ETAPOD: A forecast model for prediction of black pod disease outbreak in Nigeria

Food poisoning and environmental pollution are products of excessive chemical usage in Agriculture. In Nigeria, cocoa farmers apply fungicides frequently to control black pod disease (BPD), this practice is life threatening and lethal to the environment. The development of a warning system to detect BPD outbreak can help minimize excessive usage of fungicide by farmers. 8 models (MRM1-MRM8) were developed and 5 (MRM1-MRM5) selected for optimization and performance check. MRM5 (ETAPOD) performed better than the other forecast models. ETAPOD had 100% performance rating for BPD prediction in Ekiti (2009, 2010, 2011 and 2015) with model efficiency of 95–100%. The performance of the model was rated 80% in 2010 and 2015 (Ondo) with model efficiency of 85–90%, 70% in 2011 (Osun) with model efficiency of 81–84%, 60% in 2010 (Ondo and Osun) and 2015 (Osun) with model efficiency of 75–80%, 40% in 2009 (Osun) with model efficiency of 65–69% and 0% 1n 2011 (Ondo) with model efficiency between 0 and 49%. ETAPOD is a simplified BPD detection device for the past, present and future.


Introduction
Global warming, food poisoning and environmental pollution are current problems emanating from excessive exposure to and combustion of chemical substances. The management of BPD is a major challenge to cocoa farmers in Nigeria as they frequently apply fungicide to safeguard their crops without consideration for the safety of life and the environment [1]. BPD is more established in West Africa than in any other parts of the world [2]. Adegbola [2] in his review of Africa estimated the average occurrence of the disease as 40% in several parts of West Africa and up to 90% in certain places in Nigeria [3]. In Nigeria, cocoa export made over 80% GNI before the 1960s [3], it was reduced to 37.9% in 1997 [4] due to BPD infection and other factors, yet cocoa export remained more profitable than Rubber, Palm fruits, Groundnut, Yam, Cassava, Maize, Millet, and Sorghum [3]. Cocoa yield decline started in [1971][1972] (255,000 to 241,000Mt), through 1978-1979 to 1986 (58,700Mt), with an increase from 165,000-180,000Mt between 2000 and 2003 [5]; [6]. The increase in cocoa production was entirely due to the expansion in production area rather than increases in cocoa yield [7]. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Global climate change is one of the major factors responsible for the inconsistent fluctuations in BPD outbreak experienced annually worldwide, due to its influence on the physiology of the pathogen(s), the suitability of the environment for microbial activities and the susceptibility of the host plant(s) to microbial attack [8]. The irregular rainfall pattern and inconsistent mode of BPD occurrence in Nigeria makes it nearly impossible to control it effectively. The efficacy of the existing management strategies (cultural, physical, biological and chemical control measures) are fast declining, as such increased fungicide dosages and frequent applications are methods devised by indigenous cocoa farmers to protect their crops from the disease. Hence, an urgent need for modern approach in the control of BPD in West Africa is imminent so as to reduce the level of exposure of cocoa pods to chemicals and also decrease the amount of chemical residues in the environment [9].
Plant disease forecasting (advance disease management strategy) advocates the use of plethora management techniques directed by a rational decision making system, such that indigenous cocoa farmers worldwide will be duly informed whenever BPD outbreak is suspected and the intensity of the outbreak quantified to avoid excessive use of preventive chemicals. This research seeks to develop a forecast model for BPD prediction so as to provide information on its outbreak and the areas suspected to be under severe invasion. In the nearest future, the quantity of preventive control measures required to combat the disease will be determined using simple computer algorithms in order to minimize fungicide misuse, reduce the level of chemical pollutants in the environment, increase cocoa productivity, and reduce the risk of chemical poisoning or deaths associated with consumption of toxic chemicals substances.

The area of focus
The area of research focus was the Western part of Africa with specific preference to Nigeria (Fig 1). The Southwestern region of Nigeria was used as a case study for result validation; this region was clearly described in Fig 2. The co-ordinates of the area of focus were determined using the blackberry mobile Global Positioning System (GPS) device (version 6.0) and a mobile satellite GPS receiver model GARMIN Etrex 10 obtained from the Department of Botany, Faculty of Science, University of Ibadan, Ibadan, Nigeria. Cocoa producing States in Nigeria were shown in Table 1

Black pod disease (BPD) data
Documented reports of BPD outbreaks within Southwestern Nigeria was obtained from Cocoa Research Institute of Nigeria (CRIN), Ìdí-Ayunrė, Ibadan, Oyo State, Nigeria and the report of Lawal and Emaku [10]. The total data collected spanned from 1985 to 2014. These served as secondary data.

