NOT FOR QUOTATION WITHOUT PERMISSION OF THE AUTHOR
DECISION SUPPORT SYSTEM MINE
-
THE MANAG-NT MODELS. Kaden I. Michels K. Tiemer
February
1986 CP-86-9
Cbllaborative P a p e r s r e p o r t work which has not been performed solely
at
t h e International Institute f o r Applied Systems Analysis and which has received only limited review. Views o r opinions expressed h e r e i n do not necessarily r e p r e s e n t those of t h e Insti- tute, i t s National Member Organizations, o r o t h e r organizations supporting t h e work.INTERNATIONAL INSTITUTE FOR APPLIED SYSTEMS ANALYSIS 2361 Laxenburg, Austria
PREFACE
The Regional Water Policies project of IIASA
w a s
focused on inten- sively developed regions where both groundwater and s u r f a c ewater
r e s o u r c e s are integrating elements of t h e environment. The r e s e a r c h w a s directed towards t h e development of methods and modelsto
support t h e resolution of conflicts within such socio-economic environmental systems.For t h a t reason Decision Support Systems have been developed and imple- mented for important
test
areas.The complex problems of such regional policy analysis
are
nottract-
able in one model using any of existing computational methods. That is whya
heuristic two-level model approach has been applied. Simplified first-level models t o g e t h e r with interactive procedures f o r multi-criteria analysis are used f o r screening analysis of rational long-term policies. The more comprehensive second-level models s e r v e f o r t h e verification and specifi- cation of t h e r e s u l t s of screening analysis. They a r e usedto
check t h e managerial feasibility of estimated strategies.One of o u r
case
studies deals with open-pit lignite miningareas.
The developed Decision Support System MINE has been implemented fora test
region in t h e Lusatian Lignite District of t h e GDR. The p a p e r describes t h e approaches f o r t h e second-level models (Management Model) of t h a t DSS.This r e s e a r c h h a s been done within t h e framework of a collaborative agree- ment between IIASA and t h e Institute for
Water
Management in Berlin. This p a p e r i s t h e final r e p o r t for t h e third (last) stage of collaboration.Sergei Orlovski P r o j e c t Leader
Regional
Water
Policies P r o j e c tABSTRACT
The Decision Support System MINE has been developed f o r t h e analysis of regional water policies in open-pit lignite mining
areas.
I t is based ona
two-level model approach. The first-level p l a n n i n g model is used f o r t h e estimation of rational s t r a t e g i e s of long-term development applying dynamic multi-criteria analysis. Therefor simplified submodelsare
used fora
rough time discretization (yearly time s t e p s and l a r g e r ) . The second-level management model considers managerial/operational a s p e c t s f o r s h o r t e r time s t e p s (monthly and yearly) employing more comprehensive submodels.I t i s a classic simulation model. For selected submodels stochastic simulation (Monte Carlo method) is used in o r d e r t o consider random inputs (e.g.
hydrological inflow and
water
demand). This model serves for t h e verifica- tion of s t r a t e g i e s obtained in t h e planning model, f o r t h e verification of simplified submodels used in t h e first-level model, and for t h e specification of strategies.Starting with t h e description of t h e position of t h e management model within t h e DSS MINE t h e s t r u c t u r e of t h e management model is given. The used submodels f o r s u r f a c e water/groundwater interaction and
water
qual- ity are described. In t h e Appendix computer subroutines ofsome
submodels are given being suitable for amore
general application.CONTENTS
1. Introduction
1.1 Background and Objectives f o r t h e DSS MINE 1.2 General S t r u c t u r e of t h e DSS MINE
1.3 The GDR Test Area
2. Stochastic Simulation of Management S t r a t e g i e s 2.1 Basics
2.2 Stochastic simulation of input d a t a 2.1.1 Hydrological inflow
2.1.2 Water demand 2.3 Management r u l e s
2.3.1 Balancing of
water
u s e r s 2.3.2 Remaining p i t management 2.4 Simulation of systems development2.4.1 Remaining pit submodel 2.4.2 Infiltration submodel 2.4.3 Balance submodel
2.5 Monte Carlo simulation and statistical evaluation 2.6 Numerical
tests
3. Deterministic Simulation of Long-term Policies References
Appendix
1. SUBROUTINE REMPIT
-
remaining pit submodel of t h e management model 2. SUBROUTINE INFI-
infltration submodel of t h e management model 3. SUBROUTINE BALANC-
balanc submodel of t h e management modelDECISION SUPPORT S Y S T M
-
THE MANAGEMENT MODELS.
ade en',
I.h!ichels2 and
K.~ i e m e r '
1. I n t r o d u c t i o n
1.1. Background and
Objectives f o rthe
DSS JllINERegions with open-pit lignite mining a r e characterized by complex and grave interactions in the socio-economic environmental system with special regard t o groundwater and surface water resources. To illustrate this f o r the German Demo- cratic Republic as t h e country with t h e greatest lignite production (about one third of t h e world production):
1.
The annual output of lignite amounts t o more then
300mill. tons/annum.
Thereby
it isnecessary t o pump out more then
1.7bill. mg/annum water f o r dewatering t h e open-pit mines. This amounts t o about
20% oft h e stable runoff of the 'whole country.
2.
The dewatering results in regional cone shaped groundwater depressions and consequently in extensive changes of the hydrological regime and of t h e con- ditions f o r water resources use and management, also in down-stream r i v e r basins.
- Infiltration losses of s&face water caused by mine dewatering reduce t h e water supply f o r down-stream water users and increase t h e groundwater pumpage necessary f o r dewatering
oft h e lignite mines.
- significant alterations of natural groundwater recharge
arecaused by t h e extensive changes of geographical and ecological conditions in open-pit mining
areas.For example, t h e natural groundwater r e c h a r g e of
atypical agricultural a r e a is changing under t h e climatic conditions of t h e GDR from about
200mm/yr. up t o
400mm/yr., Kaden et
al.1 9 8 5 ~ .
