Forecasting Vegetation Condition with a Bayesian Auto-regressive Distributed Lags (BARDL) Model

Abstract. Droughts form a large part of climate/weather-related disasters reported globally. In Africa, pastoralists living in the Arid and Semi-Arid Lands (ASALs) are the worse affected. Prolonged dry spells that cause vegetation stress in these regions have resulted in the loss of income and livelihoods. To curb this, global initiatives like the Paris Agreement and the United Nations recognised the need to establish Early Warning Systems (EWS) to save lives and livelihoods. Existing EWS use a combination of Satellite Earth Observation (EO) based biophysical indicators like the Vegetation Condition Index (VCI) and socio-economic factors to measure and monitor droughts. Most of these EWS rely on expert knowledge in estimating upcoming drought conditions without using forecast models. Recent research has shown that the use of robust algorithms like Auto-Regression, Gaussian Processes and Artificial Neural Networks can provide very skilled models for forecasting vegetation condition at short to medium range lead times. However, to enable preparedness for early action, forecasts with a longer lead time are needed. The objective of this research work is to develop models that forecast vegetation conditions at longer lead times on the premise that vegetation condition is controlled by factors like precipitation and soil moisture. To achieve this, we used a Bayesian Auto-Regressive Distributed Lag (BARDL) modelling approach which enabled us to factor in lagged information from Precipitation and Soil moisture levels into our VCI forecast model. The results showed a ∼2-week gain in the forecast range compared to the univariate AR model used as a baseline. The R2 scores for the Bayesian ARDL model were 0.94, 0.85 and 0.74, compared to the AR model's R2 of 0.88, 0.77 and 0.65 for 6, 8 and 10 weeks lead time respectively.

Download Full-text

Communicating Complex Forecasts for Enhanced Early Warning in Nepal

10.5194/gc-2019-3 ◽

2019 ◽

Cited By ~ 1

Author(s):

Mirianna Budimir ◽

Amy Donovan ◽

Sarah Brown ◽

Puja Shakya ◽

Dilip Gautam ◽

...

Keyword(s):

Early Warning ◽

Lead Time ◽

Warning System ◽

Early Warning Systems ◽

Warning Systems ◽

Time Data ◽

Deterministic Models ◽

Early Action ◽

Probabilistic Forecasts ◽

Barriers And Challenges

Abstract. Early warning systems have the potential to save lives and improve resilience. Simple early warning systems rely on real-time data and deterministic models to generate evacuation warnings; these simple deterministic models enable life-saving action, but provide limited lead time for resilience-building early action. More complex early warning systems supported by forecasts, including probabilistic forecasts, can provide additional lead time for preparation. However, barriers and challenges remain in disseminating and communicating these more complex warnings to community members and individuals at risk. Research was undertaken to analyse and understand the current early warning system in Nepal, considering available data and forecasts, information flows, early warning dissemination and decision making for early action. The research reviewed the availability and utilisation of complex forecasts in Nepal, their integration into dissemination (Department of Hydrology and Meteorology (DHM) bulletins and SMS warnings), and decision support tools (Common Alerting Protocols and Standard Operating Procedures), considering their impact on improving early action to increase the resilience of vulnerable communities to flooding.

Download Full-text

Multi-parametric observations of debris-flow initiation at the headwaters of the Gadria catchment (eastern Italian Alps)

10.5194/egusphere-egu2020-8720 ◽

2020 ◽

Author(s):

Velio Coviello ◽

Matteo Berti ◽

Lorenzo Marchi ◽

Francesco Comiti ◽

Giulia Marchetti ◽

...

