Droughts are slow-moving natural hazards that gradually spread over large areas and capable of extending to continental scales, leading to severe socio-economic damage. A key challenge is developing accurate drought forecast model and understanding a models' capability to examine different drought characteristics. Traditionally, forecasting techniques have used various time-series approaches and machine learning models. However, the use of deep learning methods have not been tested extensively despite its potential to improve our understanding of drought characteristics. The present study uses a deep learning approach, specifically the Long Short-Term Memory (LSTM) to predict a commonly used drought measure, the Standard Precipitation Evaporation Index (SPEI) at two different time scales (SPEI 1, SPEI 3). The model was compared with other common machine learning method, Random Forests, Artificial Neural Networks and applied over the New South Wales (NSW) region of Australia, using hydro-meteorological variables as predictors. The drought index and predictor data were collected from the Climatic Research Unit (CRU) dataset spanning from 1901 to 2018. We analysed the LSTM forecasted results in terms of several drought characteristics (drought intensity, drought category, or spatial variation) to better understand how drought forecasting was improved. Evaluation of the drought intensity forecasting capabilities of the model were based on three different statistical metrics, Coefficient of Determination (R2), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). The model achieved R2 value of more than 0.99 for both SPEI 1 and SPEI 3 cases. The variation in drought category forecasted results were studied using a multi-class Receiver Operating Characteristic based Area under Curves (ROC-AUC) approach. The analysis revealed an AUC value of 0.83 and 0.82 for SPEI 1 and SPEI 3 respectively. The spatial variation between observed and forecasted values were analysed for the summer months of 2016-2018. The findings from the study show an improvement relative to machine learning models for a lead time of 1 month in terms of different drought characteristics. The results from this work can be used for drought mitigation purposes and different models need to be tested to further enhance our capabilities.
Healthy farming systems play a vital role in improving agricultural productivity and sustainable food production. The present study aimed to propose an efficient framework to evaluate ecologically viable and economically sound farming systems using a matrix-based analytic hierarchy process (AHP) and weighted linear combination method with geo-informatics tools. The proposed framework has been developed and tested in the Central Highlands of Sri Lanka. Results reveal that more than 50% of farming systems demonstrated moderate status in terms of ecological and economic aspects. However, two vulnerable farming systems on the western slopes of the Central Highlands, named WL1a and WM1a, were identified as very poor status. These farming systems should be a top priority for restoration planning and soil conservation to prevent further deterioration. Findings indicate that a combination of ecologically viable (nine indicators) and economical sound (four indicators) criteria are a practical method to scrutinize farming systems and decision making on soil conservation and sustainable land management. In addition, this research introduces a novel approach to delineate the farming systems based on agro-ecological regions and cropping areas using geo-informatics technology. This framework and methodology can be employed to evaluate the farming systems of other parts of the country and elsewhere to identify ecologically viable and economically sound farming systems concerning soil erosion hazards. The proposed approach addresses a new dimension of the decision-making process by evaluating the farming systems relating to soil erosion hazards and suggests introducing policies on priority-based planning for conservation with low-cost strategies for sustainable land management.
The spatial variation of soil erosion is essential for farming system management and resilience development, specifically in the high climate hazard vulnerable tropical countries like Sri Lanka. This study aimed to investigate climate and human-induced soil erosion through spatial modeling. Remote sensing was used for spatial modeling to detect soil erosion, crop diversity, and rainfall variation. The study employed a time-series analysis of several variables such as rainfall, land-use land-cover (LULC) and crop diversity to detect the spatial variability of soil erosion in farming systems. Rain-use efficiency (RUE) and residual trend analysis (RESTREND) combined with a regression approach were applied to partition the soil erosion due to human and climate-induced land degradation. Results showed that soil erosion has increased from 9.08 Mg/ha/yr to 11.08 Mg/ha/yr from 2000 to 2019 in the Central Highlands of Sri Lanka. The average annual rainfall has increased in the western part of the Central Highlands, and soil erosion hazards such as landslides incidence also increased during this period. However, crop diversity has been decreasing in farming systems, namely wet zone low country (WL1a) and wet zone mid-country (WM1a), in the western part of the Central Highlands. The RUE and RESTREND analyses reveal climate-induced soil erosion is responsible for land degradation in these farming systems and is a threat to sustainable food production in the farming systems of the Central Highlands.