Incorporating Geological Uncertainty to define Weathering Contacts and grades in

Medium- and Short-Term Scheduling

Euler Victoria, Mine Planning Engineer, EUROCHEM S.A

Rodrigo Peroni, Professor, Mining Engineering Department, Federal University of Rio Grande do Sul.

ABSTRACT

Knowing the depth of contacts between weathering zones in phosphate mines is determinant to define the
stripping ratios and the mineralized zones. Contacts are gradational and difficult to define even in the
drillhole samples, so the combination with geostatistical simulation techniques helps to define the contact
zone and the grade distribution, providing a way to assess the probability of the occurrence and the
positioning of contacts between the barren waste and the ore zones. These inputs are used in the
scheduling, anticipating risks associated with mine planning scenarios. The results from an application to
a case study showed a good improvement when compared to traditional techniques.

INTRODUCTION

There is a growing need for a better understanding regarding some limitations of the traditional mine
planning methods based on deterministic models, as well as in the optimization approaches and their
potential effects on the project’s decision-making process. Surprises in the operational routine of open pit
mines caused by variability associated to grades, geological models, operating costs, among other
variables that impact cash flow may not be foreseen using conventional planning techniques. When using
models where the uncertainty of the estimates is not provided, it becomes difficult to address potential
variabilities of these attributes, ignoring the risk inherent to the nature of the data, causing a deceiving
certainty for the entire production chain.
In the construction of models, the knowledge of the average value of a block or a mining panel, and
consequently of the mining plan adopted to extract this panel may not be decisive, but access to multiple
mining alternatives generated by equiprobable models can be of great value in the risk and financial return
analysis of this mining panel (Peroni, 2002).
In long-term mine planning, according to Dimitrakopoulos (2002), the presence of geological risk in
mining projects is well known. During the last few years, some technological advances allowed to access
and model the geological risk. As a result of these approaches, several issues were raised, including the
integration of grade variability in pit optimization, technologies and algorithms for modeling geological
uncertainty in final pit optimization and production scheduling.
The generation of multiple equiprobable models allows us to equate the problem of evaluating reserves
and mine planning according to probabilistic approaches. The transformation of a model simply
containing grade/quality values of the parameters is evaluated in economic terms by a profit function that
incorporates all costs and benefits arising from a block. Comparison of the response produced by different
stochastic simulation algorithms is performed by delineating the final pit using a traditional algorithm and
sequencing it within the phases and operational advances created by the algorithm (Peroni, 2002).
A more recent approach from an operational point of view, according to Capponi (2019), would be the
selection of ore and waste, in the short-term mine planning stage, which can be considered as one of the
most important decisions to guarantee the sustainability of the mining enterprise. This decision becomes
even more complex due to the uncertainty inherent to the quality of the available mineral resource, which,
in turn, is a consequence of limited knowledge about the real grades in each sector of the deposit.

1
METHODOLOGY

The first step for the development of this study consists in working the database in the following ways:
regularizing the sample support, determining the geostatistical domains, delustering the samples and
finally determining the spatial continuity. Geostatistical simulations are carried out to model and assess
the uncertainty of P2O5 grade, weathering contacts, and ore/waste classification, according to a cut-of
grade limit.
The Turning Bands method is used to obtain simulations of P2O5 grades, due to the continuous nature of
the variable. The weathering contacts will be obtained through sequential indicator simulation (SISIM)
since they are categorical variables. The methods applied here can be applied to any other mineral deposit,
since continuous and categorical variables are present in any type of deposit, whether in the form of
grades, lithologies and typologies.
After generating the simulations, those models are compared against the reconciliation results of the areas
already mined, helping in the interpretation of the stochastic variables and allowing the construction of
modified mining phases by grouping the regions that show clusters of blocks with spatial and economic
similarities. With the construction of the phases, combining conventional methods and knowing the
potential variations of geological classification, the blocks are sequenced using deterministic and
probabilistic approaches. For a five-year mine plan, the variables that will represent the risks will be used,
and with them to propose the lowest possible risks, for the first years of the plan, in addition to the use of
other conventional variables. Figure 1 schematically outlines the activities to be developed. To illustrate
the proposed methodology, the technique will be applied to a large open-pit phosphate mine.