Keyword(s):

Soil Moisture ◽

Debris Flow ◽

Early Warning ◽

Debris Flows ◽

Lead Time ◽

Early Warning Systems ◽

Force Balance ◽

Italian Alps ◽

Alarm Systems ◽

Debris Flow Initiation

The complete understanding of the mechanisms controlling debris-flow initiation is still an open challenge in landslide research. Most debris-flow models assume that motion suddenly begins when a large force imbalance is imposed by slope instabilities or the substrate saturation that causes the collapse of the channel sediment cover. In the real world, the initiation of debris flows usually results from the perturbation of the static force balance that retains sediment masses in steep channels. These perturbations are primarily generated by the increasing runoff and by the progressive erosion of the deposits. Therefore, great part of regional early warning systems for debris flows are based on critical rainfall thresholds. However, these systems are affected by large spatial-temporal uncertainties due to the inadequate number and distribution of rain gauges. In addition, rainfall analysis alone does not explain the dynamics of sediment fluxes at the catchment scale: short-term variations in the sediment sources strongly influence the triggering of debris flows, even in catchments characterized by unlimited sediment supply.In this work, we present multi-parametric observations of debris flows at the headwaters of the Gadria catchment (eastern Italian Alps). In 2018, we installed a monitoring network composed of geophones, three soil moisture probes, one tensiometer and two rain-triggered videocameras in a 30-m wide steep channel located at about 2200 m a.s.l. Most sensors lie on the lateral ridges of this channel, except for the tensiometer and the soil moisture probes that are installed in the channel bed at different depths. This network recorded four flow events in two years, two of which occurred at night. Specifically, the debris flows that occurred on 21 July 2018 and 26 July 2019 produced remarkable geomorphic changes in the monitored channel, with up to 1-m deep erosion. For all events, we measured peak values of soil water content that are far from saturation (<0.25 at -20 cm, <0.15 at -40 cm, <0.1 at -60 cm). We derived the time of occurrence and the duration of these events from the analysis of the seismic signals. Combining these pieces of information with data gathered at the monitoring station located about 2 km downstream, we could determine the flow kinematics along the main channel.These results, although still preliminary, show the relevance of a multi-parametric detection of debris-flow initiation processes and may have valuable implications for risk management. Alarm systems for debris flows are becoming more and more attractive due the continuous development of compact and low-cost distributed sensor networks. The main challenge for operational alarm systems is the short lead-time, which is few tens of seconds for closing a transportation route or tens of minutes for evacuating settlements. Lead-time would significantly increase installing a detection system in the upper part of a catchment, where the debris flow initiates. The combination of hydro-meteorological monitoring in the source areas and seismic detection of channelized flows may be a reliable approach for developing an integrated early warning - alarm system.

Download Full-text

Assessing the performance of eight real-time updating models and procedures for the Brosna River

Hydrology and Earth System Sciences ◽

10.5194/hess-9-394-2005 ◽

2005 ◽

Vol 9 (4) ◽

pp. 394-411 ◽

Cited By ~ 52

Author(s):

M. Goswami ◽

K. M. O'Connor ◽

K. P. Bhattarai ◽

A. Y. Shamseldin

Keyword(s):

Neural Network ◽

Linear Model ◽

Lead Time ◽

Ar Model ◽

Linear Perturbation ◽

Discharge Data ◽

Perturbation Model ◽

Simple Linear Model ◽

Auto Regressive ◽

Linear Perturbation Model

Abstract. The flow forecasting performance of eight updating models, incorporated in the Galway River Flow Modelling and Forecasting System (GFMFS), was assessed using daily data (rainfall, evaporation and discharge) of the Irish Brosna catchment (1207 km2), considering their one to six days lead-time discharge forecasts. The Perfect Forecast of Input over the Forecast Lead-time scenario was adopted, where required, in place of actual rainfall forecasts. The eight updating models were: (i) the standard linear Auto-Regressive (AR) model, applied to the forecast errors (residuals) of a simulation (non-updating) rainfall-runoff model; (ii) the Neural Network Updating (NNU) model, also using such residuals as input; (iii) the Linear Transfer Function (LTF) model, applied to the simulated and the recently observed discharges; (iv) the Non-linear Auto-Regressive eXogenous-Input Model (NARXM), also a neural network-type structure, but having wide options of using recently observed values of one or more of the three data series, together with non-updated simulated outflows, as inputs; (v) the Parametric Simple Linear Model (PSLM), of LTF-type, using recent rainfall and observed discharge data; (vi) the Parametric Linear perturbation Model (PLPM), also of LTF-type, using recent rainfall and observed discharge data, (vii) n-AR, an AR model applied to the observed discharge series only, as a naïve updating model; and (viii) n-NARXM, a naive form of the NARXM, using only the observed discharge data, excluding exogenous inputs. The five GFMFS simulation (non-updating) models used were the non-parametric and parametric forms of the Simple Linear Model and of the Linear Perturbation Model, the Linearly-Varying Gain Factor Model, the Artificial Neural Network Model, and the conceptual Soil Moisture Accounting and Routing (SMAR) model. As the SMAR model performance was found to be the best among these models, in terms of the Nash-Sutcliffe R2 value, both in calibration and in verification, the simulated outflows of this model only were selected for the subsequent exercise of producing updated discharge forecasts. All the eight forms of updating models for producing lead-time discharge forecasts were found to be capable of producing relatively good lead-1 (1-day ahead) forecasts, with R2 values almost 90% or above. However, for higher lead time forecasts, only three updating models, viz., NARXM, LTF, and NNU, were found to be suitable, with lead-6 values of R2 about 90% or higher. Graphical comparisons were made of the lead-time forecasts for the two largest floods, one in the calibration period and the other in the verification period.