Meteorological data
Weather data from 1985 to 2016 within Southwest, Nigeria were also collected from the report of Lawal and Emaku [10], National Bureau of Statistics (NBS) Ibadan, Oyo State, the Meteorological Station of Cocoa Research Institute of Nigeria (CRIN), Ìdí-Ayunrė, Ibadan, Oyo State, Nigerian Meteorological Station (Nimet), and the Department of Geography, University of Ibadan, Ibadan, Oyo State, Nigeria. These were also classified as secondary data.

Data analysis
Qualitative data were represented as charts and graphs plotted using SPSS, version 20.0 for 32 bits resolution, while the analysis of variance was carried out using COSTAT 6.451 software. The homogeneity of means was determined using Duncan Multiple Range Test (DMRT). The proposed forecast model(s) were templates of multiple regression equation(s) developed from the meteorological and BPD data (Secondary data). The models were designed using Minitab 16.0 software and programmed on Microsoft Excel Worksheet 2007 service pack for easy access. Model selection was aided by Pearson's Product Moment of Correlation (PPMC) "R-Sq", the Adjusted Coefficient of Correlation of the developed models (R-Sq Adj. ), the Standard Error of Regression (SER) and Root Mean Square Error of Prediction (RMSE pred. ). The Error of BPD prediction was determined using E = (Y-Ŷ) 2 .

Model validation
The data used for confirmation of the predicted BPD outbreak by the developed forecast model was obtained from the research work of Oyekale [11], [12]. The template for validation (accuracy check) was stated in Table 2.

Results
The BPD function was developed using simple mathematical rule Then; D ¼ FðH; P; EÞ ¼ aðH; P; EÞ þ b ð1Þ In any case the influence of man and vectors (Ants, Termites, and Rodents etc.) serve as constants in the equation because they influence the spread of BPD in the field, coupled with the timely combination of the key factors responsible for BPD development.
Therefore, the development of BPD forecast system for cocoa required an equation encompassing all the predictors necessary for the disease development. An example of such model was given thus: In any case the individual predictors were tested against the response variable to ascertain their role(s) in black pod disease outbreak.
Rainfall and average relative humidity had a positive correlation with BPD outbreak i.e. r = 0.445 and 0.477 (Fig 4), and r 2 = 0.105 ( Fig 5) and 0.295 (Fig 6), respectively. The average temperature, sunshine duration and the year of observation had negative association with BPD outbreak in Southwest, Nigeria (r = -0.420, -0.364 and -0.018 (Fig 4), and r 2 = 0.265 (Fig  7), 0.360 (Fig 8) and 0.035 (Fig 9), respectively). It was however observed that there was no relationship between the locations of cocoa farms (Fig 10), the specific period (month) when the disease was observed (Fig 11) and BPD outbreak in Nigeria.

The weather pattern of Southwest, Nigeria and how it affects BPD development
The weather pattern for Southwest, Nigeria in the late 1900s (20 th Century) showed that the height of rainfall across the four (4) States investigated was between the months of March and October from 1991 to 1995, suggesting the possibility of infection within these periods (Table 3). Phytophthora megakarya thrives better between 20˚C and 30˚C, therefore the specific periods of the year that favoured such temperature values in Ogun, Ondo, Osun and Oyo States were June, July, August, and September in 1991 to 1995 (Table 4). On the Contrary, the  (Table 5). A relative humidity value of 75% and above favoured the establishment of BPD, therefore, periods of the year that had high relative humidity were March through October from the early morning readings taken 1991 to 1995, suggesting the possibility of infection within these periods also (Table 6). Judging by the trend of afternoon readings the periods of June through September across all the years favoured BPD proliferation (Table 7). These periods possibly served as an interlude for proliferation and spread of the pathogen leading to possible infection of predisposed cocoa plants judging from the BPD occurrence report given by the Cocoa

Development of prediction models for black pod disease in Nigeria
Several models were developed to predict BPD outbreak in Southwest, Nigeria. ETAPOD: A forecast model for prediction of black pod disease outbreak

Model selection
The posthoc analysis conducted showed that MRM 5 was the preferred model for BPD prediction followed by MRM 4

Prediction of annual BPD outbreak by ETAPOD and confirmation of forecast results
The annual BPD outbreak for Ekiti, Ondo and Osun States (Southwest, Nigeria) were used to test the developed BPD forecast model (ETAPOD). In 2009, Ekiti, Ondo and Osun States had    Table 9).

