Professional Documents
Culture Documents
Textbook Advances in Complex Societal Environmental and Engineered Systems 1St Edition Mohamed Nemiche Ebook All Chapter PDF
Textbook Advances in Complex Societal Environmental and Engineered Systems 1St Edition Mohamed Nemiche Ebook All Chapter PDF
https://textbookfull.com/product/advances-in-complex-societal-
environmental-and-engineered-systems-1st-edition-mohamed-
nemiche-2/
https://textbookfull.com/product/mathematical-advances-towards-
sustainable-environmental-systems-1st-edition-james-n-furze/
https://textbookfull.com/product/an-introduction-to-agent-based-
modeling-modeling-natural-social-and-engineered-complex-systems-
with-netlogo-the-mit-press-uri-wilensky-william-rand/
https://textbookfull.com/product/bioethanol-and-natural-
resources-substrates-chemistry-and-engineered-systems-1st-
edition-ruben-michael-ceballos/
Deep Learning for Vision Systems 1st Edition Mohamed
Elgendy
https://textbookfull.com/product/deep-learning-for-vision-
systems-1st-edition-mohamed-elgendy/
https://textbookfull.com/product/deep-learning-for-vision-
systems-1st-edition-mohamed-elgendy-2/
https://textbookfull.com/product/deep-learning-for-vision-
systems-1st-edition-mohamed-elgendy-3/
https://textbookfull.com/product/advances-in-energy-and-
environmental-materials-yafang-han/
https://textbookfull.com/product/the-crisis-of-the-14th-century-
teleconnections-between-environmental-and-societal-change-bauch/
Nonlinear Systems and Complexity
Series Editor: Albert C. J. Luo
Mohamed Nemiche
Mohammad Essaaidi Editors
Advances in
Complex Societal,
Environmental
and Engineered
Systems
Nonlinear Systems and Complexity
Volume 18
Series Editor
Albert C. J. Luo
Southern Illinois University
Edwardsville, IL, USA
Advances in Complex
Societal, Environmental
and Engineered Systems
123
Editors
Mohamed Nemiche Mohammad Essaaidi
Ibn Zohr University ENSIAS College of Engineering
Agadir, Morocco Mohamed V University of Rabat
Rabat, Morocco
This book addresses recent technological progress that has led to an increased
complexity in societal, ecological, and engineered systems. This complexity is
characterized by the emergence of new proprieties and structures resulting from
nonlinear interactions among system elements and between system and its envi-
ronment. This volume provides researchers and managers with qualitative and
quantitative methods, bottom-up and holistic approaches for handling many features
of the complex contemporary reality. This book is composed of three parts with
a total of 13 chapters: Part I (Chaps. 1–6) focuses on societal and ecological
systems, Part II (Chaps. 7–12) deals with approaches for understanding, modeling,
forecasting, and mastering complex systems, and Part III (Chap. 13) includes real-
life examples. Each chapter of this book has its own special features; it is a
self-contained contribution of researchers working in different fields of science
and technology relevant to the study of complex systems including Agent-Based
Modeling, General Systems Theory, and Mathematical Modeling.
In the chapter “ProtestLab: A Computational Laboratory for Studying Street
Protests,” Lemos, Coelho, and Lopes present an Agent-Based model for the
simulation of street protests, with multiple types of agents (protesters, police, and
“media”) and scenario features (attraction points, obstacles, and entrances/exits). In
this model agents can have multiple “personalities,” goals, and possible states. The
model includes quantitative measures of emergent crowd patterns, protest intensity,
police effectiveness, and potential “news impact,” which can be used to compare
simulation outputs with estimates from videos of real protests for parameterization
and validation. ProtestLab was applied to a scenario of policemen defending a
government building from protesters and reproduced many features observed in real
events, such as clustering of “active” and “violent” protesters, formation of moving
confrontation lines, occasional fights and arrests, “media” agents wiggling around
“hot spots,” and policemen with defensive or offensive behavior.
In the chapter “A Generic Agent Based Model of Historical Social Behavior
Change,” Ahmed M’hamdi et al. describe and discuss how human societies change
over time. The main objective of this work is to build a generic agent-based model
v
vi Preface
train railway systems. Meanwhile, the delineation of traffic impact areas could be
spatially targeted at priorities of traffic improvement for city planners.
In the chapter “Logic, Mathematics and Consistency in Literature: Searching for
Don Quixote’s Place,” Montero et al. combine fuzzy logic with other techniques
to analyze the consistency of the linguistic discourse about the village Miguel de
Cervantes (1547–1616) decided not to reveal in his classical Don Quixote’s novel.
In particular, the authors consider Cervantes linguistic description of Don Quixote’s
trips from and to that place in order to check if such information is consistent with
the map of La Mancha, and allowing a more or less constant walking speed march
per day. From this complex system of information, it is then concluded that there
is in fact a small region in the center of Campo de Montiel that meets all estimated
walking times per day, showing that Cervantes linguistic description of the trips
involving the hidden place is in this sense consistent.
In the chapter “Energy-Efficient Buildings as Socio-Technical Complex Systems:
Approaches and Challenges,” Lachhab et al. present important metrics that assess
performance and occupants’ comfort in energy-efficient buildings and study their
relationships with building physical properties, equipment control, outdoor envi-
ronment, and occupants’ behavior and activities. The authors analyze occupants’
actions and behaviors in context taking into account the complex interlinked entities,
situations, processes, and their dynamics. Lachhab et al. then review existing control
approaches and solutions for energy efficiency in complex buildings. They highlight
simulation tools that aim to study and analyze approaches for energy consumption,
occupants’ comfort, and CO2 emissions. They also introduce their ongoing work
related to modeling and control of these complex systems by highlighting the
necessity of the development of intelligent building management systems that could
include run-time processing techniques of large amount of data for deploying
context-aware event-triggered control techniques.
