Where will the next oil spill incident in the Niger Delta region of Nigeria occur?

Oil spill incidents are almost a daily occurrence within the Niger Delta region of Nigeria with far-reaching environmental, economic and social consequences. This study aimed at understanding the spatial and temporal context of the problem as a panacea for forecasting likely locations of oil spill incidents within the region. About 76.77% of crude oil spilt in the Niger Delta is lost to the environment with only about 23% of the crude oil recovered from the environment. This represents a very worrying statistic in terms of the known and unknown negative impacts of oil spills. Space Time Pattern Mining (STPM) tools were adapted to explore and interrogate historical spill data. Time series forecasting was then used for forecasting possible locations of future oil spills within the region. Results show that there is a pattern of oil spill occurrences in the Niger Delta with statistically significant hotspots identified in Rivers State, Bayelsa State and Delta State. The forecast root mean square error (RMSE) and forecast validation RMSE are −1.016328 and 1.035992 respectively. This suggests an ability of the model to fairly predict likely locations of future oil spills. This was further verified by counting the number of spills that occur within any area based on the predicted likelihood of spill occurrence. This study has shown that STPM tools can be deployed to understand the occurrence and prediction of oil spill incidents. This will ultimately aid in the deployment of scarce management resources to where they are most needed.


Introduction
The challenges presented by the occurrence of oil spills in the Niger Delta region of Nigeria frequently present a unique set of societal and environmental dilemmas that are dynamic in their manifestations.These challenges are still poorly understood in terms of their occurrence and distribution (Watts andZalik 2020, Wekpe et al 2022).An understanding of the spatial occurrence as well as the temporal distribution of oil spill events in the Niger Delta is the most important variable in identifying the likely and most vulnerable human and environmental receptors in the event of an oil spill (Park et al 2016, Whanda et al 2016).
Oil spills occur almost daily in the Niger Delta, especially in those areas that have one or more types of oil exploration and exploitation infrastructure (Mba et al 2019, Ugwu et al 2021).These spills when they occur are followed by dire consequences for the natural environment and the people who depend on the products of the environment for their survival and maintenance (Mba et al 2019, Akpokodje 2020, Ugwu et al 2021).These spills result from varying sources and activities with frequently reported cases of severe environmental degradation and negative socio-economic consequences for people who depend on the products of the environment for sustenance (Nwilo Badejo 2006, Aroh et al 2010, Ite et al 2013, Ansah et al 2022).Akinwumiju et al (2020), report that between 2006-2019, there was a total of 7,943 oil spills in the region.62% of these spills were attributed to inland corrosion and sabotage of the oil pipelines.Pipelines are very important as they represent critical infrastructure for the effective and efficient running of the oil and gas industry, which are in turn pivotal to continued energy supply and sustained economic development (Osuji et al 2010).Significant negative environmental impacts can emanate when oil pipelines are damaged during operation, such as spills and blowouts which can result in significant long-lasting environmental problems (Walker et al 1995).Equipment failure accounts for a further 7% of oil spills recorded in Nigeria (Akinwumiju et al 2020).An increasingly alarming quantity of spills in the Niger Delta is being attributed to activities of kpofire merchants (artisanal and illegal refineries), with significant losses of crude oil to the environment (Badejo and Nwilo 2004, Naanen and Tolani 2014, Yeeles and Akporiaye 2016, Mba et al 2019, Akinwumiju et al 2020, Ugwu et al 2021).
The spatial and temporal analysis of oil spill events is an important approach to addressing some of the most critical challenges presented by consistent oil spills in the Niger Delta (Whanda et al 2016).Data from various studies indicate that human error accounts for about 80% of the major oil spills in the world (Mehanna et al 2013, Alpers et al 2017, Mba et al 2019).
1.1.Oil Interdiction in the niger delta Badejo and Nwilo (2004) contend that 50% of oil spills emanate from weak pipelines due to corrosion, while 28% is attributed to sabotage.21% of oil spills arise from oil production operations and 1% of oil spills originate from operational failures such as engineering drills, inability to effectively control oil wells, failure of machines, and inadequate care in loading and unloading oil vessels.Authors such as Mba et al (2019) and Ugwu et al (2021), opine that interdiction, which is the willful theft and destruction of oil facilities, is becoming the major factor in oil spill pollution in the Niger Delta.Interdiction is the willful damage to oil pipelines for the purposes of theft, sabotage or bunkering (Anifowose et al 2012, Mba et al 2019, Ugwu et al 2021, Umar et al 2021).
Romson (2022) argues that interdiction is a crime just like human trafficking, drug trade and terrorism.Since interdiction is a crime, it should be treated as such.Crime is a human phenomenon; therefore, its distribution in space is not random.There is a pattern and a systematic application of criminal tendencies in space and time in response to certain confounding opportunities and enabling conditions.Traditionally, crime is a violation of criminal law (Lynch 2020).As such, anyone committing a crime is a criminal.For a crime to occur, several factors must hold true.These factors include; 1. a motivated and capable wrongdoer (the criminal) or tendency to commit a crime 2. a target (helpless and accessible individual or asset) or simply opportunity.
3. absence of capable authority to prevent the actions of the motivated criminal (Olanrewaju and Chimenwo 2019).
All crimes are committed in a given space when all the conditions listed above are met.In the Niger Delta, poverty, hunger and a sense of marginalization have given people the tendency to resort to criminal activities such as oil pipeline interdiction either as a form of economic survival or some form of social protest against perceived political and economic neglect (Oviasuyi and Uwadiae 2010, UNEP 2011, Wekpe et al (2018), Bodo and Gimah 2019).This is further exacerbated by the difficult environment of the region with its network of creeks and swamps through which the oil pipelines meander.These interconnected network of creeks and swamps provide the transport and movement of the siphoned crude oil, from where it is loaded onto barges for either export or to locations for illegal artisanal refining (kpo fire).
Due to the lag-time in the adequate response to crude oil theft crimes in the NDR, often caused by inadequate response personnel and capable authority (police, civil defense, armed forces and customs services), the crime of pipeline interdiction is on the increase in the NDR, resulting in the loss of valuable revenue to the government amounting to hundreds of millions of dollars daily (Romson 2022).The ability of the capable authorities to respond to incidences of oil theft, which usually results in oil spills, can be improved with knowledge of likely locations where the next spills are likely to occur.This is the main question to which this study aims to provide an answer through the adoption of space time pattern mining (STPM) tools and spatial statistics to better understand the spatial and temporal distribution of oil spills within the study area.This is a starting point for forecasting the likely future locations of oil spills in the region due to pipeline interdiction by employing the time series forecasting technique.