Source: Author's.

Figure 1. Workflow to implement the method.

To understand the contribution of each of the variables used in the risk assessment in mine planning, the
classification flowchart of materials as ore and waste is shown in Figure 2.

2
Source: Author's.

Figure 2. Ore and waste classification flowchart.

The cut-off grade used is 3.5% of P2O5, based on the result of balancing total costs and revenue, materials
that are below this value but have content above 2.2% and are not classified as fresh rock and as inferred
resource, will be identified as Potential ore.

RISK ANALYSIS COMPOSITION

To establish methods to quantify the uncertainty associated with simulations of geological variables, some
uncertainty indices can be addressed to provide the probabilities of achieving a certain parameter (Koppe,
2009; Pilger et al., 2000). The probabilities will be used during the mine sequencing in three ways,
depending on the study variable:

 P2O5 grades, which will be determined by the probability of the grade estimated by ordinary
kriging to be greater than the result obtained after processing the material (reconciliation);
 Ore/Waste contact zone definition based on grades above or below the cut-off grade (COG);
 Weathering Class, split into four categories where each one will be assigned a probability of
occurrence to be a certain class or not.

Fifty equiprobable scenarios were generated using Turning Bands (TB) simulation to quantify the
variability of the P2O5 grades. During the validation process, the data variography and statistics were
checked and reproduced to accept the models for each domain divided by weathering zones.
For the categorical variables, we used the Sequential Indicator Simulation (SISIM) to calculate the
probability of each block so that they have a grade above or below the COG, where the fifty models were
filled with values 0 and 1, where the value 1 represents a value greater than the COG and zero represents
values lower than the COG.
The SISIM will also be used to determine the weathering category. They will be subdivided into four
categories, according to the weathering degree; the four weathering zones, which, from top to bottom of
the deposit, are:

3
  Alloterite (ALOT), the region closest to the surface with the highest weathering level, is
considered to be ore but with the worst metallurgical performance;
 Isalterite (ISALT), the region just below ALOT, also considered to be ore, is the most abundant
in the reserve, excavated mechanically;
 Altered Rock (AR), region just below the ISALT, presenting some characteristics of fresh rock
but still able to be mechanically excavated;
 Fresh Rock (FR), due to the higher mechanical resistance, this material is not mined and
consequently not being treated as an ore region, and that is why modeling the contact zone
between AR and FR is important.

Therefore, the variables that will represent the probabilities are described below:

 PR_P2O5: is the variable that represents the probability of the grade to be greater than the
estimated grade, obtained by the TB method. The variable is calculated according equation (1):

𝑃𝑅_𝑃2𝑂5 = (1)

Where:
Nm= Numbers of simulation where the results are greater than the estimated grade;
Ns= Number of simulations.

 SI_RF: contains information about the probability of a given block to belong to the FR
weathering zone, obtained from SISIM when evaluating the contact between ISAL, ALOT and
AR;
 SI_RA: contains information about the probability of a given block to belong to the AR
weathering zone, obtained from SISIM when evaluating the contact between ISAL, ALOT and
FR.
 SI_ALOT: contains information about the probability of a given block to belong to ALOT,
obtained from SISIM when evaluating the contact between ISAL, AR and FR;
 SI_COG: For the risk assessment variable of erroneously classifying the block based on the COG
through the SISIM.

Figure 3 shows the risk analysis flowchart, as well as the interpretation of what can be considered as a risk
from the mine planning perspective, considering the beneficial results that can be assumed as
opportunities for improvements, when misclassified waste blocks reveal to be ore blocks.

4
Source: Author's.

Figure 3. Risk analysis flowchart

DEFINING WEATHERING CONTACTS

Using the results of the accumulated probabilities of belonging to each weathering zone category
combined with field data from a mined-out region, where the contacts are known, an interpretative study
of risk /opportunity was created to improve the model of the contact zones. Figure 4 shows the regions of
ore losses (in blue) and the regions of gains (in red). The losses came from the projection of friable
material, originally classified by the long-term geological model as ore that revealed to be fresh rock (FR),
thus classified as waste. On the other hand, material that was classified as FR and turned out to be friable
material will be the gains of ore mass. In both situations, analyzes of the probability distributions were
carried out for the two regions, and the results were compared with the reality mapped in the field. This is
not a result from the probabilities, this is a result from the field mapping at the pit bottom determining the
differences observed in terms of errors from the old modelling against the reality.