In the chapter “Modeling Space-Time-Action Modularity and Evolution of
Living Systems,” Pierre Bricage develops a new paradigm of “the gauge invariance
of living systems” which allows him to define functionally and dynamically the
hierarchical fractal organization of all living systems, from the quantum of Planck to
the whole universe. The ontogeny of interactions and the interaction of ontogenies
are topologically related through a power law of exponent 3/2, which is basically
related to the Brownian motion. His work was motivated by the need to understand
why all living systems are emerging through juxtaposition and embedment of
previous ones in a new blueprint that is always an Association for the Reciprocal
and Mutual Sharing of Advantages and Disadvantages: “interaction is construction
and construction is interaction” both for the spaces and times, which are juxtaposed
and embedded simultaneously into limited spatial and temporal networks, within
an “independent of mass-entropy relationship inter-active optimal surface flow of
flows” organizing control.
Preface ix
The editors would like to thank all the authors who contributed to this book.
Special thanks also go to Mr. Christopher Coughlin and Mr. Abdeljalil Haib for
their collaboration in this book project.
xi
xii Contents
Editors
xiii
xiv List of Editors and Contributors
Contributors
1 Introduction
Street protests and riots are manifestations of social conflict, which results from
feelings of frustration due to relative deprivation (RD) and cleavages within the
society (Gurr 1968). They may be large or small, peaceful of violent, organized or
– How do the features of the protest space and the crowd size, composition and
initial placement affect the crowd formation patterns (wandering, clustering and
fighting) and protest intensity?
– How does violent confrontation arise? Once initiated, does it spread to the bulk
of the crowd or remains confined to specific types and clusters of protesters?
– How can different police actions influence the intensity of violent confrontation
and the ability for defending a perimeter?
The novel features of ProtestLab with respect to other ABM are: (i) multiple
types of agents (‘protesters’, ‘cops’ and ‘media’), with multiple behaviours repre-
sented by agent subtypes; (ii) several types of scenario features (attraction points,
obstacles, entrances and exits); (iii) a simple but efficient reactive agent architecture
which allows all agents to have multiple goals and spatially oriented interactions;
and (iv) quantitative measures of emergent crowd patterns, protest intensity, police
vs protesters effectiveness and potential “media” coverage.
Three subtypes of ‘protester’ agents were considered, ‘hardcore’, ‘hanger-
on’ and ‘bystander’, which broadly correspond to typical protester behaviours
well characterized in micro-situational theories of violence (Collins 2008, 2009;
Wikström and Treiber 2009), ABM of clustering and fighting (Jager et al. 2001), and
also easily identified in many videos of real protests. Four subtypes of ‘cop’ agents
were included, namely ‘command’, ‘offensive’, ‘defensive’ and ‘multi role’, which
correspond to different mission profiles and assigned tasks in a police force. All
agents can be one of four states, ‘quiet’, ‘active’ (waving, shouting, etc.), ‘violent’
(ripe for confrontation) or ‘fighting’. Fights only occur when ‘violent’ protesters
and cops are in contact, such as in body fighting (fights with weapons such as stones
and tear gas are not considered in the present version of the model).
The advantage of ProtestLab with respect to existing ABM of crowd dynamics
and violent confrontation is the capability of running realistic simulations of street
protests with several hundred agents using small computers. The present version of
ProtestLab is not intended for modelling fights between two groups (e.g. supporters
of rival parties or football teams) mediated by a police force or events of collective
violence that are not politically motivated, although the model can be extended to
treat these cases. Consequently, the theories and models related to pedestrian flow,
panic evacuation, ethnic violence or violence in two-party crowds (e.g. in football
stadiums) will be mentioned only in the aspects relevant for the issues considered in
ProtestLab.
The remainder of this paper is organized as follows. In Sect. 2 we present the
theoretical background for ProtestLab. In Sect. 3 we describe the model entities,
design concepts, quantitative measures of the emergent properties and issues
related to parametrization and validation. In Sect. 4 we present the results of a
set of simulations for a scenario of a police force defending the entrance of a
government building from protesters, for different crowd densities and compositions
(proportions of subtypes of ‘protester’ agents), average aggressiveness, and police
‘personality’ or mission profile and the action of a police “command” agent.
Section 5 contains the discussion of the model results with emphasis on the model’s
6 C.M. Lemos et al.
Several ABM have been proposed for the simulation of conflict phenomena of
different types and scales. Epstein et al. (2001) introduced an ‘abstract’ ABM of
large-scale rebellion against a central authority (Model I) or ethnic violence between
two rival groups (Model II) with two types of agents, population and cops in a
torus space environment. Population agents can be in three states (‘quiet’, ‘active’
or ‘jailed’). Both types of agents have one movement rule M (move to a random
empty cell within the agent’s vision radius) and one action rule. Population agents
that are not ‘jailed’ become ‘active’ according to the following threshold (action)
rule:
P D 1 exp.k.C=A/v / (2)
as a function of deprivation and the attractiveness of the sites for rioting. The choice
of the rioting site is modelled via an utility function which takes into account
the distance from the rioter’s present position, the attractiveness and the level of
deterrence expected at each site. As rioters interact with police they may be arrested
with probability described by a function similar to that in Eq. (2). Although models
of this type (e.g. Davies et al. 2013, Parunak et al. 2014) provide a representation
of rioting events and include context-specific information for their parametrization
and validation, they do not describe the micro-scale features of the interactions
between protesters and the police forces (such as clustering of violent, active and
quiet protesters, the formation of confrontation lines with fights and arrests, etc.).
