A generalized computer vision approach to mapping crop fields in heterogeneous agricultural landscapes

Smallholder farms dominate in many parts of the world, particularly Sub-Saharan Africa. These systems are characterized by small, heterogeneous, and often indistinct field patterns, requiring a specialized methodology to map agricultural land cover. Using a variety of sites in South Africa, we present a new approach to mapping agricultural fields, based on efficient extraction of a vast set of simple, highly correlated, and interdependent features, followed by a random forest classifier. We achieved similar high performance across agricultural types, including the spectrally indistinct smallholder fields as well as the more easily distinguishable commercial fields, and demonstrated the ability to generalize performance across large geographic areas. In sensitivity analyses, we determined multi-temporal information provided greater gains in accuracy than multi-spectral information.

Download Full-text

Integrating active learning and crowdsourcing into large-scale supervised landcover mapping algorithms

10.7287/peerj.preprints.3004v1 ◽

2017 ◽

Cited By ~ 1

Author(s):

Stephanie R Debats ◽

Lyndon D Estes ◽

David R Thompson ◽

Kelly K Caylor

Keyword(s):

Active Learning ◽

Large Scale ◽

Learning Algorithm ◽

Training Data ◽

Sub Saharan Africa ◽

Data Sets ◽

Field Patterns ◽

Sub Saharan ◽

Highly Correlated ◽

Computational Resources

Sub-Saharan Africa and other developing regions of the world are dominated by smallholder farms, which are characterized by small, heterogeneous, and often indistinct field patterns. In previous work, we developed an algorithm for mapping both smallholder and commercial agricultural fields that includes efficient extraction of a vast set of simple, highly correlated, and interdependent features, followed by a random forest classifier. In this paper, we demonstrated how active learning can be incorporated in the algorithm to create smaller, more efficient training data sets, which reduced computational resources, minimized the need for humans to hand-label data, and boosted performance. We designed a patch-based uncertainty metric to drive the active learning framework, based on the regular grid of a crowdsourcing platform, and demonstrated how subject matter experts can be replaced with fleets of crowdsourcing workers. Our active learning algorithm achieved similar performance as an algorithm trained with randomly selected data, but with 62% less data samples.

Download Full-text

Integrating active learning and crowdsourcing into large-scale supervised landcover mapping algorithms

10.7287/peerj.preprints.3004 ◽

2017 ◽

Author(s):

Stephanie R Debats ◽

Lyndon D Estes ◽

David R Thompson ◽

Kelly K Caylor

Keyword(s):

Active Learning ◽

Large Scale ◽

Learning Algorithm ◽

Training Data ◽

Sub Saharan Africa ◽

Data Sets ◽

Field Patterns ◽

Sub Saharan ◽

Highly Correlated ◽

Computational Resources

Sub-Saharan Africa and other developing regions of the world are dominated by smallholder farms, which are characterized by small, heterogeneous, and often indistinct field patterns. In previous work, we developed an algorithm for mapping both smallholder and commercial agricultural fields that includes efficient extraction of a vast set of simple, highly correlated, and interdependent features, followed by a random forest classifier. In this paper, we demonstrated how active learning can be incorporated in the algorithm to create smaller, more efficient training data sets, which reduced computational resources, minimized the need for humans to hand-label data, and boosted performance. We designed a patch-based uncertainty metric to drive the active learning framework, based on the regular grid of a crowdsourcing platform, and demonstrated how subject matter experts can be replaced with fleets of crowdsourcing workers. Our active learning algorithm achieved similar performance as an algorithm trained with randomly selected data, but with 62% less data samples.

Download Full-text

Conservation research output in sub-Saharan Africa is increasing, but only in a few countries

Oryx ◽

10.1017/s0030605320000046 ◽

2020 ◽

pp. 1-10

Author(s):

P. Christy Pototsky ◽

Will Cresswell

Keyword(s):

South African ◽

Agricultural Land ◽

Research Output ◽

Sub Saharan Africa ◽

Search Term ◽

African Countries ◽

Research Papers ◽

North American Country ◽

Conservation Research ◽

Sub Saharan

Abstract We tested if peer-reviewed conservation research output has increased in sub-Saharan African countries over the last 30 years in response to increased development. We carried out a bibliometric analysis to identify the number of conservation research papers published by national authors of 41 sub-Saharan African countries during 1987–2017, to provide an index of national conservation research output. We identified country-specific development factors influencing these totals, using general linear modelling. There were positive relationships between conservation research output and population size, GDP, literacy rate, international tourism receipts and population growth rate, and negative relationships with urban population and agricultural land cover, in total explaining 77% of variation. Thirty-eight per cent of countries contributed < 30 conservation research papers (of 12,701) in 30 years. Analysis of trends in primary authorship in a random subsample of 2,374 of these papers showed that primary authorship by sub-Saharan African authors has increased significantly over time but is now at a lower rate than primary authorship for authors from countries outside the country associated with the search term, usually a European or North American country. Overall, 46% of papers had national primary authors, but 67% of these were South African. The results show that conservation research output in sub-Saharan Africa overall is increasing but only significantly in a few countries, and is still dominated by non-national scientists, probably as a result of a lack of socio-economic development.