Study area
The Niger Delta Region (NDR) defines a region found in the South-South geopolitical zone of Nigeria (figure 1).The region generally encompasses the areas where the river channel of the River Niger attains base level and bifurcates into multiple distributaries, disposing of discharge and sediment load into the Atlantic Ocean (James et al 2007, Abam andFubara 2022).The Niger Delta of Nigeria hosts the oil industry of the country as well as serves as a home to the largest mangrove forest in Africa, the third largest in the world and occupies about 10,000 square kilometers (Nwilo 2013).The NDR covers an area of about 70,000 km 2 (Mmom and Arokoyu 2010, Ocholi 2017, Izah 2018, Anyadiegwu and Uwaezuoke 2015).It is the largest river delta in Africa and the third largest in the world (Wetlands International 2016).The NDR has been described as mostly a flat swampy basin, crisscrossed by an intricate and dense network of rivers, creeks and streams.It is home to diverse species of mangrove forests, rainforests and freshwater swamps (Abam 2001, Abam and Ngah 2016, Olufemi et al 2020).Its topography, geology and soil properties, hydrodynamics and heavy rainfall make the region highly vulnerable to incidences of annual flooding and erosion; throw in a mix of people who are deeply suspicious of outsiders and any attempt to research the region becomes quite challenging (Oviasuyi and Uwadiae 2010, Akpokodje 2020).