Modelling micro-scale processes in crowd dynamics is a problem of great
theoretical and practical importance. There are two types of crowd models relevant
for the present work: models of pedestrian flow and crowd density, and models
of riotous crowds. The study of pedestrian flow and the formation of spots with
dangerous crowd density in panic evacuation has been intensely studied using meth-
ods from fluid dynamics (Helbing and Molnár 1995), cellular automata (Klüpfel
2003) and particle dynamics (Leggett 2004). Several well developed models are
available for such applications (see Still 2000, and TSO 2010a,b for a review of
existing models), which use a continuous space and time description of the motion
of pedestrians as self-propelled particles subject to a force with three components
(Social Force Model, Helbing and Molnár 1995) that express individual tendencies
to (i) maintain a desired speed towards a wanted destination point; (ii) keep
clear from other pedestrians or obstacles (repulsion force, decaying exponentially);
and (iii) approach attractive features (such as other persons or displays) (Helbing
and Molnár 1995; Helbing et al. 2001). The main difficulties are the need for
algorithms for collision avoidance, description of the contact interactions between
pedestrians and obstacles, and the need for integrating the equations of motion for
each pedestrian (Pelechano et al. 2007). Discrete space and time models, in which
pedestrians move in discrete steps and only occupy empty cells within their move
range are simpler, because integration of the equations of motion for each pedestrian
is not necessary (Klüpfel 2003). If the grid cell size is large enough for the repulsion
forces to be negligible and only one agent per grid cell is allowed, then the agents’
movement can be modelled in a very simple way. This latter approach was used in
ProtestLab due to its simplicity and advantage for handling large numbers of agents.
Models for describing processes specific to riotous crowds, such as the formation
of confrontation lines between protesters and police cordons, patterns of clustering,
simple provocation or physical fights and arrests, are comparatively less developed
than models of pedestrian flow and panic evacuation. Jager et al. (2001) presented
an ABM of clustering and fighting between two parties (e.g. hooligans supporting
two different football teams) with three kinds of agents, ‘hardcore’, ‘hangers-on’
and ‘bystanders’. These subtypes are consistent with typical behaviours identified
in micro-situational theories of violence (e.g. Collins 2008, 2009, Wikström and
Treiber 2009) and can be confirmed in many videos of real protests. The agents’
internal state is characterized by an ‘aggression motivation’ that increases/decreases
depending on the difference between the number of visible agents of own and
8 C.M. Lemos et al.
opposing groups. Agents approach members of own group with preference for their
‘acquaintances’ and, depending on their ‘aggression motivation’, may approach and
start fighting with agents of the other group. The simulated results showed that
fighting occurs when there are large asymmetric groups with significant proportions
of ‘hardcore’, and that ‘bystanders’ are victimised. This model has the advantages
of including aggression as a state variable, memory effects in the ‘aggression
motivation’, and purposeful movement to simulate the collective behaviour of get-
ting support before engaging in violence. However, it does not include connection
between aggressiveness and social context variables and deterrence effects (there
are no ‘cop’ agents).
Durupınar (2010) presented a sophisticated ABM for different types of crowds
with psychological factors (personality, mood and emotions), based on a more
sophisticated agent architecture which allows representation of emotion contagion.
This model includes several types of agent personalities (protesters, agitators, and
police agents) and events (such as fires and explosions), and has important features
such as emotion contagion, multiple actors and multiple actions (waving, kicking
and punching). Although the model is computationally demanding, Durupınar
presents simulations with several hundreds of agents, including protesters marching
in a street watched by policemen at the sides. However, it is not clear how the
model would represent clustering, moving confrontation lines, fights and arrests in
a realistic way over many time steps, and also lacks quantitative measures of these
emergent features.
Torrens and McDaniel (2013) proposed an ABM for simulation of riotous crowds
with four types of agents, ‘Civil’, ‘Rebel’, ‘Jailed’ and ‘Police’1 and the model space
is a 2D torus in which sites can be blocked to represent obstacles. The attributes
of the first three types of agents are the same as in Epstein’s model (H, G, Pa , N
and T) and the transition from ‘Civil’ to ‘Rebel’ is done using Eqs. (1)–(2). The
geographic behaviour is implemented with three modules: spatial cognition, loco-
motion and steering. The spatial strategy (goal-driven approach/avoid movement)
is implemented by defining weights for ‘Rebel’ and ‘Police’ agents to approach or
avoid the other ‘Police’ and ‘Rebel’ agents. Although this model has sophisticated
locomotion and steering modules and is able to represent both low and high density
crowds, it falls short in other aspects critical for simulation of street protests.
For instance, the set of agent types and their states is the same as in Epstein’s
‘abstract’ ABM. Also, the approach/avoid weights are constant, whereas in reality
they depend on local superiority or inferiority of protesters and police agents. The
results reported in Torrens and McDaniel (2013) do not show the crowd patterns
typical of street protests with violent confrontation, and for a simulation scenario
of mass protest with 5000 citizens 42 % of the crowd ends up arrested, which is
obviously unrealistic.
1
In fact there are only two types of agents. ‘Civil’, ‘Rebel’ and ‘Jailed’ are different states of a
single agent type, as in Epstein’s ABM (Epstein 2002; Epstein et al. 2001).
A Computational Laboratory for Studying Street Protests 9
Lacko et al. (2013) proposed an ABM for the simulation of urban riots, with
police, protesters and a remotely operated vehicle (called “BOŽENA RIOT”)
designed for riot control, which considers cognitive modelling of agents, collision
avoidance and crowd pressure calculations. The authors claim to have implemented
the PECS (Physical Conditions, Emotional State, Cognitive Capabilities and Social
Status) agent architecture (Schmidt 2002), and model fear in terms of the number of
other persons within an agent’s “intimate zone”. Pressure calculations are performed
by dividing the crowd in groups, then defining a group direction which is used to
compute the pressure for individual agents in a virtual grid. However, these authors
do not give details of their PECS implementation and do not report any results
in Lacko et al. (2013). Thus, it is not possible to know whether or not the model
reproduces realistic patterns of confrontation, how it handles variable density and
pressure within the crowd, or how many agents can be included in one simulation.