')~nternational I n s t i t u t e f o r Applied S y s t e m s A n a l y s i s Laxenburg, Austria q n s t i t u t e f o r Water Management, Berlin, CDR
-
The rate ofwater
pumped from t h e mining area into t h e s u r f a c ewater
system amounts t o about 30-50 % of t h e total r i v e r discharge (70% under low flow conditions).3. In lignite mining areas t h e groundwater quality and consequently t h e quality of mine drainage water, as well as t h e water quality in remaining pits i s strongly affected by t h e oxidation of f e r r o u s minerals (e.g. pyrite) in t h e sub- soil. With t h e n a t u r a l groundwater r e c h a r g e t h e oxidation products
are
flushed out, and t h e percolatedwater
becomes v e r y acid. Consequently t h e acidity of t h e groundwater increases. In t h e post-mining period t h esame
effect o c c u r s by t h e raising of groundwater t a b l e and t h e leaching of acid products.From t h e mentioned processes caused by open-pit lignite mining originate signifi- cant conflicts between different i n t e r e s t groups. Figure 1 illustrates t h e most important interdependencies between
water
u s e r s and t h ewater
r e s o u r c e s subsys-t e m s
in a n impact diagram.Nature
'y-L-y'
Figure 1: Water r e s o u r c e s impact diagram f o r lignite mining regions
Due t o t h e complexity of t h e socio-economic environmental prooesses in mining a r e a s , t h e design
of
regional water policies andwater
use technologiesas
w e l l as mine drainage can only done properly based on a p p r o p r i a t e mathematical models.From a c r i t i c a l analysis of t h e state-of-the-art of modeling in lignite mining areis i t h a s been concluded, t h a t above a11 methods and models are r e q u i r e d t o s u p p o r t t h e analysis and implementation of r a t i o n a l long-term r e g i o n a l w a t e r policies in open-pit lignite mining a r e a s , t o achieve a p r o p e r balance between economic
welfare and t h e
state
of t h e environment, Kadenet
al. 1985b.Towards t h a t goal t h e r e s e a r c h of t h e Regional Water Policies p r o j e c t of IIASA, in collaboration with r e s e a r c h institutes in t h e GDR, and in Poland, in t h e period 1984-1985
w a s
directed. One of i t s major products i s t h eDecision
SupportSystem MINK see Kaden
et
al. 1985a, 1985c, Kaden 1986.1.2.
General Structure ofthe
DSS MINEThe analysis of regional
water
policies in mining regions i s a problem of dynamic multi-criteria choice. I t i s embedded in a complicated policy making pro- cess. An advanced system of decision aids i s needed which allows:-
t o consider t h e controversy among different water u s e r s and i n t e r e s t groups,-
t o include multiple c r i t e r i a some of which c a n not b e evaluated quantitatively,-
t o t a k e into t h e account t h e uncertainty and t h e stochastic c h a r a c t e r of t h e system inputsas
w e l l as t h e limited possibilities t o analyze all t h e decisive n a t u r a l and socibeconomic p r o c e s s e s and impacts,-
t o o f f e r aset
of decision alternatives, demonstrating t h e n e c e s s a r y trade-offs between d i f f e r e n twater
u s e r s and i n t e r e s t groups.A t p r e s e n t no mathematical methods are available o r p r a c t i c a l applicable consid- e r i n g all t h e s e problems in o n e single model. E.g. t h i s holds
true
f o r a n y nonlinear stochastic multi-criteria analysis. Cmly hierarchic& model systems c a n satis*all requirements.
In general, dynamic problems of regional
water
managementare
approached by time-discrete dynamic systems models. The step-size and t h e available mathematical methods are t h e s t r u c t u r a l f a c t o r s of t h e n e c e s s a r y model h i e r a r - chy. Frequently a l r e a d ya
two-level m o d e l h i e r a r c h y satisfies most requirements.F o r t h e D S S MINE such a two-level system h a s been realized, combining
a
f i r s t - level Planning Model witha
second-level Management Model.The first-level Planning Model realizes a dynamic multi-criteria analysis f o r a relatively small number of planning periods , j 4 , .
. .
,Jas
t h e time s t e p f o r principalmanagement/technological
decisions. The time s t e p depends on t h e vari- ability in time of r e l e v a n t processes, on t h e r e q u i r e d c r i t e r i a and t h e i r reliability, and o n t h e frequency of decisions. A sa
compromise between a c c u r a c y and both, d a t a p r e p a r a t i o n and computational e f f o r t , f o r t h e D S S MINE v a r i a b l e time s t e p sare
used, s t a r t i n g with o n e y e a r and increasing with time up t o 15 y e a r s . This h a s been done taking into account t h e uncertainties in predicting model inputs and t h e r e q u i r e d a c c u r a c y of model r e s u l t s , decreasing with time.The planning m o d e l s e r v e s f o r t h e estimation of rational s t r a t e g i e s of long-
term
systems development. These s t r a t e g i e sare
selected by multi-criteria analysis considering a number of criteria. The c r i t e r i a have t o ' b e chosen from a givenset
of indicators, e.g. c o s t of water supply, cost of mine drainage, satisfaction ofwater
demand and environmental requirements. These indicatorsare
assumed t o b e integral values o v e r t h e whole planning horizon.In Figure 2 a block scheme of t h e planning model i s given. According t o t h e Figure t h e systems
state
i s c h a r a c t e r i z e d by state variables depending on previ- ous systemsstate
and by state descriptive parameters. Thelatter are
auxiliary p a r a m e t e r s with r e s p e c t t o t h e multi-criteria analysis but although r e s u l t s of t h a t analysis being of i n t e r e s t f o r t h e m o d e l user. Thestate
variables are t r e a t e das
control variables (decis!ons) in t h e multi-criteria analysis.
Indiators of systems dowlopmsnt O(
...,
l(j),D(j),Sv(j),Sd(j),..-
1<
m x 0--.. -.-.
