Tana. To
map the types of crops, MODIS EVI data integrated with Traditional Cropping Calendar (TCC)
and a decision tree based classification algorithm were used. The overall approach of the study
is summarized in four major steps; 1) Separating cropland from the non-cropland, 2)
identifying and mapping the different types of crops, 3) characterizing the vegetation dynamics
of different crops over time and space and 4) Assessing major ecosystem service and disservice
of growing different crops in the study areas. The proposed methodology revealed promising
results in separating cropland from the non-cropland ones with 75% of an overall accuracy
whereas crop type mapping was done with 60% of accuracy. Relatively the poor classification
accuracy for crop type map is due to the high landscape variability of the study area, smaller
spatial resolution of the MODIS data which is less than individual plot. Nonetheless, crop type
mapping accuracy can further be improved if more criteria are established. Lake Tana Sub
Basin (LTSB) cropland is dominated by Teff cultivation that exposed the region for more soil
erosion risk. The final outcomes of the present study could support researchers to develop a
crop phenology from multi-temporal data like MODIS. Moreover, maps of different crops
growing in the basin could help to evaluate the economic benefit of different land use practices.
Using the developed crop type maps, the spatio-temporal surface cover condition and biomass
over the cropland could be inter-related with several ecosystem disservices mainly soil loss and
sediment load in a particular watershed.
Description
Keywords
Cropland, Ecosystem Services, Modis Data, Multi-Temporal, Traditional Cropping Calendar