Download Full-text

A non-linear neural network technique for updating of river flow forecasts

Hydrology and Earth System Sciences ◽

10.5194/hess-5-577-2001 ◽

2001 ◽

Vol 5 (4) ◽

pp. 577-598 ◽

Cited By ~ 67

Author(s):

A. Y. Shamseldin ◽

K. M. O’Connor

Keyword(s):

Neural Network ◽

Soil Moisture ◽

River Flow ◽

Model Updating ◽

Ar Model ◽

Soil Moisture Accounting ◽

Non Linear ◽

Auto Regressive ◽

Input Model ◽

Updating Procedure

Abstract. A non-linear Auto-Regressive Exogenous-input model (NARXM) river flow forecasting output-updating procedure is presented. This updating procedure is based on the structure of a multi-layer neural network. The NARXM-neural network updating procedure is tested using the daily discharge forecasts of the soil moisture accounting and routing (SMAR) conceptual model operating on five catchments having different climatic conditions. The performance of the NARXM-neural network updating procedure is compared with that of the linear Auto-Regressive Exogenous-input (ARXM) model updating procedure, the latter being a generalisation of the widely used Auto-Regressive (AR) model forecast error updating procedure. The results of the comparison indicate that the NARXM procedure performs better than the ARXM procedure. Keywords: Auto-Regressive Exogenous-input model, neural network, output-updating procedure, soil moisture accounting and routing (SMAR) model

Download Full-text

Benchmarking ensemble streamflow prediction skill in the UK

Hydrology and Earth System Sciences ◽

10.5194/hess-22-2023-2018 ◽

2018 ◽

Vol 22 (3) ◽

pp. 2023-2039 ◽

Cited By ~ 29

Author(s):

Shaun Harrigan ◽

Christel Prudhomme ◽

Simon Parry ◽

Katie Smith ◽

Maliko Tanguy

Keyword(s):