5
Source: Author's.

Figure 4. Map of gains and losses of ores, assigned to the RF contact.

The solids were obtained by the intersection of the current topography and the final pit design. The
material obtained above the final pit design was considered an addition, and the material below the pit was
considered as losses; the area analyzed corresponds to 12% of the total area of the five-year mining plan.
When comparing the probability of blocks classified as friable material (other than FR) present in the
reconciliation solid related to losses, we observe an increase in the variable SI_RF; in the average of the
reserve presents a 10.3% chance to be FR (Figure 4) changing to 19.6%, thus presenting good capacity to
anticipate negative results, which can be treated as a risk according to the variable.
To analyze the region where volume gains were observed, the SI_RA variable is used. Inside the red solid
(Figure 4), the blocks classified as FR in the model obtained an average probability value of 45.7%, while
in the reserve the same group of blocks have an average of 23.7%, also showing a good ability to be
monitored by the mining plan.
Regarding the ALOT, it presents lower processing performance when compared to the other mineralized
typologies, so the number of ore coming from this domain must be limited. The value that will be
interpreted as the ideal average value for a block not to be ALOT will be the average value of the reserve
of the result of SI_ALOT regarding the blocks not classified as ALOT.

P2O5 GRADE VARIABILITY

When analyzing the variation of P2O5 grades, the objective is to anticipate impacts on the production of
concentrate. The stochastic data used will again be the probability; however, as the data analyzed is
continuous, the purpose is to calculate the chance of the estimated grade via ordinary kriging to be higher
than the real grade, assumed as true by sampling the plant feed grade. Figure 5 allows the analysis of, how
the probability distribution behaves across the reserve (represented by a 3D accumulated block model in a
plan view) against the drilling pattern.

6
Source: Author's.
Figure 5. Probability map of the estimated P2O5 grade to be higher than the actual grade.

For a better understanding of the purpose of this assessment, the model was compared to the reconciliation
of grades in a monthly basis, and the results can be observed in Figure 6. It shows a comparison between
the estimated model (obtained by ordinary kriging) and the real results, assuming the results indicated by
the plant as the true values. The graph in Figure 6 shows the monthly results observed by the mine
operation on the horizontal axis, while the vertical axis on the left shows the result of the P2O5 grade of the
theoretical and actual model (red line actual result, yellow line estimated model). On the right-hand axis,
the results of the of the accumulated probabilities (bar graph) of the value estimated by ordinary kriging
being greater than the value reported by the plant. When examining the months of May, June, and August
in Figure 6, the probabilistic model is correct when predicting a chance of the actual grade to be lower
than the estimated grade, since the probability shown is less than 50%, which represents the probability
that the actual grade is greater than the estimated grade, values close to 50% indicate low error values in
relation to the actual content (3% relative error) and the estimated one, as shown in the graph . For a better
construction of the method, more evaluations should be performed, with an extensive database and
consisting of more periods, in order to calibrate the simulation method used for risk analysis purposes.

7
Source: Author's own.

Figure 6. Result of reconciliation of P2O5 grade against stochastic data.

To improve the analysis, the method must be maintained, assuming that a larger database will help to be
more representative and to calibrate the results given by the probabilistic model. The SI_COG variable
will be used to signal waste blocks that have the potential to become ore, due to the high probability that
the P2O5 grade is greater than the COG. It is then agreed that waste blocks classified exclusively by the
COG and presenting values greater than 50% of the SI_COG variable will be treated as candidates to
change their classification, representing a potential addition of ore.

MINE SCHEDULING RESULTS

The phases were built with the aim of carrying out the medium-term sequencing in an annual basis,
detailing the mining horizons closest to the sequencing. The map showing the distribution of the phases
can be seen in Figure 7, the colors show the distribution of the mining phases after operationalizing the
pushbacks and combining with the physical limits imposed by the modifying factors (mining claims,
environmental permits, landowner, etc.).

Source: Author's.