Hydrological/socio- economic input
I ( i )
Figure 2: Block schema of t h e planning model
I
With t h e purpose of a unified model being independent on t h e chosen c r i t e r i a all indicators are bounded and t r e a t e d
as
constraints. Based on t h a t t h e following multi-criteria problem f o r a s u b s e t Ol L EL of t h e indicators 0 ( Q, , L
=I,.. . ,L)
i sdefined:
Or
=
Minimum !L
EL,^ (1.1) s u b j e c tto
inequality constraints0 s maxO (1.2)
& ( j ) S O , j = l ,
...,
J equality constraintsC , , ( j )
=
0 , j = 1 ,...,
J&(j)
-mu) =
0 , j = 1 ,...,
J boundsb j -1
The planning model as a first-level screening model i s based on a s e r i e s of more
o r
less s t r o n g simplifications in o r d e rto
obtain a manageable system being suitable f o r multi-criteria analysis. The major simplifications a r e :+
Staeoftbsyrtem
State descriptive parameters Sd(jl = fSd(j, I. D.SVI
4 7
S m variables j + . l +I '
+ t
r
ConstraintsI
Decisions D ( j ) ; D,
t w
S,,W =fsV(i,sv(j
-
11, D ~ s ~ ) .-
t h e discretization of t h e planning horizon into a small number of planning periods; all model data, e.g. decisions. s t a t e variablesare
assumedto
be con- stant within t h e planning period,-
t h e neglection of uncertainties in model inputs,-
t h e application of simplified environmental submodels based on comprehensive models,-
t h e neglection of relevant environmental subprocesses as t h e interaction between groundwater and surfacewater
depending on t h e surfacewater
table.That i s why a second-level h a g e m e n t
Model
for t h e simulation of systems behavior f o ra
l a r g e r number of smaller management periods (monthly and yearly time steps) is applied. I t is usedto
analyze managerial decisions by t h e help of sto- chastic simulation andto
verify results obtained with t h e planning model.In t h e given p a p e r t h e Management Model will be described in detail. This r e s e a r c h has been c a r r i e d out in t h e framework of t h e collaborative agreement between IIASA and t h e Institute f o r
Water
Management in Berlin, GDR.1.3. The
GDR T e s t A r e a
The DSS MINE h a s been developed with special r e g a r d
to a test
region in t h e German Democratic Republic. It is a n about 500 kme largearea
in t h e Lusatian Lig- nite District. A detailed description is given in Kadenet
al., 1985a. W e considera
planning horizon of 50 years, divided into 1 0 planning periods. In Figure 3a
scheme of t h etest
region i s depicted, illustrating t h e essential decisions on sys-t e m s
development.The following decisions are taken into t h e account (the indices are explained in Figure 3):
B.P
-
fluxf r o m
ato 6
C q a
-
supply of lime hydrate for water treatmentAtmd
-
duration of mine drainage mine D before starting i t s operationmaxh, -
maximumwater
level in t h e remaining pit The systems s t a t e is characterized by t h e following parameters '1:water table in t h e remaining pit
concentration of component 1 in t h e remaining pit L=I + f i e + , 1=2 -,
H +
s t o r a g e volume in t h e remaining pit.
groundwater flow
to
ainfiltration balance segment
As
a,p representative groundwater tableconcentration of component
L
in t h e flowto
a concentration of componeritL
in drainagewater
a f t e r treatmentflux/
water
tableat
balance profile bp a concentration of component L in t h e flux through balance profile bp aquantity of industrial waste water
concentration of component
L
in t h e industrialwaste water.
9
Paremetere t y p e d bold a r e s t a t e v a r i a b l e s o f t h e planning model.Figure 3: Detailed scheme of t h e
test
regionWe consider a planning horizon of 50 y e a r s , divided into 10 planning periods. The long-term development is above all determined by t h e mine drainage. This i s a con- tinuous process without relevant medium- and short-term (within t h e y e a r ) varia- tions. Therefor i t is assumed t h a t all decisions and systems descriptive values related
to
mine drainage are sufficiently described by mean values o v e r planning periods (or linear interpolated between planning periods).The systems variability within t h e y e a r s r e s u l t s from t h e hydrological inflow into t h e region and t h e fluctuating water demand. In this case t h e r e l a t e d deci- sions, s t a t e variables and systems descriptive values depend on managerial a s p e c t s
to
b e considered o n a monthly basis within t h e management model. In t h e Figure 3those parameters being of interest f o r t h e management model are special signed.
2. Stochastic Simulation of Management Strategies
2.1.
BasicsAccording t o t h e f i r s t simplifications t h e planning model considers principal
management/technological
decisions f o r estimated input values (expectation values). The feasibility of t h e estimated decisions is checked only in t h e mean f o r planning periods (by t h e help of constraints C,(j), C,, (j) and bounds, compare Eq. (1.2)-(1.4)).Problems a r i s e if t h e principal decisions are superimposed by managerial decisions f o r s h o r t e r time intervals, depending on t h e actual partly random sys-
t e m s
development. This is especially typical f o rwater
demand/supply. Both, t h e models f o r t h e actualwater
demand, and f o r t h e availablewater
r e s o u r c e s have t o consider autocorrelated and random components. Thewater
demand has t o b e satisfied according t o its variations between and within y e a r s . I t i s not sufficient t o satisfy t h e water demand in t h e mean o v e r planning periods. E.g.water
f o r sup- plementary irrigation is needed in t h e vegetation period but not even distributed o v e r t h e year.Consider t h e
water
u s e r s l,
l = I , .. .
,L with t h e water demand demr and t h ewater
supply supl. For t h e planning model t h e following c r i t e r i a i s used:Result of t h e multi-criteria analysis is some rational supply s t r a t e g y [ s u p l ( j ) ,
1
=l,...,
J , l =l,...,L 1.
This strategy has t o b e transformed by a n a p p r o p r i a t e management r u l e into t h e actualwater
supply strategy f o r all month k in t h e y e a r s i , i d l . . . , I: [ s u p i ( i ,k ), i =1,...
, I , k =1,...,
1 21.
The common c r i t e r i a f o r t h e satisfaction of water demand - f o r long-term
water
management and planning i s as follows:with i
-
y e a r , k-
month.Now it has t o b e checked whether t h e strategy obtained based on c r i t e r i a (2.1) satisfies c r i t e r i a (2.2). And this i s t h e f i r s t task of t h e second-level management model. By t h e help of stochastic simulation based on t h e Monte Carlo method t h e j k a s i b i l i t y of s t r a t e g i e s
is
verified and t h e s t r a t e g i e s a r e statistically evaluated.The model realizes t h e following steps:
1. Stochastic simulation of uncertain hydrological/socio-economic inputs, i.e.
inflow and water demand.
2. Simulation of monthly systems development based on stochastic inputs and management r u l e s considering rational s t r a t e g i e s estimated with t h e plan- ning model.
3. Statistical analysis of selected decisions,
state
variables, descriptive values and indicators f o r probabilistic assessment of t h e management strategy.The statistical reliability of r e s u l t s depends on t h e number of realizations of t h e Monte Carlo simulations and t h e d e g r e e of influence of stochastic inputs. In most c a s e s 100 realizations should b e sufficient. Nevertheless t h e numerical e f f o r t is high. For t h e 50-years planning horizon in this case t h e simulation has t o b e done
f o r 60000 month.
