University of Twente Student Theses
A Remotely Sensed Based Comparison of Three Area Stratification Methods to Improve Estimation of Crop Area Statistics, A Case Study in Fragmented Landscapes of Ethiopia
Bamud, Shafi (2022) A Remotely Sensed Based Comparison of Three Area Stratification Methods to Improve Estimation of Crop Area Statistics, A Case Study in Fragmented Landscapes of Ethiopia.
|
PDF
3MB |
| Abstract: | Agricultural statistics, particularly crop production data, are crucial for food security analysis. Crop production is a product of crop area and yield. Thus, accurate and reliable crop area estimation is a critical component of agricultural statistics because it is vital in assessing food production. The optimal method of crop area is determined by various operational criteria, including land arrangement, crop field shape, crop type, cropping pattern, and available skills and resources. In Sub-Saharan African countries, farmlands are small, dispersed, and complex, making estimation of crop area challenging. In addition, high landscape fragmentation, the extreme variability of the environment over space and time and a mixed cropping system are challenging for crop area estimation. Many developing countries, especially sub-Saharan African countries, are affected by the early warning system of food security because of inadequate crop area statistics. Even though there have been various types of crop area estimates, the uncertainty of each type of crop area estimate varies with the topographic nature of the area and other related factors. Therefore, area stratification might be based on areas that do not properly represent the areas of actual crop fields, which leads to the biased representation of area stratification for crop area statistics in fragmented landscapes. To get an accurate and reliable crop area estimate, identifying the most appropriate area stratification method is very important, which in turn reduces the problem of food security that is caused by a lack of accurate crop area estimation. This research compared three existing area stratification methods for Oromia, Ethiopia to improve crop area estimation using high-resolution NDVI imagery. The three existing strata maps (CPS zone, LH zone and administrative unit map) were intersected to prepare a sampling scheme of Planet satellite imagery that specified clusters of tiles (1 km2) of sampled images. As many as 400 sample areas (tiles) were selected as sampling areas. The Planet monthly composite imagery of 2021 was used to obtain the time series (12 month) NDVI images of the 400 tiles. Then ISODATA unsupervised classification method was used to classify the NDVI time series images into NDVI clusters. Area fraction of the clusters by tiles were generated to analyse the relationship and variability of the NDVI clusters and the strata map(s). The Pearson Ch-square(X2) test was employed to analyse the relationships between the NDVI clusters and the strata map. The result revealed that all strata maps had a significant relationship with the NDVI clusters at a 5% significance level. A one-way ANOVA was used to analyse the variability of the area fraction of crop field NDVI clusters and the strata map. The result showed general within and between strata differences, such that the administrative unit (woreda) map had the lowest within strata difference of the others. Finally, maximum variability (counts of significantly different pairs of strata) between strata was determined according to the statistical significance in area fraction of crop field clusters at a 5% significance value using ANOVA, Tukey-HSD, and Hochberg’s GT2 post hoc test analysis of pairwise comparison. The statistical analysis result demonstrated that 60% of the CPS zone pairs of strata had a significant difference, 38% of the LH zone pairs of strata had a significant difference, and 51% of the administrative unit (woreda) map had a significant difference. The overall analysis result of ANOVA indicated that the large significant differences in the CPS zone map consider spatial variability between strata that can identify uniform areas and help to generalize large areas. On the other hand, the administrative unit map had minimum variability within strata than the others, which indicates the administrative unit map provides samples from a small area that can accurately represent the area. To reduce uncertainty in crop area estimation, area stratification is supposed to be representative, relatively uniform within and generally different from its neighbors. The findings support integrating the CPS zone map and the administrative unit map can provide a stratification method for fragmented landscapes like Ethiopia, reducing the uncertainty in crop area estimation. Keywords: Stratification Map, Tiles, NDVI, ISODATA, Crop Field, Planet, Cluster Variability |
| Item Type: | Essay (Master) |
| Faculty: | ITC: Faculty of Geo-information Science and Earth Observation |
| Subject: | 48 agricultural science |
| Programme: | Geoinformation Science and Earth Observation MSc (75014) |
| Link to this item: | https://purl.utwente.nl/essays/92884 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page