Data source and data description
The oil spill data set for this analysis was obtained from the oil spill data sets collected and archived by the National Oil Spill Detection and Response Agency (NOSDRA).NOSDRA is a government agency of the Federal Republic of Nigeria tasked with the responsibility of detecting and swiftly responding to all incidents of oil spills in Nigeria.The data set contains records of all spills in the country as reported by the oil-producing companies and communities and duly validated by a joint investigative visit (JIV).The data set is typically in a spreadsheet format composed of geographically referenced spill sites, dates and times of the events, estimated quantity of the spill and a general description of the spill area.Other information captured in the spill data sets includes facility owner, type of spill and likely cause of the spill, clean-up status, local government area where the spill occurred, images of the spill site and oil recovery status.The key components of the dataset needed for this research are the spill locations (latitude and longitude), the estimated volume of oil spilled and the oil spill occurrence dates.4,810 oil spill incidents reported between March 2012 to March 2022 were used in the analysis.The data set is publicly available on the Nigeria oil spill monitor maintained by NOSDRA on https://oilspillmonitor.ng/.

Space time pattern mining (STPM)
Space Time Pattern Mining (STPM), aids in the analysis of the distribution of geographically referenced data in space and time ESRI (2020).It uses a series of sophisticated statistical calculations to identify patterns in the context of events across space and time.The STPM tools operate on the premise that every event occurs somewhere and at a particular time.STPM analysis tools help in understanding where, when and at times why events occur.STPM is an emergent, but very important geospatial tool that enables the identification of statistically significant hot and cold spots within a study region of interest.The technique has been applied to studies in criminology for the identification of crime patterns, disease outbreaks and other emerging environmental or ecological issues (Harris et al 2017, Morckel and Durst 2021, Purwanto et al 2021).

Emerging hotspot analysis
Spatial data mining techniques such as hotspot analysis help to identify unknown and unclear patterns in a spatial and temporal data set.Hotspot analysis aids in the clustering of variables in order to better visualize changes in a given factor (Nallan et al 2015, Obida et al 2017).The basic goal of any hotspot analysis is to identify groupings within an area.These groupings can either represent high or low values of a given variable, which correspond to hot and cold spots, respectively (Sánchez-Martín et al 2019, Umar et al 2021).This research adopted the use of the optimized hotspot analysis tool in ArcGIS Pro.The optimized hotspot analysis tool in ArcMap was deployed to identify statistically significant clusters within the oil spill data and the locations of the reported and recorded incidents (figure 2).The optimized hotspot analysis is an upgrade on the Getis-OrdGi * hotspot analysis tool (Ghodousi et al 2020).
The optimized hotspot analysis tool generates its results by using parameters derived from the characteristics of the input data (figure 2).The Optimized Hot Spot Analysis tool interrogates the input data in order to obtain the settings that will yield optimal hot spot results (ArcGIS Pro help 2021).The tool works by identifying statistically significant spatial clusters of high values (hot spots) and corresponding low values (cold spots).Thereafter, it automatically aggregates incident data and identifies an appropriate scale of analysis, and corrects for both multiple testing and spatial dependence.This tool interrogates the input data in order to determine settings that will produce optimal results (ArcGIS Pro help 2021).The emerging hot spot analysis begins with a conceptualization of the spatial and temporal relationships with a defined geographic dataset, specified by the user, to calculate a Getis-Ord Gi * statistic for each bin in the space-time cube.The emergent result is defined by a z-score, p-value, and hot spot classification per bin ESRI (2020).
Hotspots define areas (locations) shown on a map where events are occurring at a rather higher normalized magnitude or frequency when compared to the entire area of interest or study (Maciejewski et al 2010).Hotspot analysis is a very important tool that helps to identify and pinpoint the location of clustering and dispersion in a spatial data set.This becomes very helpful in the case of dealing with a large dataset of incidents, where many incidents overlap one another, thereby making it difficult to visually determine exactly where the 'hot' and 'cold' spots are in the data.Hotspot analysis can also be deployed in temporal analysis, in order to determine seasonal locational shifts in the data being examined.The formula for hotspot analysis is given by the equation below.
Where; Gi * = Getis-Ord Gi * x j = attribute value for feature j w i , j = the spatial weight between feature I and j. n = total number of features and: The purpose of conducting an emerging hot spot analysis is to investigate a phenomenon of interest and identify patterns or trends over space in conjunction with identifiable patterns over time.The summary fields across the study area are then explored to identify clustering patterns of point densities (Maciejewski et al 2010).
The Emerging Hot Spot Analysis tool begins by using the results of a successfully executed space time cube (STC) as an input to conduct a hot spot analysis using the Getis-Ord Gi * statistic for each bin already created in the space time cube (Khan et al 2023).The Neighborhood Distance and Neighborhood Time Step parameters define how many surrounding bins, in both space and time, will be considered when calculating the statistics for a specific bin.Then, the hot and cold spot trends detected by the Getis-Ord Gi * hot spot analysis are evaluated with the Mann-Kendall test to determine whether trends are persistent, increasing, or decreasing over time (Khan et al 2023).The results are symbolized by seven different categories describing the statistical significance of hot or cold spots and the location's trend over time.