In summary, existing ABM of riotous crowds have limited ability for modelling
the variety of actors, states and micro-interactions in typical street protests. It would
be desirable to integrate in one model different types of actors (protesters and
policemen with differentiated behaviour), interactions (fights and arrests), scenario
features, and quantitative measures of crowd patterns and protest intensity, for more
realistic simulations. Thus, the purpose of ProtestLab is to overcome some of these
drawbacks.
3 Model Description
ProtestLab was implemented in NetLogo and has the following entities: observer
(world, global variables), turtles (agents) and patches (space cells in the
model space). Three types of agents were implemented, ‘cops’, ‘protesters’ and
‘media’. The first two types correspond to the main interacting players and the latter
is necessary for describing potential ‘news impact’ of protest events. The presence
of “media” agents also changes the microscopic context by attracting curious
protesters. Three subtypes of ‘protesters’ were considered, ‘hardcore’, ‘hanger-on’
and ‘bystanders’ which broadly correspond to characteristic behaviours observed
in protests and reported in the literature (Collins 2008; Jager et al. 2001), and four
subtypes of ‘cop’ agents, ‘defensive’, ‘offensive’, ‘multi-role’ and ‘command’, to
reflect mission profile differentiation and allow a combination of command-oriented
10 C.M. Lemos et al.
Fig. 1 Simplified class diagram for the agents and scenario features considered in ProtestLab
ProtestLab is based on the following design concepts: (i) agents are reactive and
rule-based, move in discrete time and space increments (at most one agent per grid
cell) and change their behavioural state and position depending on the local context;
(ii) the agents’ goal-directed movement is implemented by minimizing a penalty
function that involves weights for avoiding or approaching other agents and scenario
features (as in Ilachinsky 2004); and (iii) agents are activated in random order with a
2
However, if the global variable move-radius is increased from its default value of 1 a velocity
varying in steps of 1/move-radius can be introduced.
A Computational Laboratory for Studying Street Protests 11
frequency that depends on their type (as explained in Sect. 3.3) and behavioural state
and perform a scan-plan-behave sequence. Upon activation agents scan the
environment to form their percept PrAi (other agents and scenario features in sight),
plan actions by updating their internal state, and behave by moving and doing an
action that depends on their type and state, such as “taking pictures” of fights in the
case of ‘media’ agents, or starting fights in the case of ‘cops’ or ‘protester’ agents.
The agent attributes common to all agents are their vision radius and angle (which
define the vision cone for detecting nearby agents and scenario features) and move
radius (distance an agent can travel in one time step measured in units of patch
length), which are defined as input variables. The internal state IAi of Ai consists
of the agent’s behavioural state SAi 2 {quiet,active,violent,fighting},
personality vector !Ai (described below), current position x D .x; y/, intended
(planned) position x , and if SAi D fighting the number of time steps to the
end of fight, stored in one of two variables, attacking or defending (to
differentiate fights initiated by ‘cops’ from those initiated by ‘protesters’). ‘Cops’
also have a waypoint variable for storing an eventual patch that the ‘command
cop’ ordered to defend. Figure 2 shows the agents’ architecture used in ProtestLab
and helps understanding the workings of the scan-plan-behave sequence of the
agent cycle. In this figure A is the set of all agents, E is the set of environment
features, SAi is the agent’s behavioural state, !0;Ai .IAi ; PrAi / and !Ai .IAi ; PrAi / are
the default and context-adjusted personality vectors of agent Ai respectively (as
explained below), Ac is the agent’s action and f .x; SAi ; !Ai .IAi ; PrAi // is the agent’s
function (see e.g. Wooldridge 2009 for the terminology).
Agents move to the centre of the empty patch .Ox; yO / within their move range that
minimizes the penalty function
where
0 1
NFk
X X
SD.x; y/ D @ D.x;y/ Fk;j A (4)
Fk 2PrAi jD1
is a double summation, in which the outer sum is taken over the feature sets Fk
and the inner sum involves the distances from entity k (say one fighting ‘cop’)
belonging to a feature set Fk (in this case the agent set of visible fighting ‘cops’)
to point .x; y/, and .x0 ; y0 / is the current position of agent Ai .3 Table 1 shows the
components of the default personality vector !0;Ai for all agent types and subtypes.
Weights have absolute value between 5 (very high) and 0 (neutral, meaning that
the agent is indifferent to entities of the corresponding feature). Positive weights
result in attraction and negative weights in repulsion relative to the corresponding
feature. The method of minimizing a penalty function allows the implementation
of multi-goal driven movement with a single rule, without the complications of
the “combinatorial explosion problem” that would arise with a rule-based IF
<context> THEN <action> formulation (Ilachinsky 2004).
The penalty-minimizing method of inducing goal-driven movement weights
perceived entities with context-dependent motivations or tendencies, but the for-
mulation in Eqs. (3)–(4) is not suitable for targeted movement towards a particular
position or agent. For example, a ‘cop’ Ai may have a motivation to go to a way-
point .xwpi ; ywpi / to defend that position, support other ‘cops’ or engage a protester.
‘Cops’ may also be ordered by a command agent to go to a specific way-point
.xwpc ; ywpc / (e.g. site to be protected or violent protester to be engaged). Denoting
by .Oxwp ; yO wp / the position of the patch centre within the agent’s move range nearest
to the way-point, the next position will be .Ox; yO /
where ˛i is the weight for moving towards an individually selected target and ˛c is
the command weight (or “obedience level”) to go to .xwpc ; ywpc /. Depending on the
P
N
3
The L1 norm kxk1 D jxi j in Eq. (3) was introduced as a normalization factor, so that
iD1
personality vectors of different agents can be inspected and compared if required.
