University of Twente Student Theses


Hyper-temporal NDVI-derived indicators to estimate maize yields in Kenya

Essien, John Joseph (2012) Hyper-temporal NDVI-derived indicators to estimate maize yields in Kenya.

[img] PDF
Abstract:Food insecurity remains high in many parts of Kenya due to recurrent seasons of failed rains occurring during crop growing seasons. This has resulted in drought situations which have led to extensive loss of yields, displacement of people, malnutrition, epidemics and sometimes resulting to famine. Variability in climatic conditions is projected to be more severe in subsequent years. This calls for concern because large part of Kenya’s population is dependent on rain-fed agriculture. Maize which is a rain-fed crop accounts for >80% cereals used at household level. Accurate and timely information on crop yields estimation in the stages of crop growth is fundamental towards ensuring food security. The use of satellite derived NDVI data for monitoring of crops is of key importance in providing accurate and timely estimation of yields to warn against impending poor or failed harvest. This study is centred on exploiting effectively the use of hyper-temporal SPOT-VGT NDVI-derived indicators to estimate maize yields. Three distinct steps, the ISODATA clustering technique, disaggregation of published maize statistics and simple linear regression model using hyper-temporal NDVI derived indicators as independent variable and maize yields as dependent variables were performed in this study. The ISODATA clustering produced useful temporal NDVI profiles for classification. The disaggregation of maize areas statistic data by NDVI classes gave reasonable fractions of maize per districts. Two NDVI-derived indicators, maximum NDVI and the sum of NDVI were used in establishing correlations with maize yields. Strong and significant correlations were found from both NDVI-derived indicators in estimating maize yields. A coefficient of determination (R2) of 0.71 was obtained between maximum NDVI and maize yields while R2 of 0.84 was obtained between sum of NDVI and maize yields at district level.. The results show the effectiveness of the NDVI-derived indicators in estimating maize yields. Thus this study has demonstrated that hyper-temporal NDVI-derived indicators from SPOT -VGT data can be used for simple, early and reasonably accurate estimation of maize yields.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page