Time series forecasting (TSF)
Time series forecasting defines methods or tools used to predict or forecast the future outcome, value or an event over a period of time.It involves the development of models based on historical data and calibrating them to make useful observations and therefore inform, guide and direct strategic future decisions of managers or decision-makers.Time series forecasting holds under the assumption that the future conditions will hold true to historical trends or patterns (Hyndman and Athanasopoulos 2018).
In this study, we see oil theft as a crime and therefore treat it as such.For a crime to occur, there are several factors that must hold.These factors include; a motivated and capable wrongdoer (the criminal); a target (helpless and accessible individual or asset); and absence of a capable authority to prevent the actions of the motivated criminal (Olanrewaju and Chimenwo 2019).Forecasting problems that involve a time component requires time series forecasting as an approach to provide viable solutions to the problem by applying a datadriven approach for effective and efficient planning purposes (Hyndman and Athanasopoulos 2018).
Given a spatiotemporal dataset given as O i (oil spill data) of historic oil spill events (data provided by NOSDRA), for the Niger Delta area (Bayelsa, Delta and Rivers State) with feature set L i (location Where T (time) is the independent variable or predictor and L (location) is the dependent variable.The aim here is to predict to an appreciable extent and accuracy, the location of the next oil spill.This is based on earlier findings from the hotspot analysis and supporting literature that most modern spills in the Niger Delta are a result of interdiction.The objective is to achieve the desired results with minimal classification errors along with a confidence value indicating the confidence of the prediction.The proposed crime prediction model was conducted on three levels; oil spill history dataset preparation, oil spill hotspot identification and spill prediction approach using time series forecasting tools.
This study employed the use of the time series forecasting tool available in ArcGIS Pro to build a model for predicting the likely locations for future occurrences of oil spill cases based on available historical data.The time series forecasting tools (TSFT) enable forecasting and estimation of future values of the probability of events occurring within locations of a STC.Further evaluation of the results and comparison of the output can also be performed for each model for each location in order to determine the best-performing model.There are three distinct TSFC available in the ArcGIS Pro toolbox.They include simple curve fitting, forest-based forecasting and exponential smoothing forecasting methods (ArcGIS Pro Help 2020).
The three different time series forecasting methods used for this study were then subjected to validation to determine the best performing model by applying the evaluate forecasts by location tool available in ArcGIS Pro.The evaluation tool was used to create multiple forecast cubes using the three different forecasting tools and parameters (simple curve fitting, forest-based forecasting and exponential smoothing).The tool identified the best performing forecast for each location using either forecast root mean square error (RMSE) or validation RMSE.