Table 1 Components of the default personality vector for all agent types and subtypes
Fighting Violent Active Quiet Fighting Violent Active Quiet Fighting Quiet
protester protester protester protester cop cop cop cop media media Flag Obstacle Entrance/exit
!0 !1 !2 !3 !4 !5 !6 !7 !8 !9 !10 !11 !12
‘Hardcore’ protester 2 2 1 0 5 4 3 2 0 0 5 1 0
‘Hanger-on’ protester 2 1 0 0 2 1 1 1 2 3 1 1 0
‘Bystander’ protester 5 3 1 0 5 3 1 0 5 1 0 1 1
‘Command’ cop 0 0 0 0 0 0 0 0 0 0 0 0 0
‘Defensive’ cop 0 0 0 0 5 5 5 5 0 0 5 5 0
A Computational Laboratory for Studying Street Protests
‘Offensive’ cop 5 5 1 0 5 5 3 1 0 0 1 1 0
‘Multi-role’ cop 3 3 1 0 3 3 1 1 0 0 3 3 0
‘Media’ 3 1 0 1 3 1 1 0 1 1 1 1 0
A positive sign means attraction and a negative sign means repulsion
13
14 C.M. Lemos et al.
internal state IAi and percept PAi the components of the personality vector of agent
Ai may change by the application of context rules that provide adaptation to the local
context (as in Ilachinsky 2004). These context rules are different for each agent type
and subtype.
Table 2 shows the names of the context rules and the components of the
personality vector that change by their application for all agent types and subtypes
in the model. The ‘hardcore’ cluster rule decreases attraction to flags when less then
ten ‘violent’ or ‘fighting’ protesters are visible by setting !10 D 1; the ‘pursue
cops’ rule increases attraction towards cops when protesters have a superiority of
four to one, by setting !5;6;7 D 5; and the ‘avoid cops’ rule reverses the signs of
!47 when protesters are outnumbered in a proportion equal or greater then two
to one. Notice that the conditions that activate the avoid/approach context rules are
mutually exclusive. The context rules for all other agent types and subtypes follow
the simple principles illustrated for the ‘hardcore’ context rules.
‘Offensive’ cop On-station !012 Pursue protesters !2 !6;7;10;11 Retreat !1 !6;7;10;11 Support !012
‘Multi-role’ cop On-station !1011
‘Media’ Keep-distance !07
15
Another random document with
no related content on Scribd:
THROUGH KIKUYU TO
GALLA-LAND
INTRODUCTION.
My friend, George Henry West, and myself left Cairo in the latter part
of the year 1899, with the intention of proceeding to Uganda viâ
Zanzibar and Mombasa. George was an engineer in the service of
the Irrigation Department of the Egyptian Government, and had
gained a large and varied experience on the new works on the
Barrage below Cairo, then being concluded, and in building, running,
and repairing both locomotives and launches. As a profession I had
followed the sea for three years, leaving it in 1896 in order to join the
British South African Police, then engaged in subduing the native
rebellion in Mashonaland. At the conclusion of hostilities I wandered
over South Africa, and finally found my way to Egypt, where I met
George West. A year later, accompanied by George, I was on my
way southwards again, en route for British East Africa.
When George and I left Cairo, our idea was to go up-country as far
as the Lake Victoria Nyanza, as we considered it extremely probable
that there would be something for us to do in the engineering line,
either in building launches or in the construction of small harbour
works.
We reached Mombasa in due course, and from there proceeded to
Nairobi by the railway then in course of construction to Uganda.
Nairobi is 327 miles from the coast, and is an important centre, being
the headquarters of both the Civil Administration of the Protectorate
and the Uganda railway. On our arrival, George received an offer,
which he accepted, to go up to the lake with a steamer, which was
then on the way out from England in sections, and on his arrival at
the lake with it to rebuild it. I remained in Nairobi.
In course of time I met the personage referred to in these pages
as “El Hakim,”[1] whom I had known previously by repute. He was
said to be one of the most daring and resolute, and at the same time
one of the most unassuming Englishmen in the Protectorate; a dead
shot, and a charming companion. He had been shooting in
Somaliland and the neighbourhood of Lake Rudolph for the previous
four years, and many were the stories told of his prowess among
elephant and other big game.
It was with sincere pleasure, therefore, that I found I was able to
do him sundry small services, and we soon became fast friends. In
appearance he was nothing out of the common. He was by no
means a big man—rather the reverse, in fact—and it was only on
closer acquaintance that his striking personality impressed one.
He had dark hair and eyes, and an aquiline nose. He was a man
of many and varied attainments. Primarily a member of the medical
profession, his opinions on most other subjects were listened to with
respect. A very precise speaker, he had a clear and impartial manner
of reviewing anything under discussion which never failed to impress
his hearers.
He was a leader one would have willingly followed to the end of
the earth. When, therefore, he proposed that I should accompany
him on an ivory trading expedition to Galla-land, that vast stretch of
country lying between Mount Kenia on the south and Southern
Somaliland on the north, which is nominally under the sphere of
influence of the British East African Protectorate, I jumped at the
chance; and it was so arranged. He had been over much of the
ground we intended covering, and knew the country, so that it
promised to be a most interesting trip.
About this time I heard from George that he was coming down
country, as the steamer parts had not all arrived from England, and
consequently it would probably be months before it would be ready
for building. He had also had a bad attack of malarial fever in the
unhealthy district immediately surrounding the lake at Ugowe Bay,
and altogether he was not very fit. I suggested to El Hakim that
George should join us in our proposed expedition, to which he
readily agreed; so I wrote to George to that effect.
