| Human activities have led to an increase in extreme weather events on the African continent(especially in sub-Saharan Africa).In sub-Saharan Africa,about 95%of food crops are grown in rain-fed agricultural areas.Various extreme weather events,such as floods and droughts caused by climate change,can directly affect crop growth and lead to reduced food production,which affects food security in some countries such as Mozambique.Mozambique has a diverse climate and is strategically located in one of the most flood-prone regions on the west coast of the Indian Ocean.The impact of flooding on agricultural land is significant,as it can cause damage to the harvested area and final yield of crops during the growing season,or prevent the harvesting and sowing of crops.This study makes full use of multi-source remote sensing data to monitor the distribution of cropland in the Zambezi River basin,analyze the effectiveness of remote sensing data and methods in mapping and monitoring cropland distribution using different remote sensing data and different methods,map and monitor flood duration,analyze the dynamics of affected crops and their recovery after different flood durations,and assesses the impact of floods on cropland and crop growth from two extreme flood events in Mozambique.This thesis presents a quantitative flood assessment methodology(extent and duration)that provides more detailed information for the assessment of the impact of floods on agriculture.The robustness of the random forest classifier in identifying the distribution of cropland in different agroecological zones was demonstrated,and a high-resolution spatial distribution mapping of cropland in the Zambezi River Basin(ZRB)was prepared,producing a cropland distribution map with a spatial resolution of 10 m.This is the highest resolution and most accurate cropland distribution dataset in the Zambezi River Basin to date,filling the gap in high-precision cropland datasets in the region.The analysis showed that the random forest performed most consistently and with higher accuracy,and the overall accuracy of the obtained cropland extent map was 84%with a kappa coefficient of 0.67;moreover,the classification accuracy of cropland in agro-ecological zones with low temperature was higher than that in agro-ecological zones with the warm climate,which might be caused by the low spectral distinction between cropland and grassland in the latter and the long cultivable time.Regardless of the classifier used,the characteristics of cropland will affect the classification results,o the characteristics of cropland in different agroecological zones should be considered in cropland monitoring.The method developed in this study can be used to regularly monitor the distribution of cropland in the Zambezi River Basin,which will help to analyze the impact of cropland expansion or abandonment on biodiversity and changes in the intensity of cropland use.A method based on optical remote sensing data to assess the area of inundated cropland using satellite imagery with high temporal resolution(day-by-day)but low spatial resolution is proposed to solve the assessment challenge of assessing the area of flood-inundated cropland in rapid response.This was done when high-resolution satellite data were scarce.and the methodology herein described was applied to the flood event that occurred in the Lower Limpopo River Basin,southern Mozambique in January 2013.A support vector machine(SVM)classifier was used to get flood water extent maps,which were then overlaid with cropland maps to generate the area of inundated cropland duration.The most severely flooded areas were the Chóckwè,XaiXai,Mabalane,Massingir,Chibuto,and Guijá districts,which together accounted for228,493.5 ha(92.5% of the total flooded cropland area in the study region)of flooded cropland area.The total flooded cropland area obtained herein is 79% in agreement with official reports.The results showed that the use of high spatial and temporal remote sensing techniques to rapidly assess the impact of flooding on farmland is effective for flood mapping in the Least Developed Countries(LDCs)(i.e.,Mozambique),even in the absence of high-resolution data.The integrated use of synthetic aperture radar data and optical data was explored to identify the spatial distribution of floods and their duration,and an evaluation method for flood impact on crop growth was proposed.The normalized difference vegetation index(NDVI)obtained from sentinel-2 optical data was combined with sentinel-1synthetic aperture radar data and used to obtain a pre-flood watershed area map with an overall accuracy of 94.7% and a kappa coefficient of 0.88.Then,during the flood period,the watershed area was extracted using the backscatter coefficient of the VH band of sentinel-1 images with a threshold method.The monitoring yielded inundated areas for six periods from March 13 to April 25,2019,in central Mozambique.The water surface before and after the flood was analyzed to get the extent of the flood inundation zone.Finally,it was overlaid with the cropland map to assess the extent of inundation of cropland is in different flood periods of flood duration.The results showed that the proposed threshold method for flooded area extent and duration mapping yielded an overall accuracy higher than 90%.Among all the districts in Sofala province,cropland areas in the administrative units of Buzi and Tica and Mafambisse were the most affected,with the largest extent of inundation of 23,101.1 ha in Buzi on20 March 2019 and the longest duration of flooding of over 42 days in Tica and Mafambisse.Major summer crops,including maize and rice,survived inundation for up to 12 days,while all crops were extinguished when flooding lasted for over 24 days,and the longer the inundation,the longer it took for surviving crops to return to preflood conditions,from about 20 days to up to a month after the flood.Although the exact flood loss rate will depend on several factors,such as crop type,this study provides valuable information for planning,mitigation,and re-operation activities in Mozambique. |
PDF Full Text Request