Presentation of results
Figure 3 presents the temporal distribution of oil spills in the Niger Delta between March 2012 and March 2022.Within the period under consideration, a total of 4,810 oil spill events were recorded.These spills accounted for the introduction of an estimated 363,736bbl of crude oil into the natural environment, comprising the land, vegetation, waterbodies and air.The year 2014 accounted for the highest number of spills with a contribution of 847 spill incidents with an associated 75,954.9bbl of crude oil into the environment.An estimated 22,177.4bbl was recovered, leaving about 53,777.5bbl of crude oil trapped in the environment.Of the estimated 363,736bbl of oil spilt into the environment within the period under consideration, only 109,899bbl of the spilt crude oil was recovered.This means that an estimated 253,837bbl of crude oil accounting for 76.77% of the spilt amount was lost to the environment.
Figure 4 expresses the frequency magnitude relationship that exists between oil spill counts (occurrence and frequency) and the density (volume or quantity) of oil spill incidents in the Niger Delta measured in barrels (bbl).The observed relationship is an inverse relationship in that as the volume or density of spills increases, the frequency or counts reduces.Further analysis of the data shows that the majority of oil spills within the study area fall below 758bbl per incident, accounting for just over 85% of oil spill incidents within the Niger Delta region.The implication of this is that spills which are categorized as large spills or tier 2 spills (>50bbl but <5000bbl) dominate as the main oil spill Figures 5(A) and (B) shows the results of the hotspot analysis performed to identify oil spill hotspot zones based on the frequency and cluster of oil spill events.Autocorrelation involving the aggregation of frequently occurring spill incidents and their location in space aided in clustering frequently occurring spill locations.The results show that most of the spills with statistically spatial significance (hotspots) occur along major pipelines within the region (figure 1).The hotspot's significance is interpreted as a function of the density of occurrence of oil spill events.From the results, the spill density follows a linear pattern and then radiates outwards from a central core with the hotspot significance progressively decreasing from the central core.The trans-Niger pipeline and the Ebocha/Emouha/Rumuekpe/Nkpoku pipeline are the pipelines with the most significant spill clusters.
Figure 5 (A) expresses the intensity of oil spill occurrence in the Niger Delta proxied by the number of oil spill events reported and recorded by NOSDRA.Figure 5 (B), on the other hand, denotes the proportion or quantity of the spill event expressed in barrels (a barrel equals 163.7 litres).The distribution of oil spills in the NDR is expressed both as a function of the frequency of the occurrence of events as well as the quantity of oil spilt.The results further show that the occurrence of oil spills in the NDR is not random as illustrated in figures 5(A) and (B), but rather occurs as a clustering of statistically significant events in space and time.The results represent a fourth nightly prediction for locations most likely to experience an oil spill event.Of the three States that make up the study area, Bayelsa and Rivers States have a similar outlook for possible incidents of oil spill from the last date of the input historical data.The North East region of Bayelsa State covers communities such as Mbima, Yenegwe, Biogbolo and Okarki.The North West of Rivers State covering the communities of Akinima, Obite Akabuka and Ahoada represents the locations with the greatest risks.The probability of oil spills due to interdiction is very high in Bayelsa State and Rivers State.Delta State had very lowrisk scores across all locations.
Results from the predictive model for a fourth nightly time step show that Rivers State and Bayelsa possess the locations with the highest likelihood of oil spills occurring, with significant possibilities of spills also occurring across different locations of the three states of interest.The North Western sections of Rivers State and the North Eastern sections of Bayelsa State have very high-risk chances of oil spills occurring, while there is a significant risk of oil spill events occurring across the three states of interest.

Model evaluation and validation
In order to determine the best performing model in forecasting future oil spill locations, the three different forecast models (curve fit forecast, forest based forecast and exponential forecast) were compared and merged using the evaluate forecasts by location tool available in ArcGIS Pro.The evaluate forecast tool enables for the creation of multiple forecast cubes using different forecasting tools.The tool eventually identifies the bestperforming model for each location space time cube using either the Root Mean Square Error (RMSE) or validation RMSE.The relatively small validation RMSE of 1.035992 for the exponential time series forecast model shows it is the best-performing model for prediction (figure 7, table 1).
Figure 7 shows the model prediction trend from the original value (oil spill data), the fitted values, the confidence interval and the forecasted values, which are represented in space as specific hexagonal space time cubes (figures 6(A), (B) and (C).Other information shown includes outliers above and below the fitted value.It is important to note that with increasing time steps, the accuracy of the model significantly decreases with relatively high RMSE.Hence, the use of a fourth nightly time step as this period produced the best RMSE values.
Figure 8 shows the exponential smoothing time series forecasting technique, which has been identified as the best-performing model after performing a model evaluation, using the evaluate forecasts by location tool available in the ArcGIS Pro STPM toolbox (table 1).