To render the prospect still more inviting, there existed a certain
element of mystery with regard to the river Waso Nyiro (pronounced
Wasso Nēro). It has always been supposed to rise in the Aberdare
Range, but, as I shall show, I have very good reason to believe that it
rises in the western slopes of Kenia Mountain itself. The Waso Nyiro
does not empty itself into the sea, but ends in a swamp called
Lorian, the position of which was supposed to have been fixed by an
exploring party in 1893. But, as I shall also show in the course of this
narrative, the position of Lorian varies.
The upper reaches of the Waso Nyiro were visited by the explorer
Joseph Thompson, F.R.G.S., on his way to Lake Baringo during his
memorable journey through Masai Land in 1885.
In 1887-1888 a Hungarian nobleman, Count Samuel Teleki von
Czeck, accompanied by Lieutenant Ludwig von Hohnel, of the
Imperial Austrian Navy, undertook the stupendous journey which
resulted in the discovery of Lakes Rudolph and Stephanie. Count
Teleki, on his journey north, crossed the Waso Nyiro at a point in
North-West Kenia near its source, while Lieutenant von Hohnel went
two or three days’ march still further down-stream.
A few years later, in 1892-1893, Professor J. W. Gregory, D.Sc., of
the Natural History Museum, South Kensington, made, single-
handed, a remarkable journey to Lake Baringo and Mount Kenia,
and in the teeth of almost insuperable difficulties, ascended the
western face of that mountain and climbed the peak.
At the same time, in the latter part of 1892, an American, Mr.
William Astor Chanler, accompanied by Count Teleki’s companion
and chronicler, Lieutenant von Hohnel, started from a point in
Formosa Bay on the East Coast, and made his way along the course
of Tana River to North-East Kenia, intending later to go on to Lake
Rudolph, and thence northward. He and his companion, deceived by
the reports of the natives, which led them to believe that the Waso
Nyiro emptied itself into an extensive lake, and fired by the idea of
the possible discovery of another great African lake, made their way
down to the Waso Nyiro, and after a fearful march, enduring the
greatest hardships, eventually reached Lorian. To their great
disappointment, it proved to be nothing more than a swamp, and
they turned back without examining it. A few weeks later, Lieutenant
von Hohnel, having been seriously injured by a rhinoceros, was sent
down to the coast, his life being despaired of. Shortly afterwards Mr.
Chanler’s men deserted him in a body, and returned to the coast
also, thus bringing his journey to a premature conclusion; a much-to-
be-regretted ending to a well-planned and well-equipped expedition.
As Mr. Chanler was returning to the coast he met Mr. A. H.
Neumann coming up. Mr. Neumann spent the greater part of 1893 in
shooting elephants in the Loroghi Mountains, after going north to
Lake Rudolph. He also crossed the Waso Nyiro at a point north-east
of Mount Kenia.
During the time Mr. Neumann was shooting in the Loroghi
Mountains he was obliged to make periodical visits to M’thara, in
North-East Kenia, in order to buy food from the natives, and on one
such excursion he met Dr. Kolb, a German scientist, who was
exploring North Kenia.
Dr. Kolb ascended Mount Kenia from the north, and then returned
to Europe. An interesting account of his ascent of the mountain is
published in Dr. Petermann’s “Mitteilungen” (42 Band, 1896). Dr.
Kolb then returned to Kenia in order to continue his observations, but
he was unfortunately killed by a rhinoceros a couple of marches
north of M’thara.
Lorian, therefore, with the exception of Mr. Chanler’s hurried visit,
was practically unexplored. At the commencement of our trip, El
Hakim proposed that, if an opportunity occurred of visiting Lorian, we
should take advantage of it, and endeavour to supplement Mr.
Chanler’s information. As will be seen, an opportunity did present
itself, with what result a perusal of this account of our expedition will
disclose.
FOOTNOTES:
[1] Anglice, “The Doctor.”
CHAPTER I.
PREPARATIONS AND START.
Ushanga (Beads).
sem Sem (small red Masai beads).
2 or 3 loads of sembaj (white Masai beads).
ukuta (large white opaque beads).
2 loads of mixed Venetian beads.
When all the loads were packed, they were placed in a line on the
ground; and falling the men in, we told off each to the load we
thought best suited to him. To the Swahilis, being good marching
men and not apt to straggle on the road, we apportioned our
personal equipment, tents, blankets, and table utensils. To the
Wa’Nyamwezi we entrusted the ammunition and provisions, and to
the Wa’kamba we gave the loads of wire, beads, cloth, etc. Having
settled this to our own satisfaction, we considered the matter settled,
and ordered each man to take up his load.
Then the trouble began. First one man would come to us and ask
if his load might be changed for “that other one,” while the man to
whom “that other one” had been given would object with much
excited gesticulation and forcible language to any alteration being
made, and would come to us to decide the case. We would then
arbitrate, though nine times out of ten they did not abide by our
decision. Other men’s loads were bulky, or awkward, or heavy, or
had something or other the matter with them which they wanted
rectified, so that in a short time we had forty men with forty
grievances clamouring for adjustment. We simplified matters by
referring every one to Jumbi, and having beaten an inglorious retreat
to our tents, solaced ourselves with something eatable till everything
was more or less amicably settled.
Nothing is more characteristic of the difference in the races than
the way in which they carry their loads. The Swahilis and
Wa’Nyamwezi, being used to the open main roads, carry their loads
boldly on their heads, or, in some cases, on their shoulders. The
Wa’kamba, on the other hand, in the narrow jungle paths of their
own district find it impossible, by reason of the overhanging
vegetation, to carry a load that way. They tie it up instead with a long
broad strip of hide, leaving a large loop, which is passed round the
forehead from behind, thus supporting the load, which rests in the
small of the back. When the strain on the neck becomes tiring they
lean forward, which affords considerable relief, by allowing the load
to rest still more upon the back. There were also six donkeys, the
property of El Hakim, and these were loaded up as well. A donkey
will carry 120 lbs., a weight equal to two men’s loads.
