Masters Thesis
Rodent pest outbreaks are major concern for agriculture in Africa and Tanzania in
particular, especially in drier areas such as arid and semi-arid agro ecosystems.
This is due to the fact that, if the problem cannot be treated with seriousness it deserves,
80% of the potential harvest may be lost. Crop losses occur at all stages (i.e. field to
market). However, higher losses occur at the field/harvest and storage where rodents play a major role. Severe rodent outbreaks have been reported in many areas in Tanzania e.g. the recent outbreak reported by farmers in Isimani division, Iringa, Tanzania. Based on literature, it is estimated that rodents cause 15% of the total crop pre and post-harvest losses. The problem is compounded by unpredictable rodent pest outbreaks, late control actions, and lack of adequate interventions guided by ecologically based rodent management strategies. Recently, efforts have been taken to develop an ecologically based rodent management strategy which requires knowledge about the pest species' ecology in order to reduce the damage experienced by farmers. However, this is constrained by the limited knowledge about rodent populations on individual farms to allow smarter approaches for control of rodent outbreaks.
Structural characterization and mapping of vegetation habitats could contribute
knowledge about rodent populations on individual farms. Such studies may include
describing and measuring vegetation and habitat structural component using geo spatial and statistical approaches (i.e. life form and cover types, terrain, soil and management practices) across various landscapes in different seasons and their influence to small mammal abundance. Recently, it has been reported that remote sensing derived vegetation indices could be used to explain rodent pest abundance at fine scale. Vegetation indices such as the Normalized Difference Vegetation Index (NDVI) have been reported to correlate well with vegetation productivity (i.e. biomass), forage quality and quantity (i.e. food) at moderate resolutions over a range of spatial-temporal scales. Such indices have been reported to be vital tools for studying vegetation habitat characteristics (i.e. vegetation cover) and their association with rodents in space and time. Therefore, the current study was envisaged to evaluate the potential of NDVI of common vegetation habitats derived from satellite remote sensing data for monitoring rodent population dynamics and outbreaks in order to contribute knowledge for refining ecologically based rodent management strategies. More specifically, the study was carried out to i) characterise and spatially map the vegetation habitats associated with small mammal abundance in smallholder farming agro-ecosystem; ii) determine the Normalised Difference Vegetation Index (NDVI) of common vegetation habitats and rainfall patterns in the study area; and iii) establish a relationship between NDVI of the common vegetation habitats and small mammals distribution and abundance in space and time.
The study was conducted between September 2015 and June 2016 in Isimani Division,
Iringa Tanzania. A combination of field survey and Geospatial approach including the use
of Multi temporal Landsat 8 (Operational Land Imager (OLI)) images were applied to
identify and map the vegetation habitats and estimation of NDVI. The NDVI of common
vegetation habitats and rainfall patterns were also explored. Small mammals were trapped in the mapped vegetation habitat units, and counted for abundances. In total, nine main types of vegetation habitats were investigated. A total of 597 small mammals, potentially related to major rodent pests were trapped. Different levels of scales and resolutions were considered. Linear regression analysis was employed to establish the important habitat characteristics (predictor variables) for relating rodent abundance with rainfall and NDVI.
Linear regression analysis was also used to clarify the relationships between ground
measured rodents and predicted rodent abundance from rainfall and NDVI across seasons, and calculation of the Pearson correlation coefficient (r) at P ≤ 0.05.
Results show that, vegetation habitats identified based on land use/cover types were
largely dominated by agriculture covering about 60% of the plateau landscape with
intensive maize cultivation and frequent reported rodent outbreaks. The findings show
further that, the plateau habitats support a large number of small mammals (80%) than the rest of the habitats in the other landscapes. A strong correlation (r=0.96) was obtained between ground measured point rainfall data and the real time Tropical Rainfall Measuring Mission (TRMM) Precipitation Analysis rainfall data across the identified vegetation habitats. Spatial variability of mean NDVI values with seasonal pattern across the studied vegetation habitat units were obtained whereby, higher values (0.2 to 0.6) were observed in wet season and lower values (0.0 to 0.2) in the dry season. The findings have demonstrated a good positive correlation between rainfall and NDVI along the elevation gradient of the studied landscape units with escarpment having higher correlation (r=0.688) than the plateau (r=0.653) and the valley floor (r =0.652).
This relationship suggests that rainfall patterns could be easily predicted from a link
between NDVI and elevation as predictor variables.
Results also show that, NDVI and rainfall derived from satellite data (Landsat 8 (OLI)
images) have positive influence on the rodent abundance over the studied seasons. It was observed that 98% of the predicted rodent abundance was explained by NDVI while
rainfall explained only 85%. NDVI predicted rodent abundance showed a strong positive
correlation (r=0.99) with the field measured rodent abundance. The obtained NDVI
values provide a robust measure of the presence and abundance of vegetation across the studied vegetation habitats which could be very useful in monitoring rainfall dynamics and as a proxy for predicting rodent pest abundance. The findings have revealed that rainfall, NDVI, and elevation were important predictor variables that could be considered for predicting small mammals or rodent pest abundance in the study area.
These results support the hypotheses that NDVI of common vegetation habitats has the
potential for monitoring rodent population dynamics under smallholder farming
agro-ecosystems. Hence, NDVI could be used to model rodent outbreaks within a
reasonable short time compared to the sparse and not readily available rainfall data.
Further research is required to explore the existing relationship between vegetation
habitats with their associated microclimate and rodent pests in the hotspot areas.
In addition, relationship between NDVI and rodent pest species composition and
community structure in different habitats and seasonal rainfall patterns should be
explored.
Bill and Melinda Gates through TANURAT for provision of research support.