This a s p e c t has
to
be considered in t h e model f o r stochastic simulation as i t i s illustrated in Figure 4.Figure 4: Block scheme of t h e management model
p a r t 1
-
stochastic simulation of management s t r a t e g i e sHydrologiallcocio-
r ~ n o r n i c Input N m d k b l e uncmain hydro-
I U ) l o g l a l l ~ m i c input
O(i.k) =fO(l,k.O(i.k - 1)
....
1Only those decisions + ( i , k ) and submodels
are
included in t h e Monte Carlo simula- tion strongly depending on t h e stochastic inputs @ ( i ,k). For t h e remaining deci- sions, inputs and submodels t h e r e s u l t s of t h e planning model are used as mean values f o r planning periods. Based on t h e r e s u l t s of t h e planning model (decisions) a management ruLe+ ( i , k )
=
~ (,k ,+(i ,k - i ) , @ ( i ,k - i ) , r ( i ,k -1)) i (2.3)7
i s defined f o r t h e estimation of t h e m a n a g e r i a l d e c i s i o n s +(i ,k). Based on these decisions and t h e uncertain inputs O(i ,k ) t h e
state
variables r ( i , k ) are estimated ( f o r t h e management model is no needto
distinguish betweenstate
variables andstate
descriptive parameters):r ( i , k )
=
fI'(i,k,+(i,k), O ( i , k ) , r ( i , k ) , I ' ( i , k -1),...)
(2.4) Again, only thosestate
variables are considered strongly depending on uncertain inputs and managerial decisions.7 -
--
Stne of nM rymm ,
4
period j
smte of the rynsm y w r i, month k State -@tire pnametora
Sd(jl = fSd(i.l.D,Sv)
Unwrtain nrbbbs
1 + 1 b
-
Deciriom D(I).D,
b W-nt ruln
W(i,k) =W(i,k,Wi.k
-
ll,#(i,k - 1)+
(i,k - 1))
+
i.. . ...
-+
i,k + 1-
--
r1i.k) = flYi,k,*,O,r[i.k), r(i,k - 1). ... )
A i.k - 1-+
1 - 1
,
State warimbler1
~ " ( 1 ) = ~ ~ ( l . s ~ f i - ~ ) . D . s ~ )In Figure 5
a
simplified scheme of t h etest
region i s given illustrating t h e decisions, inputs andstate
variables being considered in t h e stochastic simulation(compare Figure 3). In compewrison
to
Figure 3a
few additional balance points have been introduced.0
Balance points0
Simulated inflow0
Fixed external inflow Water allocation from a to 8Figure 5: Simplified scheme of t h e
test
region f o r stochastic simulation qi2,3qc,s qd,s
bp 2 qs,i'qi,s REMAINING PIT
(RESERVOIR) TRIBUTARY
The major simplifications are:
1
@1,2
-
-
all mining activities are assumedto
b e constant during planning periods.S h o r t
term
variations in mine drainage and mine drainagewater
allocation are neglected.p
qi5,6
STREAM(S1 q s a
0,
P P
-
(0 P ' P
+
8
- water
quality processes are neglected; I t is assumed t h a twater
quality processes a r e damped and violations ofwater
quality requirementsare
less significant a s t h e dissatisfaction ofwater
demand int e r m s
ofwater
quantity.Water
quality alterations a r e above all caused by mine drainage, and t h a t i s taken as constant during planning periods.-
groundwater flow variations during planning periods are neglected dueto
t h e damped groundwater flow processes.That means
all
monthly varying water requirements haveto
b e satisfied from t h e stream.2.2. Stochastic simulation of input data
2.2.1.
Hydrological inflowThe inflow into t h e region is assumed
to
bea
natural hydrological process. A comprehensive analysis of s e v e r a l long duration time s e r i e s of runoff h a s shown, t h a t natural runoff under t h e climatic conditions of t h e GDR and with monthly time s t e p s posses t h e following properties, see Schramm 1975:-
i t i s nonstationary and cyclic with t h e period one y e a r ,-
i t s monthly one-dimensional distribution function can sufficiently w e l l b e approximated bya
transformed normal distribution function, e.g. a three- parametric log-normal distributionwith
Q -
mean monthly runoffX
-
transformed N(O,l)-distributed runoff q o,q, o-
parameters of t h e distribution function-
t h e process h a s Markovian c h a r a c t e r .Starting with t h e transformation
F
of t h e runoff and estimates of t h e auto- and cross-correlation a multi-dimensional runoff process i s simulated. General pur- pose programs f o r t h i s simulation:-
program SIKO f o r time s e r i e s analysis including parameter estimation-
program SIMO f o r runoff generationare
explained and listed in Kadenet
al., 1985c.For t h e
test
a r e a t h e t h r e e inflows qs,,
qs,, qs, haveto
b e simulated. This can not b e done directly because t h e balance points b p l , bp 5, bp7are
not identically with r i v e r gauges.In t h e
test
a r e a f o u r r i v e r gaugesare
located close t o t h e s e balance points.The - above mentioned assumptions have been proven for t h e runoff q
=
(q ,, q el Qs,
q J r through t h e gauges (30-years time s e r i e s of observation).Based on t h a t t h e parameters of three-parametric log-normal distributions have been estimated.
For t h e N(0.1) transformed runoff in t h e month k t h e following simulation model holds:
f o r k
=
1,. - .
,12 withA
-
0-
matrices of r e g r e s s i o n coefficients --
r e s i d u a l s t a n d a r d deviationE
-
M(0,l)-distributed random v e c t o r . The a c t u a l r u n o f f s are estimated by t h e retransformationf o r k
=
1,- . .
,12 withqo,s, -
p a r a m e t e r s of LN-3 distribution.The inflows qs q s 5, qs are weighted sums of t h e simulated runoffs, taking into t h e account t h e a c t u a l catchment areas.