Download Full-text

Institutional Factors and High-Performance Work Organisations (HPWOs) in Sub-Saharan Africa (SSA)

Management Science - Management and Industrial Engineering ◽

10.1007/978-3-030-13229-3_9 ◽

2019 ◽

pp. 199-218

Author(s):

Raphael O. Oseghale ◽

Richard B. Nyuur ◽

Yaw A. Debrah

Keyword(s):

High Performance ◽

Sub Saharan Africa ◽

Institutional Factors ◽

Sub Saharan

Download Full-text

Land Use, Environmental, and Food Consumption Patterns in Sub-Saharan Africa, 2000–2015: A Review

Sustainability ◽

10.3390/su13158200 ◽

2021 ◽

Vol 13 (15) ◽

pp. 8200

Author(s):

Jeffrey Chiwuikem Chiaka ◽

Lin Zhen

Keyword(s):

Land Use ◽

Food Supply ◽

Agricultural Land ◽

Land Use Changes ◽

Sub Saharan Africa ◽

The Sustainable Development ◽

Sparse Vegetation ◽

Local Methods ◽

Development Goals ◽

Sub Saharan

Sub-Saharan Africa (SSA) land use changes are primarily influenced by agriculture and its population. The region faces various challenges ranging from rainfall variabilities to poverty and insecurities, which further hampered food supply and production. The spatial analysis identified six land uses—agriculture, forest, grassland, wetland, urban, and others (i.e., bare land, water, and sparse vegetation), showing relative percentage changes. Additionally, information collected and analyzed shows that the Millennium Development Goals period witnessed increased agricultural land use changes in the environment to improve food supply, and farmers adopted local methods and native experiences to mitigate environmental particularities facing the region. Farmers’ landholdings are fragmented, and food supply per capita is low albeit rich in calories, and nutrition is still unbalanced, while bushmeat consumption is popular and serves as an alternative to animal-sourced protein. Concerted efforts should be made to improve food security and edge closer to the sustainable development goal during this decade.

Download Full-text

Food price volatilities and civilian victimization in Africa

Conflict Management and Peace Science ◽

10.1177/0738894217729527 ◽

2017 ◽

Vol 37 (2) ◽

pp. 193-214

Author(s):

Babak Rezaeedaryakenari ◽

Steven T. Landis ◽

Cameron G. Thies

Keyword(s):

Food Insecurity ◽

Agricultural Land ◽

Subsequent Treatment ◽

Food Price ◽

Sub Saharan Africa ◽

Rebel Groups ◽

Human Rights Organizations ◽

Violence Against Civilians ◽

Sub Saharan ◽

The Impact

This paper studies the impact of food insecurity on civilian–rebel interactions. We argue that food price volatilities affect the incentives of insurgent groups and their subsequent treatment of civilians. The hypotheses developed in this study are empirically evaluated across a battery of statistical models using monthly data from a sample of 112 first administrative districts in sub-Saharan Africa. The results show that increases in food insecurity substantially raise the likelihood of insurgent groups committing violence against civilians and that districts with a higher proportion of agricultural land are at greatest risk of civilian victimization by rebel groups during these episodes of food insecurity. The implications of this analysis suggest that the human impact of food insecurity does not simply relate to nutrition and questions of governance. Food price volatilities also incentivize the use of violence against civilians by non-state actors, which is a pertinent concern of human rights organizations and policymakers.

Download Full-text

African Biomass Burning and Its Atmospheric Impacts

Oxford Research Encyclopedia of Climate Science ◽

10.1093/acrefore/9780190228620.013.523 ◽

2020 ◽

Author(s):

Charles Ichoku

Keyword(s):

Biomass Burning ◽

Agricultural Land ◽

Water Cycle ◽

Winter Season ◽

Satellite Observations ◽

Sub Saharan Africa ◽

Atmospheric Composition ◽

Research Activities ◽

Sub Saharan ◽

Annual Distribution

Biomass burning is widespread in sub-Saharan Africa, which harbors more than half of global biomass burning activity. These African open fires are mostly induced by humans for various purposes, ranging from agricultural land clearing and residue burning to deforestation. They affect a wide variety of land ecosystems, including forests, woodlands, shrublands, savannas, grasslands, and croplands. Satellite observations show that fires are distributed almost equally between the northern and southern hemispheres of sub-Saharan Africa, with a dipole-type annual distribution pattern, peaking during the dry (winter) season of either hemisphere. The widespread nature of African biomass burning and the tremendous amounts of particulate and gas-phase emissions the fires produce have been shown to affect a variety of processes that ultimately impact the earth’s atmospheric composition and chemistry, air quality, water cycle, and climate in a significant manner. However, there is still a high level of uncertainty in the quantitative characterization of biomass burning, and its emissions and impacts in Africa and globally. These uncertainties can be potentially alleviated through improvements in the spatial and temporal resolutions of satellite observations, numerical modeling and data assimilation, complemented by occasional field campaigns. In addition, there is great need for the general public, policy makers, and funding organizations within Africa to recognize the seriousness of uncontrolled biomass burning and its potential consequences, in order to bring the necessary human and financial resources to bear on essential policies and scientific research activities that can effectively address the threats posed by the combined adverse influences of the changing climate, biomass burning, and other environmental challenges in sub-Saharan Africa.