Soil Moisture ◽

Lead Time ◽

Base Flow ◽

Ensemble Prediction ◽

Flow Index ◽

Lead Times ◽

Streamflow Prediction ◽

Skill Scores ◽

Ensemble Streamflow Prediction ◽

The Uk

Abstract. Skilful hydrological forecasts at sub-seasonal to seasonal lead times would be extremely beneficial for decision-making in water resources management, hydropower operations, and agriculture, especially during drought conditions. Ensemble streamflow prediction (ESP) is a well-established method for generating an ensemble of streamflow forecasts in the absence of skilful future meteorological predictions, instead using initial hydrologic conditions (IHCs), such as soil moisture, groundwater, and snow, as the source of skill. We benchmark when and where the ESP method is skilful across a diverse sample of 314 catchments in the UK and explore the relationship between catchment storage and ESP skill. The GR4J hydrological model was forced with historic climate sequences to produce a 51-member ensemble of streamflow hindcasts. We evaluated forecast skill seamlessly from lead times of 1 day to 12 months initialized at the first of each month over a 50-year hindcast period from 1965 to 2015. Results showed ESP was skilful against a climatology benchmark forecast in the majority of catchments across all lead times up to a year ahead, but the degree of skill was strongly conditional on lead time, forecast initialization month, and individual catchment location and storage properties. UK-wide mean ESP skill decayed exponentially as a function of lead time with continuous ranked probability skill scores across the year of 0.75, 0.20, and 0.11 for 1-day, 1-month, and 3-month lead times, respectively. However, skill was not uniform across all initialization months. For lead times up to 1 month, ESP skill was higher than average when initialized in summer and lower in winter months, whereas for longer seasonal and annual lead times skill was higher when initialized in autumn and winter months and lowest in spring. ESP was most skilful in the south and east of the UK, where slower responding catchments with higher soil moisture and groundwater storage are mainly located; correlation between catchment base flow index (BFI) and ESP skill was very strong (Spearman's rank correlation coefficient =0.90 at 1-month lead time). This was in contrast to the more highly responsive catchments in the north and west which were generally not skilful at seasonal lead times. Overall, this work provides scientific justification for when and where use of such a relatively simple forecasting approach is appropriate in the UK. This study, furthermore, creates a low cost benchmark against which potential skill improvements from more sophisticated hydro-meteorological ensemble prediction systems can be judged.

Download Full-text

Comparison of wavelet and machine learning methods for regional drought prediction

10.5194/egusphere-egu21-218 ◽

2021 ◽

Author(s):

Gilbert Hinge ◽

Ashutosh Sharma

Keyword(s):

Machine Learning ◽

Lead Time ◽

Early Warning Systems ◽

Support Vector ◽

Lead Times ◽

Model Parameters ◽

Coupled Models ◽

Drought Prediction ◽

Forecasting Performance ◽

Grid Points

Droughts are considered as one of the most catastrophic natural disasters that affect humans and their surroundings at a larger spatial scale compared to other disasters. Rajasthan, one of India's semiarid states, is drought inclined and has experienced many drought events in the past. In this study, we evaluated different preprocessing and Machine Learning (ML) approaches for drought predictions in Rajasthan for a lead-time of up to 6 months. The Standardized Precipitation Index (SPI) was used as the drought quantifying measure to identify the drought events. SPI was calculated for 3, 6, and 12-month timescales over the last 115-year using monthly rainfall data at 119 grid stations. &#160;ML techniques, namely Artificial Neural Network (ANN), Support Vector Regression (SVR), and Linear Regression (LR), were used to evaluate their accuracy in drought forecasting over different lead times. Furthermore, two data processing methods, namely the Wavelet Packet Transform (WPT) and Discrete Wavelet Transform (DWT), have also been used to enhance the aforementioned ML models' predictability. At the outset, the preprocessed SPI data from both the methods were used as inputs for LR, SVR, and ANN to form a hybrid model. The hybrid models' drought predictability for a different lead-time was evaluated and compared with the standalone ML models. The forecasting performance of all the models for all 119 grid points was assessed with three statistical indices: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Nash-Sutcliffe Efficiency (NSE). RMSE was used to select the optimal model parameters, such as the number of hidden neurons and the number of inputs in ANN, and the level of decomposition and mother wavelet in wavelet analysis. &#160;Based on these measures, the coupled model showed better forecasting performance than the standalone ML models. The coupled WPT-ANN model shows superior predictability for most of the grid points than other coupled models and standalone models. &#160;All models' performance improved as the timescale increased from 3 to 12 months for all the lead times. However, the model performance decreased as the lead time increased.&#160; These findings indicate the necessity of processing the data before the application of any machine learning technique. The hybrid model's prediction performance also shows that it can be used for drought early warning systems in the state.