Besides t h e simulated inflows in t h e s u r f a c e water system a few fixed inflows have to b e t a k e n into a c c o u n t (compare Figure 5 and s u r f a c e
water
balances in Kadenet
al., 1985a). Those are e i t h e r small t r i b u t a r i e s , waste or mine water allocations not being explicitely considered in t h e model system. Detailed informations on fluctua- tions of t h o s e inflows are not available. T h e r e f o r e t h e y are c o r r e l a t e dto
t h e simulated inflow of t h e r e s p e c t i v stream o r t r i b u t a r y , compare t h e balance model in Appendix 3.2.2.2. Water
demand
For t h e monthly w a t e r demand of any water u s e r t h e following g e n e r a l
sto-
c h a s t i c model may b e used:d e m ( i , k )
=
( t r e n d ( i , k )+
o s c i ( k )+
a u t o ( i , k ) ) . r a n d [ms/ s e c ] (2.9) witht r e n d ( i , k )
-
deterministic t r e n dosci(k)
-
deterministic oscillation component depending on typical seasonal behaviour of water u s e r sauto(i, k)
-
a u t o c o r r e l a t e d component r a n d-
random component (noise)In t h e
GDR test
region t h e following water u s e r have t o b e considered (compare Figure 3): municipal water supply (m), industrialwater
supply (i), a g r i c u l t u r a l water supply (ag), downstreamwater
u s e r (ds), environmental p r o t e c t i o n area (e).The a g r i c u l t u r a l water demand and t h e demand f o r environmental p r o t e c t i o n depend on t h e a c t u a l water t a b l e s in t h e s e regions. The models are given in Kaden
et
al.; 1905a. Random components are neglected.The model f o r t h e m u n i c i p a l w a t e r d e m a n d h a s been developed according
to
Eq. (2.9). The t r e n d i s d e s c r i b e d as a l i n e a r model, t h e a u t o c o r r e l a t e d component as a f i r s t o r d e r model. The oscillation component i s approximated by a Fourier- s e r i e s , see Kadenet al.
1985a.The industrial
water
demand is assumed t o be constant. Seasonal oscillation com- ponentsare
negligible. W e consider only a random component.The
water
demand of downstream u s e r s is slightly increasing in time. For t h e time being seasonal components are neglected.1 1
dem*(i,k)
=
[ 8.0+
O . 1 . i1
- r a n d [ms/sec] (2.12) Besides t h ewater
demand f o r downstream u s e r s a minimum flow has t o be guaranteed with r e s p e c t t o environmental aspects.For t h e r a n d o m component t h e following model is used:
r a n d
=
(1. - j a c . ~ ) (2.14)'E is a N(O.1)-distributed random number, j a c a scaling coefficient (for numerical
tests
j a c=
0.4 has been used).In o r d e r
to
realize a negative correlation between t h e water demand and t h e hydrological inflow (usually low inflow means d r o u g h t and consequently highwater
demand), f o r t h e random number E t h e number used f o r t h e stochastic inflow gen- e r a t i o n is taken.2.3.
Management rulesThe management r u l e s f o r managerial decisions have t o b e defined in o r d e r
to
satisfy t h e monthly varyingwater
demand of t h e above mentionedwater
u s e r s as good as possible. In case ofwater
deficits t h e u s e r s are ranked with r e s p e c tto
t h e i r socio-economic importance. The remaining pit can b e usedas
a r e s e r v o i rto
minimize deficits.2.3.1.
Balancingof
water useraFor t h e management model w e are only interested in t h e
water
requirementsto
t h e
stream
and t h e remaining pit. Keeping in mind t h a t t h e o t h e r supplycom-
ponents (mine drainage water and groundwater) are assumedto
be constant during a planning period t h e following balance equation holds:L
dq,,,(f , i l k )
=
d e m , ( j , i , k )-
a , , ( j ) (2.15).
L = 1with
dem, (j , i ,k )
-
total demand of u s e ru ,
planning period j , y e a r i, month k
q , , ( j )
-
supply component from s o u r c e 1to
u s e ru ,
planning period jd ( j i
,
k )-
demand of u s e ru
f o r water allocation from t h estream
( o r t h e remaining pit),in month k
,
y e a r i , planning period j.Based on this equation and on Figure 5 t h e balance equations f o r each u s e r can be given (downstream requirements can only be satisfied by t h e stream):
Thus, t h e supply requirements t o t h e
stream
and t h e remaining pit are defined.The extend t o whiqh t h e s e requirements are satisfied i s used as a c r i t e r i o n t o determine in how f a r t h e long-term s t r a t e g y estimated with t h e planning m o d e l c a n b e implemented under c o n c r e t e conditions with monthly o r seasonal fluctuations of d i s c h a r g e (inflow) and demand.
In case t h e requirements c a n not b e
m e t
in a given month t h e following lexico- g r a p h i c o r d e r i n g (ranking) i s considered:Highest priority: Municipal
water
supply Minimum downstream flow Industrialwater
supply Down-streamwater
supply Agricultural w a t e r supplyLowest priority:
Water
supply f o r environmental protection.A
more
detailed lexicographic o r d e r i n g might b e introduced splitting t h e u s e r s into s u b u s e r s as i t is usually b e dolle in long-termwater
management models,see
Kozerski 1981.2.3.2.
Remaining pit managementIn t h e
test
region a remaining pit originating a f t e r abandoning mine A in t h e planning period j , =7 c a n b e used as a r e s e r v o i r f o rwater
supply and flow augmen- tation.In o r d e r t o u s e t h e remaining p i t as a r e s e r v o i r
it
is n e c e s s a r yto
fill i t upto
t h e lower s t o r a g e limit. Both, t h e filling and t h e a c t u a l management of t h e p i t are c h a r a c t e r i z e d by extensive exchange relations between s u r f a c ewater
( t h e reser- voir) and t h e surrounding groundwater (aquifer).Due t o t h e s e exchange r e l a t i o n s t h e c o n t r o l of t h e filling p r o c e s s and t h e management of t h e remaining pit had
to
b e included as decisions into t h e planning m o d e l because of conflicting i n t e r e s t s between variouswater
u s e r s and t h e mining authority. Thewater
u s e r sare
i n t e r e s t e d in a n e a r l y usage of t h e remaining pitas
a r e s e r v o i r ; t h a tm e a n s
a f a s t a r t i f i c i a l filling of t h e pit. This, however, contrad- i c t sto
t h e i n t e r e s t s of t h e mining authority. An a c c e l e r a t e d rise of t h ewater
level in t h e remaining pit c a u s e s a considerable i n c r e a s e incost
of mine drainage in neighboured mines. Very illustrative examples f o r t h a t are given by P e u k e r tet
al., 1985.Depending on t h e p r e f e r e n c e s of m o d e l u s e r s t h e planning m o d e l w i l l g e n e r a t e some compromise solution f o r t h e remaining pit filling and management in
t e r m s
of mean values f o r planning periods.The obtained long-term s t r a t e g y h a s
to
b e transformed intoan
adequate monthly management r u l e , hoth, f o r t h e filling, and f o r t h e management.f i l l i n g phase
During t h e filling p r o c e s s t h e remaining pit
can
be considered as a commonwater
u s e r . The demand equalsto
t h e estimated allocation from t h estream to
t h e p i t dur- ing t h e planning periods.Instead of t h e constant values t h e monthly demand could b e interpolated between values f o r planning periods. In t h e given case t h e ''user" remaining pit is ranked between t h e agriculture and t h e environmental protection
area.