The overall assessment of the output quality of MRM 5 forecast model (ETAPOD)
The  Table 10.

Accuracy of ETAPOD in BPD prediction
The  (Table 12).

The probability of obtaining accurate BPD predictions
The probability of obtaining accurate results for BPD prediction was very high in Ekiti and Ondo States, but it was not consistent in Osun State (Table 13). The probability range for obtaining good results in Ekiti was 0.93�P�0.95, whereas, it was 0.50�P�0.83 in Ondo State. In Osun State, the value was a disappointing 0.00�P�0.71 range (Table 13).

Weather survey in line with BPD outbreak in Southwestern Nigeria
The weather report in the early 1900s for Southwestern Nigeria showed that there was recurrent rainfall within the months of March through October from 1991 to 1995. Also, ambient temperature was low during the day and at night, and there was much saturated water vapour in the air across the four (4) States investigated within the same period. March to October happen to be the most productive periods for Cocoa production in Southwest, Nigeria; Therefore, the observations noted gives an indication of the possibility of infection within these periods. This favourable weather pattern for black pod disease infection was earlier reported by Akrofi [14].

BPD prediction and data validation
The MRM 5 BPD forecast model (ETAPOD) selected as the best fitted model for prediction of BPD outbreak in Nigeria gave results of annual BPD outbreak that accurately quantified annual BPD outbreak in Ekiti and Ondo States but inaccurately described the situation in Osun State. This is as a result of the credibility of the data fed into the model. The observation made was in accordance with the findings of Luo [15] who also designed a forecast model for the prediction of foliar diseases of winter wheat caused by Septoria tritici across England and Wales and his predictions for the disease was seemingly not 100% accurate. The error of BPD prediction was very low in Ekiti, whereas, it was on the high side in Osun. The disparity in the credibility of the predicted outcome is solely due to the quality of the data fed into the system. This lapses was indeed identified by Luo [15] who gave a few recommendations on how a forecast system can be improved in order to obtain quality forecast results. The level of prediction accuracy was defined thus as 0.0%�Accuracy Level<100%. This was also identified by Luo [15] as he recognized the fact that no forecast system can be 100% accurate at all times and in all instances.

Recommendation
ETAPOD harnesses several potentials and possibilities that can be improved on to obtain excellent results. The accuracy of the warning system developed for the prediction of black pod disease (ETAPOD) can be perfected if: 1. Weather parameters are obtained from meteorological stations situated in the farm 2. Consistency of cocoa production within that locality is constant 3. The type of cropping system employed could be determined 4. cocoa is the major crop cultivated on the piece of land 5. Advanced digital image analysis could be used to improve measurement precision of disease prevalence and severity.  "+" Fairly Good Prediction, "++" Good Prediction, "+++" Extremely Good Prediction, and "-"Poor Prediction.
https://doi.org/10.1371/journal.pone.0209306.t010 Conclusion ETAPOD harnesses the potentials to improve the functionality of other existing management strategies for the control of BPD in Nigeria by providing timely information on its outbreak, detect areas under severe attack (AUSA), thereby discouraging fungicide misuse among local cocoa farmers. ETAPOD is unique in the sense that its primary function is not geographically restricted. Also, ETAPOD can be manipulated to provide optimum results anywhere needed in Nigeria, Africa and all around the world. Its ability to provide qualitative and quantitative description of BPD pressure makes it superior to other forms of BPD control strategies in use. Therefore, ETAPOD is a pertinent tool that can effectively minimize the prevalence of BPD in Nigeria with minimal chemical application, decreasing the risk of chemical poisoning and increasing the production of healthy cocoa products nationwide. This is the surest and fastest way to ensure sustainability of cocoa production in Nigeria and the world at large.