Convolutional neural networks (CNNs) are the gold standard in the machine learning (ML) community. As a result, most of the recent studies have relied on CNNs, which have achieved higher accuracies compared with traditional machine learning approaches. From prior research, we learned that multi-class image classification models can solve leaf disease identification problems, and multi-label image classification models can solve leaf disease quantification problems (severity analysis). Historically, maize leaf disease severity analysis or quantification has always relied on domain knowledge—that is, experts evaluate the images and train the CNN models based on their knowledge. Here, we propose a unique system that achieves the same objective while excluding input from specialists. This avoids bias and does not rely on a “human in the loop model” for disease quantification. The advantages of the proposed system are many. Notably, the conventional system of maize leaf disease quantification is labor intensive, time-consuming and prone to errors since it lacks standardized diagnosis guidelines. In this work, we present an approach to quantify maize leaf disease based on adaptive thresholding. The experimental work of our study is in three parts. First, we train a wide variety of well-known deep learning models for maize leaf disease classification, then we compare the performance of the deep learning models and finally extract the class activation heatmaps from the prediction layers of the CNN models. Second, we develop an adaptive thresholding technique that automatically extracts the regions of interest from the class activation maps without any prior knowledge. Lastly, we use these regions of interest to estimate image leaf disease severity. Experimental results show that transfer learning approaches can classify maize leaf diseases with up to 99% accuracy. With a high quantification accuracy, our proposed adaptive thresholding method for CNN class activation maps can be a valuable contribution to quantifying maize leaf diseases without relying on domain knowledge.

This study investigates crop disease monitoring with real-time information feedback to smallholder farmers. Proper crop disease diagnosis tools and information about agricultural practices are key to growth and development in the agricultural sector. The research was piloted in a rural community of smallholder farmers having 100 farmers participating in a system that performs diagnosis on cassava diseases and provides advisory recommendation services with real-time information. Here, we present a field-based recommendation system that provides real-time feedback on crop disease diagnosis. Our recommender system is based on question–answer pairs, and it is built using machine learning and natural language processing techniques. We study and experiment with various algorithms that are considered state-of-the-art in the field. The best performance is achieved with the sentence BERT model (RetBERT), which obtains a BLEU score of 50.8%, which we think is limited by the limited amount of available data. The application tool integrates both online and offline services since farmers come from remote areas where internet is limited. Success in this study will result in a large trial to validate its applicability for use in alleviating the food security problem in sub-Saharan Africa.

This paper contains a critical review of the norms employed in the design of soil and water conservation structures in the South African sugar industry and highlights research needs in order to update them. Sugarcane in South Africa is grown on wide-ranging soils, sometimes in non-ideal climates and on steep topographies where soils are vulnerable to erosion. A consequence of unsustainable soil loss is reduction in field production capacity. Sugarcane fields are protected against erosion through, inter alia, the use of engineered waterways, contour banks and spill-over roads. The South African Sugarcane Research Institute (SASRI), previously known as the South African Sugar Experiment Station (SASEX), developed a nomograph to easily compute the maximum width of field panels based on soil type, tillage method, replant method, surface structures to control runoff, surface cover and slope. This was followed by guidelines and norms for the design of soil and water conservation structures. However, the nomograph was developed based on an acceptable soil loss of 20 t·ha−1·yr−1, yet soil formation rates in South Africa range between 0.25 and 0.38 t·ha−1·yr−1. Comparisons between design norms in the National Soil Conservation Manual and norms used in the sugar industry clearly show discrepancies that need to be investigated. The design of soil conservation structures includes the design of both contour bank spacing and hydraulic capacity. The sustainable soil loss method is recommended in the design of contour spacing and it determines contour spacing based on evaluation of site-specific sheet and rill erosion potential of the planned contour spacing while the hydraulic design employs Manning’s equation. Considering that increases in both design rainfall and design floods are anticipated in South Africa, it is necessary to incorporate these projections in the design of soil and water conservation structures. Many soil loss models exist, of which empirical models are the most robust and provide stable performances. The majority of empirical models are lumped models which estimate average annual soil loss. The Modified Universal Soil Loss Equation (MUSLE) estimates event-based erosion and, given that the majority of soil erosion occurs during a few extreme events annually, the design norms should be updated using the MUSLE.

Highlights Very few sediment yield events contribute to annual sediment yield. Any rainfall, runoff, and peak discharge event has the potential to generate the most extreme sediment yield event. Twenty year return period recommended for design of conservation structures. Abstract . Design of conservation structures includes both hydrologic and hydraulic designs. Hydrologic design involves estimation of design floods which are required for the sizing of the hydraulic structures. The minimum recommended return period for the design of conservation structures is 10 years but due to the projected levels of risk, and the fact that a few large events are likely to be responsible for the majority of the erosion, the 10-year return period currently recommended may be inadequate. This study investigated system design criteria and the capital cost of varying design return periods for soil and water conservation structures in the sugar industry of South Africa. Observed rainfall data and results of runoff, peak discharge and sediment yield simulated using the Agricultural Catchments Research Unit (ACRU) model were utilized in this study. Relationships between extreme events of sediment yield and the rainfall, runoff and peak discharge events associated with them were analyzed and the capital cost of varying design return periods was also investigated. The results showed that only 0.2% of sediment yield events contributed up to 95% of the annual sediment yield simulated in the sugar production areas in South Africa and that any event of rainfall, runoff and peak discharge had the potential to generate an extreme sediment yield event provided the soil surface was not adequately protected. Based on a sustainable soil loss of 5 t ha-1, the 20-year return period was recommended for the design of soil and water conservation structures. Furthermore, the capital cost implication of varying design return periods from the minimum 10-year return period ranged from an increase of 16% to 35% across the sugar industry. Therefore, given that soil erosion is associated with adverse effects on sustainable crop production and also increases in costs of replanting destroyed crops, the 20-year return period is recommended for the design of soil and water conservation structures in the sugar industry in South Africa. Keywords: Capital cost, Design criteria, Erosion, Return perioW, Risk, Soil and water conservation.

The Agricultural Catchments Research Unit (ACRU) model is a daily time step physical-conceptual agrohydrological model with various applications, design hydrology being one of them. Model verification is a measure of model performance and streamflow, soil water content and sediment yield simulated by the ACRU model have been extensively verified against observed data in southern Africa and internationally. The primary objective of this study was to verify simulated runoff volume, peak discharge and sediment yield against observed data from small catchments, under both bare fallow conditions and sugarcane production, which were located at La Mercy in South Africa. The study area comprised 4 research catchments, 101, 102, 103 and 104, monitored both under bare fallow conditions and sugarcane production, with different management practices per catchment.  Observed data comprised: daily rainfall, maximum and minimum temperature, A-pan evaporation and runoff for the period 1978–1995, and peak discharge and sediment yield for the period 1984–1995. The data were checked for errors and and inconsistent records excluded from analysis. Runoff volume, peak discharge and sediment yield were simulated with the ACRU model and verified against the respective observed data. In general, the correlations between observed and simulated daily runoff volumes and peak discharge were acceptable (i.e. slopes of regression lines close to unity, R2 ≥ 0.6 and the Nash–Sutcliffe coefficient of efficiency close to unity). Similarly, the correlation between observed and simulated sediment yield was also good. From the results obtained, it is concluded that the ACRU model is suitable for the simulation of runoff volume, peak discharge and sediment yield from catchments under both bare fallow and sugarcane land cover in South Africa.