The management r u l e is t h e
s a m e
as f o r any user. The possible allocation is compared with t h ewater
demand. If t h e possible allocation i s l a r g e r then t h e demand t h e demand i s satisfied, otherwise a deficit o c c u r s (and is recorded f o r statistical evaluation). Thisrule
i s a "pessimistic" one, because any deficit can not be compensatedlater.
This meansfor
t h e remaining pit, t h a t t h e filling goal can not b e satisfied.In difference t o common
water
u s e r s f o r t h e remaining pit deficits can b ecom-
pensated, because a surplus of allocation can be realized (if i t i s available). In this case t h e allocation is not controlled by t h ewater
demand according t o Eq. (2.20) but by t h ewater
level.Define J + ( i , k ) t h e
water
level in t h e remaining pit estimated in t h e planning model (the monthly valuesare
obtained by linear interpolation between t h e solu- tions f o r planning periods). The monthly allocationto
t h e remaining pit i s aimed towards t h e realization of this water level. Therefor t h e required amount of inflow dq:,(i .k) has t o b e estimated in o r d e rto
increase t h ewater
level in t h e remaining pit from t h e actual value h,'(i ,k -1) in month k -1to
t h e goal4
(i .k ). Now t h e remaining p i t i s considered as a u s e r with t h e demanddq;'
(i ,k ).In Figure 6 t h e different outcome of t h e given management r u l e s is illustrated.
Both alternatives can b e checked with t h e simulation model.
Water table rem. pit
"mharge control led"
1
Artificial
-
recharge"water table controlled"
1 I I b
Year
Po.-- I I I
7 I
I I
---A I L,,-
[ m 3 / ~ . l
Figure 6: Management r u l e s f o r t h e filling phase of t h e remaining pit
Management phase
If t h e
water
level in t h e remaining pit has reached t h elower
s t o r a g e limit t h e pit c a n b e managedas
a r e s e r v o i r . T h e r e is a l a r g e amount of concepts and models f o r r e s e r v o i r management available, both, f o r flood protection, and f o r leveling ofwater
deficits. In t h e given study directed towards rational long-term s t r a t e g i e s t h e use of t h e s t o r a g e f o r flood protection i s less important. For long-term management modeling, periods of low flow conditions are considered as t h e signifi- cant events. A s a f i r s t alternative t h e following simple management r u l e is imple- mented:In case of downstream
water
deficits t h e r e s e r v o i r is used f o r flow augmentation in o r d e r t o level up t h e deficit. Any surplus of runoff in t h estream
i s used f o r filling up t h e remaining pitto
t h e upper s t o r a g e limit. Consequently, i t i sa
strategy of maximum s t o r a g e parsimony on t h e one hand and of possibly full compensating of deficits on t h e o t h e r one.But, t h e r e i s a significant difference t o common storages. Due t o t h e low groundwater table around t h e remaining pit caused by mine drainage in neigh- boured mines t h e r e s e r v o i r permanently loses
water to
t h e ground. The loss increases with increasingwater
level in t h e remaining pit. That means, high water levelsare
less economically not only because of lost discharge t o t h e pit, but although because of increased pumpage f o r mine drainage in neighbouring mines.In t h e time being management r u l e s should be studied taking this into t h e account.
Obviously a compromise between t h e reliability of satisfaction of
water
demand and t h e storage volume (thewater
level) hasto
b e found.2.4. Simulation of syaterns development 2.4.1. W i n i n g pit aubmodel
The submodel of t h e remaining pit has
to
describe t h e essential interrelations between t h e s u r f a c ewater
in t h e pit and t h e surrounding groundmiter in t h e filling phaseas
wellas
in t h e management phase. That is why t h e common balance equation f o r r e s e r v o i r management in its usual formwith
A S -
change of s t o r a g e volumeP -
precipitationZ
-
inflowE -
evapotranspirationR -
outflowcan not b e used h e r e . The equation has t o be extended by a term t h a t t a k e s into t h e account t h e infiltration into t h e groundwater o r t h e flux of groundwater into t h e r e s e r v o i r . This i s illustrated in Figure 7. I t demonstrates t h e effect of different management strategies during one y e a r on t h e
state
of t h e r e s e r v o i rat
t h e end of t h e year.The deviations depend on t h e difference between t h e
state
variables water Level remaining pit and water tabLe in the aquifkr. Consequently t h e dynamics of t h e groundwater system haveto
b e considered in additionto
t h e dynamics of t h e storage. This w a s done by computing management alternatives by means of a com- plex comprehensive groundwater flow model and by deducing a reduced grey-box model. The methodology i s described in detail by Kadenet
91.. 1985c. The result of model reduction are models f o r yearly and f o r monthly time steps. The monthly model has been developed in such a way t h a t i t provides t h e same r e s u l t sat
t h e end of one y e a ras
t h e yearly model if t h e inflow i s constant during t h e year.The monthly model needed f o r t h e management model has t h e following form:
with
Estimation of yearly values (yearly const. recharges)
.--0.-
... ... ..., 1
Estimation of monthly values Years--. ... -
4.2 Water level
remaining pit
Artificial recharge remaining pit
Figure 7 : Impact of different management s t r a t e g i e s on t h e water level in t h e remaining pit
$(i,k)
- water
level in t h e remaining pit [ m ] ,at
t h e end of month k in t h e y e a r ihi(i ,k
- water
level in t h e remaining p i t under n a t u r a l conditions (q,a), at
t h e end of month k , y e a r i!lp(i.k)
-
flux between stream and remaining p i t [ m 9 / sec], month i,
y e a r ka i * a p b o b i
-
parameters.Precipitation and evaporation a r e considered in t h e
water
t a b l e under n a t u r a l conditions. Their a l t e r a t i o n during month and due t o different s u r f a c e of t h e pit incase
of a c c e l e r a t e d filling a r e negligible.This model i s used in realizing t h e management r u l e s given in Section 2.3.2.