Download Full-text

Forecast of temperature-attributable mortality at lead times of up to 15 days for a very large ensemble of European regions

10.5194/egusphere-egu21-4107 ◽

2021 ◽

Author(s):

Marcos Quijal-Zamorano ◽

Desislava Petrova ◽

Xavier Rodó ◽

Èrica Martinez-Solanas ◽

Joan Ballester

Keyword(s):

Early Warning ◽

Lead Time ◽

Early Warning Systems ◽

Attributable Mortality ◽

Lead Times ◽

Warning Systems ◽

Epidemiological Models ◽

European Regions ◽

Large Ensemble ◽

Weather Forecasts

Implementing adequate health preventing measures is essential for public health decision making, particularly in the current context of rising temperatures. Most of the early warning systems are only based on climate data, and in very few cases they truly model the actual impact of the climate phenomena.Here we establish, for the first-time, the theoretical basis for the development of operational heat-health early warning systems that combine climate and health data. We studied the predictability of Temperature Attributable Mortality (TAM) at lead times of up to 15 days for a very large ensemble of European regions. To achieve this goal, we analysed daily counts of all-cause mortality for the period 1998-2012 in 147 NUTS2 regions in 16 European countries, representing more than 400 million people, and daily high-resolution weather forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF). We applied epidemiological models for the fitting of the temperature-mortality relationship in each of the regions, accounting for the different vulnerabilities and socio-demographic characteristics existing in Europe. We compared the predictive skill of the temperature and health forecasts on seasons and days with higher mortality risk.&#160;We conclude that the predictability of temperature can be used to issue skilful forecasts of TAM. In general, the predictability limit of temperature is similar to the one of TAM, which implies that the use of epidemiological models to transform the climate variables into health information does not reduce the lead time limit with significant forecast skill. Nonetheless, the spatial heterogeneity of the predictability lead time for TAM is higher than for temperature, especially in summer, where the complex shape of the temperature-mortality association amplifies the forecast errors. Overall, we find&#160; a nearly-linear relationship between the predictability of temperature and TAM for different seasons and regions, suggesting that future improvements in the predictability of temperature could automatically lead to improvements in the predictability of TAM.

Download Full-text

Designing a multi-objective framework for forecast-based action of extreme rains in Peru

10.5194/egusphere-egu2020-6003 ◽

2020 ◽

Author(s):

Jonathan Lala ◽

Juan Bazo ◽

Paul Block

Keyword(s):

Disaster Preparedness ◽

Disaster Response ◽

Lead Time ◽

Early Warning Systems ◽

Red Cross ◽

Warning Systems ◽

Short Term ◽

Early Action ◽

Wide Range ◽

Technological Constraints

The last few years have seen a major innovation within disaster management and financing through the emergence of standardized forecast-based action protocols. Given sufficient forecasting skill and lead time, financial resources can be shifted from disaster response to disaster preparedness, potentially saving both lives and property. Short-term (hours to days) early warning systems are common worldwide; however, longer-term (months to seasons) early actions are still relatively under-studied. Seeking to address both, the Peruvian Red Cross has developed an Early Action Protocol (EAP) for El Ni&#241;o-related extreme precipitation and floods. The EAP has well-defined risk metrics, forecast triggers, and early actions ranging from 5 days to 3 months before a forecasted disaster. Changes in climate regimes, forecast technology, or institutional and financial constraints, however, may significantly alter expected impacts of these early actions. A robust sensitivity analysis of situational and technological constraints is thus conducted to identify benefits and tradeoffs of various actions given various future scenarios, ensuring an adaptive and effective protocol that can be used for a wide range of changing circumstances.

Download Full-text

Ability to Forecast Regional Soil Moisture with a Distributed Hydrological Model Using ECMWF Rainfall Forecasts

Journal of Hydrometeorology ◽

10.1175/2008jhm1035.1 ◽

2009 ◽

Vol 10 (2) ◽

pp. 544-554

Author(s):

J. M. Schuurmans ◽

M. F. P. Bierkens

Keyword(s):