Download Full-text

Rwenzori Score (RS): A Benthic Macroinvertebrate Index for Biomonitoring Rivers and Streams in the Rwenzori Region, Uganda

Sustainability ◽

10.3390/su122410473 ◽

2020 ◽

Vol 12 (24) ◽

pp. 10473

Author(s):

Peace Liz Sasha Musonge ◽

Pieter Boets ◽

Koen Lock ◽

Minar Naomi Damanik Ambarita ◽

Marie Anne Eurie Forio ◽

...

Keyword(s):

Freshwater Ecosystems ◽

Sub Saharan Africa ◽

Biotic Index ◽

Average Score ◽

Canonical Ordination ◽

The Sustainable Development ◽

Pollution Tolerance ◽

System P ◽

Sub Saharan ◽

Highly Correlated

The Rwenzori region in Uganda, a global biodiversity hotspot, is currently undergoing exponential economic and population growth, which puts continuous stress on its freshwater ecosystems. In Sub-Saharan Africa, biomonitoring campaigns using region-specific biotic indices is limited, particularly in Uganda. In this research, we present the Rwenzori Score (RS), a new macroinvertebrate-based biotic index developed to specifically assess the aquatic health of Rwenzori streams and rivers. We collected and measured both biological and physicochemical variables and identified 34,202 macroinvertebrates, belonging to 64 different taxa. The RS was developed in two steps. First, using canonical ordination, we identified chemical variables that correlated significantly with gradients in macroinvertebrate assemblage distribution and diversity. Second, based on selected variables and weighted averages, we determined specific family indicator values and assigned pollution tolerance values (varying from 1: tolerant; to 10: sensitive) to a family. Finally, we established four water quality classes: poor, fair, good, and excellent. The RS is highly correlated with the Average Score Per Taxon System (p < 0.05), a well-known and widely used biotic index. The RS has 5 unique taxa that are not included in other regional indices. In this regard, the development of the RS is a beneficial tool for tailor-made biomonitoring that can contribute to the sustainable development of the Rwenzori stream and river basins.

Download Full-text

Projected Impact and Cost-effectiveness of Community-based Versus Targeted Azithromycin Administration Strategies for Reducing Child Mortality in Sub-Saharan Africa

Clinical Infectious Diseases ◽

10.1093/cid/ciz1220 ◽

2020 ◽

Cited By ~ 1

Author(s):

Rebecca L Brander ◽

Marcia R Weaver ◽

Patricia B Pavlinac ◽

Grace C John-Stewart ◽

Stephen E Hawes ◽

...

Keyword(s):

Cost Effectiveness ◽

Child Mortality ◽

Mortality Risk ◽

Macrolide Resistance ◽

Sub Saharan Africa ◽

Sensitivity Analyses ◽

Base Case ◽

Published Data ◽

Risk Of Death ◽

Sub Saharan

Abstract Background Trials of mass drug administration (MDA) of azithromycin (AZM) report reductions in child mortality in sub-Saharan Africa. AZM targeted to high-risk children may preserve benefit while minimizing antibiotic exposure. We modeled the cost-effectiveness of MDA to children 1–59 months of age, MDA to children 1–5 months of age, AZM administered at hospital discharge, and the combination of MDA and postdischarge AZM. Methods Cost-effectiveness was modeled from a payer perspective with a 1-year time horizon, and was presented as cost per disability-adjusted life-year (DALY) averted and death averted, with probabilistic sensitivity analyses. The model included parameters for macrolide resistance, adverse events, hospitalization, and mortality sourced from published data. Results Assuming a base-case 1.64% mortality risk among children 1–59 months old, 3.1% among children 1–5 months old, 4.4% mortality risk postdischarge, and 13.5% mortality reduction per trial data, MDA would avert ~267 000 deaths at a cost of $14.26/DALY averted (95% uncertainty interval [UI], 8.72–27.08). MDA to only children 1–5 months old would avert ~186 000 deaths at a cost of $4.89/DALY averted (95% UI, 2.88–11.42), and postdischarge AZM would avert ~45 000 deaths, at a cost of $2.84/DALY (95% UI, 1.71–5.57) averted. Cost-effectiveness decreased with presumed diminished efficacy due to macrolide resistance. Conclusions Targeting AZM to children at highest risk of death may be an antibiotic-sparing and highly cost-effective, or even cost-saving, strategy to reduce child mortality. However, targeted AZM averts fewer absolute deaths and may not reach all children who would benefit. Any AZM administration decision must consider implications for antibiotic resistance.

Download Full-text