During t h e filling p h a s e t h e m o d e l above i s applied from month
to
month with t h e given inflow. For t h e management phase t h e remaining pit computation h a s t o b e divided into two p a r t s .In t h e f i r s t s t e p t h e
total
usable s t o r a g e volume i s imaginary added t o t h e n a t u r a l d i s c h a r g e in t h e s t r e a m and considered t o b e available f o r downstream o r o t h e r users. In o r d e r t o consider t h e exchange p r o c e s s e s between groundwater and remaining p i t t h e following algorithm h a s t o b e used, compare Eq. (2.22):c = h p ~ ( i . k )
+
a l . & ( i . k - 1 )+
a e . & , ( i . k - 2 )+
b l - q p ( i , k - 1 ) (2.24) withqiPf(i * k )
-
potential maximum d i s c h a r g e [ m 3 / s e c ] , month i , y e a r khp"" -
lower s t o r a g e limit [ m ]~p"p," -
maximum allocation capacity [ m a / s e c ] .After balancing off all u s e r s in t h e system t h e final r e s e r v o i r
state
i s computed interms
of t h e a c t u a l r e q u i r e d d i s c h a r g e t o t h e stream and t h ewater
levelat
t h e end of t h e studied month. This i s based on t h e minmum of t h e f r e e d i s c h a r g e (not used) qs &(i ,k ) of all balance points.If q s m b n ( i , k )
> c$yi(i
, k ) t h e r e i s a s u r p l u s of s u r f a c ewater
( i , k )=
qs ,h(i ,k )-
qr:(z ,k )to
b e allocated t o t h e stream. For t h e poten- t i a l maximum allocationto
t h e pit holds, analogously t o Eq. (2.24)with
q f $ ( i , k )
-
potential maximum inflow [ m 3 / s e c ] , month i , y e a r khpmX
-
u p p e r s t o r a g e limit [ m ]4:pa
-
maximum allocation capacity [ m 3 / s e c ] Using Eq. (2.24) and (2.25)w e
obtain:I
d q , , ( i , k ) f o r d q , , ( i , k )<
q!,;(i,k)q , , p ( i , k )
=
(2.26)q f $ ( i , k ) f o r d q , , ( i , k ) r q f $ ( i , k )
If qs*(i , k ) r q,~,:(i , k ) t h e r e i s a deficit of s u r f a c e
water.
For t h e d i s c h a r g e t o t h e s t r e a m holds:q p . , ( i , k )
=
q r : ( i , k )-
~ s * ( i , k ) ( 2 2 7 ) With Eq. (2.22) t h e finalwater
level in t h e remaining p i tat
t h e end of month k i s estimated.In Appendix A t h e computer program f o r t h i s submodel i s given.
2.4.2. Infiltration dmodel
The major p a r t of common long-term
water
management modeling i s t h e balanc- ing ofwater
u s e r s accordingto
t h e i r local distribution and lexicographic ranking in o r d e r t om e e t
t h e i rwater
demand. The ranking is considered by t h e help of a respective temporal sequence of balancing.This approach causes difficulties if t h e ranking sequence does not coincide with t h e local distribution (upstream user before downstream user) and if t h e satisfaction of t h e requirements of lower-priority upstream u s e r s effect t h e supply of downstream u s e r s with higher priority. And this happens if t h e
stream
i s characterized by discharge dependent infiltration losses ( o r base flow!).In t h e mining
test
regionwater
level alterations in thestream
cause signifi- cant changes in t h e infiltration and consequently in t h e discharge between various balance profiles. This process and ways f o r modeling have been discussed in detail by Kaden e t al., 1985a. 1985c. For each balance segment (compare Figure 3)black-box models of t h e foLlowing s t r u c t u r e have been developed:
with
Agi ( i , k )
-
infiltration betweenstream
and groundwater dueto
water level changes in t h estream
[m '/ sec ] h 8 ( i , k )-
actualwater
level in t h e stream over bottom(mean value f o r t h e balance segment) [m], y e a r 5 , month k
K8 (i
-
averagewater
level in t h e stream over bottom (mean value for t h e balance segment) [m1,
y e a r i (mean value of t h e respective planning period).
The impulse
u
, k ) f o r a balance segment [a, gl results f r o m t h e actual dischargeat
t h e respectivestream
profile. Since this profile might change between upstream and downstream balance point f o r t h e effective impulse a weighted mean i s used. The weighting f a c t o r i s y , OSySl,usuaLlyy=0.5.With Eq. (2.29) a feed-back between infiltration and changes in discharge is real- ized. The infiltration f o r t h e balance segment effects t h e upstream balance profile as a consequence of
water
level changes. Additionally external inflows/outflows within t h e balance segment haveto
be considered f o r t h e infiltration calculation.For t h e impulse
at
t h e downstream profile holds:with
q s ,(i , k )
-
dischargeat
t h e upstream balance profile, y e a r i , month kq i ,,&i
1 -
infiltration f o r t h e balance segmentat
meanwater
level, y e a r idq &i , k )
-
external inflow/outflowat
t h e downstream balance profile, y e a r i , month kI h 8 . p
- water
level key function f o r t h e downstream profile.The actual discharge
at
t h e downstream profile6
can be estimated iterativelyapplying Eq. (2.28)
-
(2.30). The above mentioned difficulty in balancing under con- sideration of infiltration becomes now obvious.In Appendix 2 t h e computer code f o r t h e infiltration submodel of t h e management model is given.