Soil Moisture ◽

Spatial Pattern ◽

Lead Time ◽

Hydrological Model ◽

Root Zone ◽

Weather Prediction ◽

Numerical Weather Prediction Model ◽

Lead Times ◽

Spatial Average ◽

Distributed Hydrological Model

Abstract This study mimics an online forecast system to provide nine day-ahead forecasts of regional soil moisture. It uses modified ensemble rainfall forecasts from the numerical weather prediction model of the European Centre for Medium-Range Weather Forecasts (ECMWF), which is provided by the Royal Netherlands Meteorological Office (KNMI). Both the individual ensembles as well as the mean of the ensembles are used as input for a hydrological model of a 70-km2 study area during March–November 2006. The outcomes are compared to the model run with high-resolution rainfall fields (based on 14 rain gauges within the study area and meteorological radar) as input. It is shown that the total spatial mean rainfall is forecasted very well for all lead times. The measured rainfall (spatial mean) shows a distribution with peaks at 0–1 and >10 mm day−1. These peaks are underestimated by the ensemble mean of the forecasts and this underestimation increases with lead time. This is not the case when ensemble members are used. Besides, the modeled uncertainty in rainfall by ECMWF underestimates the true uncertainty for all lead times and the number of rainfall events (thresholds 0.1, 0.5, and 1.0 mm) is overestimated. Absolute temporal mean bias values in root zone storage—that is, soil moisture—larger than 1 mm start to show for lead times longer than 3 days. The lower and upper bounds of bias for a lead time of 9 days are approximately −4 and 7 mm, respectively (negative values mean the forecasted soil moisture is underestimated). The bias in root zone storage shows a spatial pattern that represents the spatial pattern of total rainfall: areas with less rainfall than spatial average show a negative bias and vice versa. Local differences within this spatial pattern are due to land use and soil type. The results suggest that ensemble forecasts of soil moisture using ensemble rainfall forecasts from ECMWF are of practical use for water management, even at regional scales.

Download Full-text

Benchmarking Ensemble Streamflow Prediction skill in the UK

10.5194/hess-2017-449 ◽

2017 ◽

Cited By ~ 6

Author(s):

Shaun Harrigan ◽

Christel Prudhomme ◽

Simon Parry ◽

Katie Smith ◽

Maliko Tanguy

Keyword(s):

Soil Moisture ◽

Lead Time ◽

Mean Squared Error ◽

Low Cost ◽

Ensemble Prediction ◽

Lead Times ◽

Streamflow Prediction ◽

Skill Scores ◽

Ensemble Streamflow Prediction ◽

The Uk

Abstract. Skilful hydrological forecasts at sub-seasonal to seasonal lead times would be extremely beneficial for decision-making in water resources management, hydropower operations, and agriculture, especially during drought conditions. Ensemble Streamflow Prediction (ESP) is a well-established method for generating an ensemble of streamflow forecasts in the absence of skilful future meteorological predictions, instead using Initial Hydrological Conditions (IHCs), such as soil moisture, groundwater, and snow, as the source of skill. We benchmark when and where the ESP method is skilful across a diverse sample of 314 catchments in the UK and explore the relationship between catchment storage and ESP skill. The GR4J hydrological model was forced with historic climate sequences to produce 51-member ensemble of streamflow hindcasts. We evaluated forecast skill seamlessly from lead times of 1-day to 12-months initialised at the first of each month over a 50-year hindcast period from 1965–2015. Results show ESP was skilful against a climatology benchmark forecast in the majority of catchments across all lead times up to a year ahead, but the degree of skill was strongly conditional on lead time, forecast initialisation month, and individual catchment location and storage properties. UK-wide mean ESP skill decays exponentially as a function of lead time with mean squared error skill scores across the year of 0.839, 0.303, and 0.179 for 1-day, 1-month, and 3-month lead times, respectively. However, skill was not uniform across all initialisation months. For lead times up to 1-month, ESP skill was higher than average when initialised in summer and lower in winter, whereas for longer seasonal and annual lead times skill was highest when initialised in autumn and winter months and lowest in April. ESP is most skilful in the south and east of the UK, where slower responding catchments with higher soil moisture and groundwater storage are mainly located; correlation between catchment Baseflow Index (BFI) and ESP skill was very strong (ρ = 0.896 at 1-month lead time). This is in contrast to the more highly responsive catchments in the north and west which are generally not skilful at seasonal lead times. Overall, this work provides a scientifically defensible justification for when and where use of such a relatively simple forecasting approach is appropriate in the UK and creates a low cost benchmark against which potential skill improvements from more sophisticated hydro-meteorological ensemble prediction systems can be judged.

Download Full-text