Spatio-temporal dimension of oil spills in the niger delta
Various reasons have been advanced to account for the continued oil spill problem experienced in the Niger Delta.These reasons range from operational failure, pipeline rupture, human errors and oil theft or interdiction (Mbah et al 2019, Ugwu et al 2021).Results from the analysis show that hotspots and clusters of significant oil spill incidents in terms of the frequency of spill occurrences and the volume of oil spilt occur along pipelines (figures 1 and 5).This agrees with the work of Mbah et al (2019) and Ugwu et al (2021), who found that most oil spill interdiction occurs at pipeline locations through which both crude oil and refined products are transported either from oil fields or to export terminals.
Of the numerous pipelines that crisscross the region, three of these pipelines recorded statistically significant spills (hotspots).These pipelines include the Trans Niger pipeline (TNP) owned and operated by SHELL.The pipeline is a major pipeline within the region which carries between 180,000 to 200,000 barrels of crude oil (Bonny light) per day through Port Harcourt, Eleme and Ogoni to the export terminal at Bonny South of Port Harcourt.The other pipeline is the Nembe Creek Trunk-line which is operated and managed by AITEO and operates mostly in Bayelsa state with termination and loading bays at Brass.The third pipeline of major interest is the Ebocha-Emouha-Rumuekpe Trunk-line which feeds into the Trans Niger pipeline at Rumuekpe, all in Rivers State.The third pipeline of interest is the Trans-Forcadoes pipeline.The spatial concentration of hotspots of oil spills is not random in their manifestation and must therefore be treated as a prelude to future spills, as those sections provide an enabling environment for crude oil theft to occur (Ngada and Bowers 2018).
The volumetric distribution of oil spills in the Niger Delta also follows a similar pattern to the spatial clustering shown for the intensity of spill events (figure 5(A)).Figure 5(B) clearly shows an agglomeration of spill incidents in the study area as a function of the quantity of crude oil spilled within that particular region in space and time.It also shows an agglomeration of spills of less than 5000bbl as the most dominant type of spill in terms of the volume spilled (figure 4).The implication of this clustering is that there is a pattern to the volume of oil lost to the environment due to oil spill counts.This assertion agrees with the findings of Whanda et al (2016), who found there exist clusters of oil spill events in respect of both the frequency and the intensity by volume.
Further evaluation of the results presents an interesting set of relationship and pattern emergence in the distribution and clustering of oil spill counts on one hand and the volumetric (magnitude or quantity) distribution of oil spills in space on the other hand.Figures 5(A) and (B) shows that areas in the Niger Delta associated with high-frequency counts (hotspot of events) are also the same areas with statistically significant quantities of oil spills within the study area.This finding is given further credence by the work of Whanda et al (2016), who found significant cases of oil spills to be clustered.
Several factors are integral to these clustering patterns observed in the distribution of oil spill cases in the Niger Delta (in terms of the frequency or counts and its associated quantity or magnitude).It has earlier been established that the majority of oil spill cases in the study area emanate from cases of interdiction for the purposes of crude oil theft (Mba et al 2019, Ugwu et al 2021).These interdiction activities rely on several favorable conditions needed for the relative ease of sabotaging oil facilities, removing and transporting the stolen  crude oil.The most significant of these conditions is the availability of a transport route for the easy movement of the siphoned crude oil, which is also followed by the relative obscurity of the location for the interdiction activities.