Finally, we had to register our porters at the Sub-Commissioner’s
office, as no safaris are allowed to proceed until that important
ceremony has been concluded, and the Government has pouched
the attendant fees. In our case, however, there appeared to be a
certain amount of difficulty. On delivering my application I was told to
wait for an answer, which I should receive in the course of the day. I
waited. In the afternoon a most important-looking official document
was brought to me by a Nubian orderly. In fear and trembling I
opened the envelope, and breathed a heartfelt sigh of relief when I
found that the Government had refused to register our porters, giving
as their reason that the districts we intended visiting were unsettled
and, in their opinion, unsafe, and therefore we should proceed only
at our own risk. We did not mind that, and we saved the registration
fee anyhow. The Government had already refused to register the
Somali’s porters, and they intimated, very rightly, that they could not
make any difference in our case.
Jamah Mahomet, who was in command of the Somali safari,
started off that day. He had with him Ismail Robli as second in
command. A smaller safari, under Noor Adam, had started a week
previously. Both these safaris intended visiting the same districts as
ourselves. We were fated to hear a great deal more of them before
the end of our trip.
In the evening I received a private note from one of the
Government officers, informing me that we were likely to have a
certain amount of trouble in getting across the river Thika-Thika
without fighting, as the natives of that district were very turbulent,
and advising us to go another way. My informant cited the case of
Messrs. Finlay and Gibbons by way of a cheerful moral.
Finlay and Gibbons were two Englishmen who had been trading
somewhere to the north of the Tana River. They had forty men or so,
and were trading for ivory with the A’kikuyu, when they were
suddenly and treacherously attacked and driven into their “boma”
(thorn stockade), and there besieged by quite six thousand natives.
From what I saw later, I can quite believe that their numbers were by
no means exaggerated. During a night attack, Finlay was speared
through the hand and again in the back, the wound in the back,
however, not proving dangerous. They managed to get a message
through to Nairobi, and some Nubian troops were sent to their relief,
which task they successfully accomplished, though only with the
greatest difficulty. It was not till six weeks after he received the
wound that Finlay was able to obtain medical assistance, and by that
time the tendons of his hand had united wrongly, so that it was
rendered permanently useless. This was a nice enlivening story,
calculated to encourage men who were setting out for the same
districts.
The following day I received a telegram from George to say that
he had arrived from Uganda at the Kedong Camp, at the foot of the
Kikuyu Escarpment, so I went up by rail to meet him. He looked very
thin and worn after his severe attack of fever. We returned to Nairobi
the same evening, and proceeded to our camp. El Hakim, who was
away when we arrived, turned up an hour later, and completed our
party. He had been to Kriger’s Farm about seven miles out. Messrs.
Kriger and Knapp were two American missionaries who had
established a mission station that distance out of Nairobi, towards
Doenyo Sabuk, or Chianjaw, as it is called by the Wa’kamba.
El Hakim, being anxious to get our men away from the pernicious
influence of the native bazaar, arranged that he would go on to
Kriger’s early on the following morning, and that George and I should
follow later in the day with the safari, and camp for the night near
Kriger’s place. Accordingly he started early in the forenoon on the
following day.
George and I proceeded to finish the packing and make final
arrangements—a much longer task than we anticipated. There were
so many things that must be done, which we found only at the last
minute, that at 3 p.m., as there was no prospect of getting away until
an hour or so later, I sent George on with the six loaded donkeys,
about thirty of El Hakim’s cattle, and a dozen men, telling him that I
would follow. George rode a mule (of which we had two), which El
Hakim had bought in Abyssinia two years before. They were
splendid animals, and, beyond an inconvenient habit, of which we
never cured them, of shying occasionally and then bolting, they had
no bad points. They generally managed to pick up a living and get fat
in a country where a horse would starve, and, taking them
altogether, they answered admirably in every way. I would not have
exchanged them for half a dozen of the best horses in the
Protectorate. One mule was larger than the other, and lighter in
colour, and was consequently known as n’yumbu m’kubwa, i.e. “the
big mule.” It was used by George and myself as occasion required.
The other, a smaller, darker animal, was known as n’yumbu m’dogo,
i.e. “the little mule.” It was ridden exclusively by El Hakim.
After George’s departure I hurried the remaining men as much as
possible, but it was already dusk when I finally started on my seven-
mile tramp. Some of the men had to be hunted out of the bazaar,
where they had lingered, with their loved ones, in a last long farewell.
There is no twilight in those latitudes (within two degrees of the
equator), so that very soon after our start we were tramping along in
the black darkness. I had no knowledge of the road; only a rough
idea of the general direction. I steered by the aid of a pocket-
compass and a box of matches. After the first hour I noticed that the
men commenced to stagger and lag behind with their lately
unaccustomed burdens, and I had to be continually on the alert to
prevent desertions. I numbered them at intervals, to make sure that
none of them had given me the slip, but an hour and a half after
starting I missed three men with their loads, in spite of all my
precautions. I shouted back into the darkness, and the men
accompanying me did the same, and, after a slight interval we were
relieved to hear an answering shout from the missing men. After
waiting a few moments, we shouted again, and were amazed to find
that the answering shout was much fainter than before. We
continued shouting, but the answers grew gradually fainter and more
faint till they died away altogether. I could not understand it at first,
but the solution gradually dawned upon me. We were on a large
plain, and a few hundred yards to the left of us was a huge belt of
forest, which echoed our shouts to such an extent that the men who
were looking for us were deceived as to our real position, and in their
search were following a path at right angles to our own. I could not
light a fire to guide them, as the grass was very long and dry, and I
should probably have started a bush fire, the consequences of which
would have been terrible. I therefore fired a gun, and was answered
by another shot, seemingly far away over the plain to the right.