Figure 7. Plan view showing the mathematical all the phases for the long term mine planning.

8
Phases 1 and 2 correspond to the areas covered by the medium-term drilling, and the others are only
present in the reserve sampled by the long-term drilling.
The positive effects of infill drilling can be observed in Table 1, considering the variable PR_P2O5
values, which are close to 50%. These values show that the methodology presents a low risk of obtaining
very different results from those estimated for P2O5. The analysis for the SI_COG variable demonstrates
that, throughout the lifespan of the mine, there are few chances of ore misclassification concerning the
grades, but some opportunities can be observed for waste blocks in phases 1 and 2, indicating values close
to a 60% probability that the P2O5 content is greater than the COG.
The categorical variables SI_RF and SI_ALOT are influenced by the shape and depth of each phase
region. For example, deeper phases with larger contact areas on the pit bottom tend to have higher values
for SI_RF, while the upper regions, closer to the surface, present higher values for the SI_ALOT.

Table 1. Characteristics of the phases

Phase 1 2 3 4 5 6 7 8 9 10 11 12 13 TOTAL
Mass (Mt) 1.5 7.9 4.4 16.4 6.8 7.8 5.4 16.6 16.7 22.8 54.5 25.6 32.8 219.2
P2O5 (%) 3.9 4.9 4.3 4.5 3.9 4.9 4.1 4.8 4.5 4.7 4.4 4.5 4.8 4.6
SI_COG (%) 56.8 69.1 57.4 59.2 49.6 60.2 56.7 60.8 55.4 59.4 56.5 56.8 57.8 57.9
Ore

SI_RF (%) 6.9 6.6 1.8 9.2 7.8 6.7 16.7 11.5 5.4 13.8 7.9 8.5 6.6 8.6
SI_ALOT (%) 9.1 8.8 7.0 7.0 10.1 6.0 5.9 8.1 7.0 8.2 6.2 12.1 9.4 8.1
PR_P2O5 (%) 46.6 46.9 40.4 42.3 41.5 41.5 37.0 39.6 41.9 40.2 38.5 43.8 36.5 40.8
Mass (Mt) 0.4 0.5 1.9 4.8 5.9 3.9 5.9 4.8 17.6 11.5 29.0 11.4 16.9 114.4
P2O5 (%) 3.6 3.9 3.2 3.3 3.3 2.7 2.7 3.3 2.6 3.1 2.8 3.1 2.7 3.0
Waste

SI_COG (%) 59.9 68.8 45.6 47.1 43.8 34.5 38.7 50.0 34.5 43.1 38.4 42.5 30.3 40.8
SI_RF (%) 6.8 4.3 1.9 6.0 5.4 6.8 13.0 7.4 4.6 16.4 5.7 7.0 4.7 7.0
SI_ALOT (%) 14.8 24.2 15.0 12.8 15.2 11.4 10.3 21.6 19.7 13.3 13.1 16.9 24.8 16.8
PR_P2O5 (%) 48.3 32.5 49.2 41.6 51.0 49.0 49.1 41.3 48.2 48.8 46.3 42.2 48.1 46.8
Source: Author's.

Based on simulation results and interpretations, a mining plan will be built. It shall consider variables that
quantify the risk, using targets to be achieved for each period and meeting the conventional objectives of
the mining plan. It must also supply ore to the plant following the requested parameters that generate
fewer surprises based on risk assessments for the first periods, keeping in mind that in future attempts,
actions should be taken to either reduce the risk or minimize its impact. Hence, the values given by the
variables SI_ALOT and SI_RF were forced to a more homogeneous distribution between the periods, as
the drilling densification is planned just by searching higher values of P2O5. It must be noted that the
coefficients of variation related to weathering are much higher than the grades, which results in a more
erratic behavior, demanding greater attention to ensure better assertiveness with lower uncertainty.
The main assumptions to obtain a medium-term mine plan, based on the controlling variables created,
were:

 PR_P2O5: When controlling this variable, we tried to standardize the results around 50%, getting
good results if the results are between 40% and 60%. For the first year, the acceptance was more
rigorous, due to the results presented in the reconciliation, the values of the variable closer to
50% indicate fewer surprises in the P2O5 result; in the following years, these results can be better
investigated, through running more drilling campaigns, for instance;
 SI_RF: For this variable, it is intended to maintain the average per period around 10% for the ore
blocks; this reference value was determined considering the results of the weathering model
reconciliation;
 SI_RA: This variable was treated as an indicator of waste blocks that could become ore blocks
due to an error in weathering classification; this variable is not used in the sequencing, but can
indicate some opportunities;
 SI_ALOT: The objective of this variable was to control an excessive contribution of ALOT in the
sequencing, trying to keep the values close to 11%;
 SI_COG: for this variable, we tried to identify potential classification switch from waste blocks
to ore when values were greater than 50%.

9
The purpose of the plan was to sequence in a medium-term timespan, determining the first five years, with
the first year split into monthly plans. The sequencing for the ore blocks is not considering the SI_COG
and SI_RA fields, because the SI_COG variable is redundant to the PR_P2O5 analysis, already
determining the chances to be above or below the COG and the reconciled grade and will only be used for
waste blocks. The SI_RA variable will not be used for the ore evaluation, the objective is only to control
the contaminants, which are measured by the SI_ALOT field, and the non-mineralized contacts
represented by the SI_RF field. The scheduled advances can be seen in Table 2.

Table 2. First year schedule result, on a monthly schedule

Period (Month) 1 2 3 4 5 6 7 8 9 10 11 12
Mass (Mt) 415,0 415,0 415,0 415,0 415,0 415,0 415,0 415,0 415,0 415,0 415,0 415,0
P2O5 (%) 4,5 4,5 4,5 4,5 4,5 4,5 4,5 4,7 4,9 4,9 4,8 4,7
Ore

PR_P2O5 (%) 44,9 47,5 49,3 46,7 45,1 50,3 45,7 50,6 46,6 44,0 48,5 47,4
SI_ALOT (%) 7,2 12,0 11,4 10,8 5,5 8,9 11,0 12,6 4,5 10,5 6,7 5,5
SI_RF (%) 5,4 8,2 3,4 5,4 7,1 3,7 3,6 2,5 5,5 4,6 4,2 6,9
Mass (kt) 110,1 93,1 57,8 60,4 50,4 45,3 47,0 49,8 23,0 5,8 61,3 29,4
Waste

SI_COG (%) 50,1 41,4 49,2 52,5 34,3 40,9 43,8 41,3 36,2 43,7 0,0 38,4
SI_RA (%) 17,0 30,9 25,3 15,2 20,7 40,5 12,7 58,5 38,3 19,1 0,0 67,9

Source: Author's.

When analyzing the result of the plan in year 1, three relevant points can be highlighted:

 Chances of the P2O5 grade being lower than indicated in months 1 and 10 (PR_P2O5 ≤ 45%), in
the remaining months, the results are closer to 50%, but there is a slight chance that they may be
lower than the P2O5 grade reported in the mining plan;
 Greater potential to exceed ALOT feeding at months 2, 3, 7 and 8 (SI_ALOT ≥ 11%);
 Low risk of appearance of FR blocks instead of weathered blocks (SI_RF ≤ 10%).

The analysis of the SI_RA variable for months 6, 8 and 12 indicate good chances of reclassifying blocks
from FR to AR as they have values above 40%; such deviation was positive, based on the reconciliation
results. The sequencing of the first year was limited to phases 1 and 2, regions with the highest density of
holes.
Figure 8 shows the monthly scheduling, limited to phases 1 and 2, assuming those regions are covered by
the medium-term drilling campaign.

10
Source: Author's.
Figure 8. Monthly mine plan, year 1.

For the remaining planned years (Figure 9), there was a need to exceed the limits of the medium-term
drilling campaign, and the possible impacts can be seen in Table 3, as well as the general result of the
plan.

Table 3. Scheduled results for the first five years on an annual basis.
Period (Month) 1 2 3 4 5

Mass (Mt) 4.980,0 9.000,0 12.000,0 12.000,0 12.000,0

P2O5 (%) 4,6 4,5 4,5 4,6 4,8
Ore

PR_P2O5 (%) 47,2 43,4 43,7 40,2 42,4

SI_ALOT (%) 8,7 8,0 6,9 7,6 9,2
SI_RF (%) 5,0 5,0 6,3 7,0 7,0
Mass (kt) 633,3 4.527,8 4.975,9 7.913,3 6.309,7
Waste

SI_COG (%) 44,8 39,2 39,5 40,5 37,9

SI_RA (%) 20,8 17,3 26,1 18,9 24,9
Source: Author's.