Predictive modelling of crude oil spill hotspots in the niger delta
This study treated the problem of oil interdiction in the Niger Delta as a crime.And like most reoccurring crimes, a pattern emerges over time (Marzan et al 2017).It is this emerging pattern that was explored using hotspot analysis, which led to the attempt to model future occurrences of oil interdiction that could lead to oil spills.As such, a time series forecasting analysis approach was adopted to achieve the goal of predicting the likely location or locations of the next spill within the Niger Delta region.A combination of the results obtained from the analysis of retrospective oil spill data (oil spill location and time) led to the development of three distinct predictive models for the identification of future oil spills in the Niger Delta by adapting the predictive capacity of time series forecasting.The space time analysis of oil spills in the study area has already revealed that there are clusters (hotspots) of oil spill events within the study area.This provides a platform for supporting the identification of high-risk areas and the formulation of prevention and suppression efforts.Figures 6 (A), (B) and (C) shows the results of the fourth nightly time series forecasts for different locations within the study area proxied by the 372 STC created by aggregating the locations of individual oil spill locations into hexagonal grids.The results further show that the probability of the occurrence of oil spills due to interdiction is very high in Bayelsa State.Of the three states that make up the study area, Bayelsa State has the greatest likelihood of experiencing an oil spill incident.Within Bayelsa State, the propensity for an oil spill incident to occur is greatest around Yenogoa LGA and the towns of Epie and Kolo Creek.This is followed by reduced predicted risk values for Rivers State and Delta State with the least likely occurrences of oil events resulting from pipeline interdiction.
Results from the predictive models show that Rivers State and Bayelsa State are most at risk for possible cases of oil spills.The North Eastern region of Rivers State and the North Western region of Bayelsa State are shown to be the most probable locations for oil spills to occur.The reason for this may be due to the greater availability of oil infrastructure within these states and the increasing ease of moving stolen crude through the interconnected waterways and forest tap lines that crisscross these states.This agrees with the findings of Mba et al (2019), and Ugwu et al (2021), who noted that theft accounts for the majority of oil spill events in the Niger Delta.

Conclusion and recommendations
This study investigated the use of space time pattern mining techniques in identifying hotspot zones of oil spill incidents in the Niger Delta region of Nigeria with a view to forecasting the locations of future occurrences using historical oil spill data reported for the region.This study has been able to demonstrate that the STPM techniques can be a useful tool in proactively identifying hotspot zones and predicting likely locations of oil spill incidents that arise due to interdiction.Findings show that oil spills that emanate due to interdiction can arguably be predicted to a relatively accurate extent.
This research has amply demonstrated that time series forecasting can be applied to identifying potential locations of future oil spills.However, this knowledge can only be useful in managing oil spill problems when combined with effective policing and deployment of responsive units to target regions.Arising from this, it is recommended that NOSDRA and NSCDC deploy the use of monitoring drones to keep a close eye on areas and pipeline sections that have been identified as hotspots.

Figure 1 .
Figure 1.The Niger Delta Region showing oil spill locations and pipeline network.

Figure 2 .
Figure 2. Space time cube (STC) for detecting patterns in space and time.Data processing and workflow for performing emerging hotspot analysis (A) data sourcing, conversation, aggregation and storage (B) data pre-processing and processing (C) points are aggregated into space time bins and individual point counts are tallied (D) data extracted for area of interest (E) data transformed into CDF data cube format (F) for each bin in the cube, the Getis-Ord Gi statistic is estimated and the Mann-Kendall trend test is generated for each bin's time series to produce an emerging hot spot map.(Source: Harris et al 2017).Reproduced from Harris et al. (2017) © IOP Publishing Ltd.CC BY 3.0.

Figure 4 .
Figure 4. Frequency magnitude curve of oil spill incidents in the Niger Delta (data is log at base 5).

Figure 5 .
Figure 5. Oil spill hotspot zones in the Niger Delta by count (A) and by volume (B).

Figure 6 .
Figure 6.Time series forecasting of likely oil spill locations across the Niger Delta using different forecasting methods: (A) Is a curve fit forecast method for predicting oil spill incidents; (B) is a forest-based forecasting method for predicting likely locations of oil spills; (C) Is an exponential time series forecast method for identifying possible oil spill occurrence locations in the Niger Delta.

Figure 7 .
Figure 7. Model prediction of possible oil spill location for specific STC in the study area.

Table 1 .
Summary table for model evaluation and validation.