Telling the men to sit down and rest themselves on the path, I
ordered Jumbi to follow me, and, after carefully taking my bearings
by compass, started to walk quickly across the plain to intercept
them.
It was by no means a pleasant experience, trotting across those
plains in the pitchy blackness, with the grass up to my waist, and
huge boulders scattered about ready and willing to trip me up. I got
very heated and quite unreasonably angry, and expressed my
feelings to Jumbi very freely. I was in the midst of a violent diatribe
against all natives generally, and Swahili porters in particular, which I
must admit he bore with commendable patience, when the earth
gave way beneath me, and I was precipitated down some apparently
frightful abyss, landing in a heap at the bottom, with all the breath
knocked out of my body. I laid there for a little while, and
endeavoured to collect my scattered faculties. Soon I stood up, and
struck a match, and discovered that I had fallen into an old game-pit,
about 8 feet deep. It was shaped like a cone, with a small opening at
the top, similar to the old-fashioned oubliette. I looked at the floor,
and shuddered when I realized what a narrow squeak I might have
had; for on the centre of the floor were the mouldering remains of a
pointed stake, which had been originally fixed upright in the earth
floor on the place where I had fallen.
“Is Bwana (master) hurt?” said the voice of Jumbi from somewhere
in the black darkness above.
I replied that I was not hurt, but that I could not get out without
assistance; whereupon Jumbi lowered his rifle, and, to the
accompaniment of a vast amount of scrambling and kicking, hauled
me bodily out.
We were by this time very near to the men for whom we were
searching, as we could hear their voices raised in argument about
the path. We stopped and called to them, and presently they joined
us, and we all set off together to join my main party. We reached it
without further mishap, and resumed our interrupted march.
It was very dark indeed. I could not see my hand when I held it a
couple of feet from my face. One of the men happening to remark
that he had been over the path some years before, I immediately
placed him in the van as guide, threatening him with all sorts of pains
and penalties if he did not land us at our destination some time
before midnight.
I was particularly anxious to rejoin George, as I had the tents,
blankets, and food, and he would have a very uninteresting time
without me. We marched, therefore, with renewed vigour, as our
impromptu guide stated that he thought one more hour’s march
would do the business. It didn’t, though. For two solid hours we
groped blindly through belts of forest, across open spaces, and up
and down wooded ravines, until somewhere about eleven p.m.,
when we reached a very large and terribly steep ravine, thickly
clothed with trees, creepers, and dense undergrowth. We could hear
the rushing noise of a considerable volume of water at the bottom,
and in the darkness it sounded very, very far down.
I halted at the top to consider whether to go on or not, but the
thought of George waiting patiently for my appearance with supper
and blankets made me so uncomfortable that I decided to push on if
it took me all night. We thereupon commenced the difficult descent,
but halfway down my doubts as to the advisability of the proceeding
were completely set at rest by one of the men falling down in some
kind of a fit from over-fatigue. The others were little better, so I
reluctantly decided to wait for daylight before proceeding further. I
tried to find something to eat among the multifarious loads, and
fortunately discovered a piece of dry bread that had been thrown in
with the cooking utensils at the last moment. I greedily devoured it,
and, wrapping myself in my blankets, endeavoured to sleep as well
as I was able on a slope of forty-five degrees. A thought concerning
George struck me just before I dropped off to sleep, which comforted
me greatly. “George knows enough to go in when it rains,” I thought.
“He will leave the men with the cattle, and go over to Kriger’s place
and have a hot supper and a soft bed, and all kinds of good things
like that,” and I drew my blankets more closely round me and
shivered, and felt quite annoyed with him when I thought of it.
At daylight we were up and off again, and, descending the ravine,
crossed the river at the bottom, and continued the march. On the
way I shot a guinea-fowl, called by the Swahilis “kanga,” and after an
hour and a half of quick walking I came up with George.
He had passed a miserable night, without food, blankets, or fire,
and, to make matters worse, it had drizzled all night, while he sat on
a stone and kept watch and ward over the cattle. The men who had
accompanied him were so tired that they had refused to build a
boma to keep the cattle in. He seemed very glad to see me. We at
once got the tent put up, a fire made, and the boma built, and soon
made things much more comfortable. In fact, we got quite gay and
festive on the bread and marmalade, washed down with tea, which
formed our breakfast.
El Hakim was at Kriger’s place, about a mile distant. We had to
wait two or three days till he was ready to start, as he had a lot of
private business to transact. We left all the cattle except nine behind,
under Kriger’s charge; we sent the nine back subsequently, as we
found they were more trouble than they were worth.
In the evening I went out to shoot guinea-fowl; at least, I intended
to shoot guinea-fowl, but unfortunately I saw none. I lost myself in
the darkness, and could not find my way back to camp. After
wandering about for some time, I at last spied the flare of the camp
fires, halfway up a slope a mile away, opposite to that on which I
stood. I made towards them, entirely forgetting the small river that
flowed at the foot of the slope. It was most unpleasantly recalled to
my memory as I suddenly stepped off the bank and plunged, with a
splash, waist deep into the icy water. Ugh!
I scrambled up the opposite bank, and reached the camp safely,
though feeling very sorry for myself. El Hakim and George thought it
a good joke. I thought they had a very low sense of humour.
On the following morning George and I sallied forth on sport intent.
George carried the shot-gun, and I the ·303. We saw no birds; but
after an arduous stalk, creeping on all fours through long, wet grass,
I secured a congoni. Congoni is the local name for the hartebeeste
(Bubalis Cokei). The meat was excellent, and much appreciated. El
Hakim joined us in the afternoon, accompanied by Mr. Kriger and Mr.
and Mrs. Knapp, who wished to inspect our camp. We did the
honours with the greatest zest, knowing it would be the last time for
many months that we should see any of our own race.