2.4.3. Balance submodel
The management rules, t h e submodels f o r input simulation, and t h e above given submodels f o r t h e remaining pit and the infiltration have
to
b e combined f o r balancing t h e surfacewater
resources.Due t o t h e discharge dependent infiltration t h e balancing of the e n t i r e system i s only possible by iterative computation. The infiltration in all balance segments has
to
be estimated before balancing of allwater
users. However, through t h eterm
dq ( i , k ) in Eq. (2.30) t h e infiltration depends on t h e u s e r balancing.The following algorithm has been developed assuming that in t h e majority of realizations t h e requirements of
all
u s e r s can be satisfied. I t consists oft w o
s e p a r a t e balance computations:1. balancing of t h e surface
water
system upstream-downstream f o r a given actualwater
demand of all users, considering t h e infiltration,2. balancing of
water users
according to t h e i r lexicographical ranking.The computation
starts
with procedure 1 f o r t h e givenwater
demand. If t h e water demand i s fully satisfied,all
parameters have been exactly estimated and an itera- tion i s redundant.If t h e
water
demand i s not satisfied, in procedure 2 t h e available r e s o u r c e sare
distributed between t h e users according t o t h e i r priority. After t h a t in procedure 1 t h e system i s balanced considering t h e reducedwater
demand from procedure 2.This computation i s continued iteratively until t h e discharges f o r all balance pro- files do not change during iteration (within a given accuracy).
In Appendix 3 t h e computer code of the subroutine balance of t h e management model is given.
2.5. Honte Carlo simulation and statistical evaluation
The major problem related
to
t h e Monte Carlo simulation i s i t s high computa- tional e f f o r t depending on t h e number of realizations NREL. Frequently fixed numbers of realizations (e.g. NREL=100) are selected. In thiscase
NREL usually will be overdimensioned in o r d e rto
ensure a certain statistical evidence and t h e numerical e f f o r t will be higheras
necessary.Principally, t h e number of realizations depends on t h e required statistical evidence of t h e results. If this evidence i s checked in the course of t h e computa- tion, t h e simulation can be stopped
as
soon as possible. For t h eDSS
MINE such atest
has been realized in t h e following simple f o r m :Every 1 0 realizations t h e mean values f o r selected parameters
5
(decisions andstate
parameters f o r each planning period)are
compared. If t h e deviation is smaller then E with r e s p e c tto
t h e mean value of t h e planning modelzp
t h e simula- tion is stopped.According
to
t h a t t h e number of realizations i s controlled by t h e factor E t o be fixed by t h e u s e r (e.g. 0.05).In Figure 8 a simplified flow chart of the subroutine controlling the Monte Carlo simulation is depicted.
t
initialization of arrays for registrationc : for all realizations c: for all planning periods
t
input of results of .planning model for period jI
7 i = ib(j),ie(j) c : for all years of period j-
k = 1,12 I c: for all months of year i c : monthly water balancet
registration ofmonthly parameters
t
registration of satisfaction water demandt
registration of densityfunction of parameters
t
estimation of yearly parameterst
estimation of parameters ofplanning period
t
estimation of total indicators1 +
registration of total indicatorsno
stowtest: statistical evidence of results sufficiently 7
$
standardization of statistical resultst
estimation of distribution function of total indicatorsFigure 8: Flow chart of Monte Carlo simulation
An important methodological problem i s the registration and statistical evaluation. The following types of registration a r e common:
-
distribution functions of reliabilities of t h e o c c u r r e n c e of defined events, e.g.t h e satisfaction of
water
demand,-
distribution functions of selected parameters, e.g. t h e total c o s t of mine drainage,-
density functions of selected parameters, e.g.water
allocation.These continuous functions c a n only b e r e g i s t e r e d empirically in a d i s c r e t e form f o r defined classes. Define ( s c l a s s (l ),1 = I , .
. .,KC)
t h e scaling f a c t o r of t h e classes, f o r convenience t h esame
f o r all functions (e.g. 0.-
0.5-
0.7-
0.8-
0.9-
0.95).The r e g i s t r a t i o n depends on t h e definition of t h e r e f e r e n c e value. I t i s advan- tageous
to
usea
r e f e r e n c e value being known in advance. In t h i s case t h e statisti- cal events have not t o b e s t o r e d and t h e empirical functions c a n b e estimated dur- ing t h e simulation. This i s necessary especially then ifa
l a r g e number of parame-ters
i sto
b e r e g i s t e r e d , e.g. f o r t h e GDRtest area
2 1 p a r a m e t e r s f o r 1 0 planning periods and 1 2 month. (1.e. f o r 100 realizations 252000 values.)For t h e
DSS
MINE t h e following empirical probabilistic functionsare
estimated:S a t i s f a c t i o n @ w a t e r d e m a n d
with dem
- water
demand, sup.-
water supply.This function i s estimated f o r all month in all planning periods. In t h i s case w e use probabilities as t h e r e f e r e n c e values f o r r e g i s t r a t i o n and sup i s normalized by dem, consequently t h e d i s c r e t e function values can b e estimated during stochastic simulation.
I n d i c a t o r s of s y s t e m s development
PI( ind0 )
=
R o b [ ind<
ind0 jA l l events have t o b e s t o r e d . The empirical distribution function i s estimated
at
t h e end of simulation. The probability i s scaled with s c l a s s , see above.F b r a m e t e r s of s y s t e m s development
For decisions,
state
p a r a m e t e r s andstate
variablesa
density function i s estimated.p p ( p a r O ) = R o b [ p a r 0 s p a r < p a r 0
+
Apar j (2.34) The function i s estimated f o r all planning periods and all month. A s t h e r e f e r e n c e valuew e
use t h e known r e s u l t of t h e planning m o d e l (mean value f o r e a c h period).In Figure 9 t h e above defined probabilistic functions and t h e i r scaling are illustrated.
2.6. Numerical tests
Basis of t h e numerical
test w a s
a typical r e s u l t of t h e planning model obtained f o r a multi-criteria analysis f o r t h e c r i t e r i a :dev -m
-
deviationwater
demand/supply municipality dev i-
deviationwater
demand/supply industry cost - m i- total
mine drainage c o s tcost --m
- cost
municipal w a t e r supply cost i-
c o s t industrialwater
supply.F o r e a c h c r i t e r i a t h e utopia-point h a s been selected as r e f e r e n c e point. These r e s u l t s have been used
as
initial values f o r t h e management model.a) Reliability of satisfaction water demand
4
Scaling:b) Distribution function of indicator
C) Density distribution of parameter
/
I
I
Scaling :--
/ / / /
par : (sclass(l), I = 1
,...,
KL;-
(2-sclars(l)), I = KL,
...,
2) r par/
/
/ Scaling:
/
/ p, : (sclass(l), I = 1
,...,
KL)Figure 9: Statistical evaluation
1
+
indThe simulation has been performed with 60 realizations. Therefor about 6 Min.
CPU-time