11
Source: Author's.
Figure 9. Five-year mine planning.

The need for drilling can be observed from the second year of the mining plan, due to the values close to
40% of the variable PR_P2O5. In the first year, the plan sought to obtain higher values for PR_P2O5,
while for the remaining variables, the values remained within the proposed limit.

CONCLUSIONS

The application of geostatistical simulation has proven to be effective in evaluating the uncertainty related
to geological data., weathering contacts and grades, since no additional drilling or sampling was required
other than those already used by the conventional geological model and estimation. The simulation results
used in conjunction with the reconciliation process demonstrated the ability to validate the models and to
provide uncertainty information about the targets to be achieved in the mining sequencing. The results
also demonstrated that the application of the proposed method allowed to redirect the planning, based on
the grade variability and errors of contacts detected by the weathering model after the use of the simulated
variables, and also confirmed by the mine operation, reducing risks for the first periods. If the risks were
not mapped and avoided, or at least mitigated with other palliative actions, the mine plan would deliver
less adherence with the operation and lower performance. For example, if blocks previously identified as
ore are, when exposed during excavation, in fact FR blocks (waste), the operation is forced to look for
tonnages and grades out of the planned regions to meet the requirements demanded by the processing
plant.
The use of this approach in the medium and short terms also stimulated us to understand the need for
improved knowledge on areas beyond the 5-year plan. This means that when the values of probabilities
presented are not acceptable, it allows for an anticipation of actions to change the strategic plan and to
make better decisions.

12
REFERENCES

CAPPONI, L. N. Introdução de parâmetros de controle de incerteza para o planejamento de lavra. 2012.

148 p. Dissertação (Mestrado) - (PPGE3M), Universidade Federal do Rio Grande do Sul, Porto Alegre,
2012.

DIMITRAKOPOULOS, R.; FARRELLY, C. T.; GODOY, M. Moving forward from traditional

optimization: grade uncertainty and risk effects in open-pit design. Mining Technology: transactions of the
institutions of mining and metallurgy, Philadelphia, USA, v. 111, n. 1, p. 82-88, 2002.

DIMITRAKOPOULOS, R; JEWBALI, A. Joint stochastic optimization of short and long term mine
production planning: method and application in a large operating gold mine. Mining Technology,
Englewood, v. 122, n. 2, p. 110-123, jun. 2013.

DIMITRAKOPOULOS, R; RAMAZAN, S. Uncertainty-based production scheduling in open pit mining.

Transactions of the Society for Mining, Metallurgy, and Exploration, Inc., v. 316, p. 106-112, 2004

FREITAS, M. B. S. Planejamento estocástico de lavra: metodologias de simulação, otimização e gestão de

risco para a mina do futuro. Dissertação (Mestrado) - Escola Politécnica da Universidade de São Paulo –
São Paulo, 2015.

HUSTRULID, W.; KUCHTA, M. Open pit mine planning and design. Rotterdam, Balkemma, 1995.

JOURNEL, A. G. Geostatistics for conditional simulation of ore bodies. Economic Geology, Littleton, v.
69, n. 5, p. 673-687, ago. 1974.

KOPPE, C. V. Metodologia para comparar a eficiência de alternativas para disposição de amostras. Tese
(Doutorado) - (PPGE3M), Universidade Federal do Rio Grande do Sul, Porto Alegre, 2009.

PERONI, R. DE L. Análise da sensibilidade do sequenciamento de lavra em função da incerteza do

modelo geológico. 143 p. Tese (Engenharia) – PPGEM, Universidade Federal do Rio Grande do Sul,
2002.

PILGER, G. G. Critérios Para Locação Amostral Baseados em Simulação Estocástica Dissertação

(Mestrado) - (PPGE3M), Universidade Federal do Rio Grande do Sul, Porto Alegre, 2000.

13