Probabilistic Drought Hazard and Risk Model: A contribution of the Risk Nexus Initiative

Transcription

1 Workshop on Developing a Drought Monitoring, Early Warning and Mitigation System for South America 8 10 August 2017 Buenos Aires, Argentina Probabilistic Drought Hazard and Risk Model: A contribution of the Risk Nexus Initiative Omar Dario Cardona / Gabriel Bernal August 9, 2017 INGENIAR Bogotá Colombia

2 The UN Risk Nexus Initiative (RNI): A new partnership Risk is the nexus between adaptation to climate change, environment management, and risk reduction policies for sustainability and resilience. RNI will go beyond the fragmented policy approaches in favor of a holistic approach to risk, which explicitly recognizes and embraces interdependence. 2

3 The UN Risk Nexus Initiative (RNI): A new partnership The principal outputs: New risk models and metrics that condense interdependence and inform investment decisions, Indicators that transcend fragmented policy approaches in order to monitor progress towards sustainability and resilience, and Enhanced risk governance based on forensic studies of risk, new conceptual frameworks, valuations of risk, and cutting edge tools of social communication. 3

4 Drought Hazard and Risk Assessment: New component of the GRM-A The objective is to develop a drought risk model to estimate the economical losses in the agricultural sector globally (water and food stress). The proposed methodology aims to quantify the loss in production (yield) of crops exposed to droughts. Any other adverse effect of droughts is not considered within the scope of this presentation. The expected outcome is the risk assessment in terms of probabilistic metrics (AAL, PML, LEC) at country level.

5 Probabilistic Risk Assessment HAZARD Set of stochastic scenarios Based on historical climate data Probabilistic representation Spatially distributed Time series All return periods are included Possibility of including CC trends Damage RISK EXPOSURE Cultivated land units Location Crop type and seasonality Cultivated area Production cost Loss Production Economic VULNERABILITY Crop yield response to water Crop characteristics Crop calendar Climate time series Loss in potential production Metrics Loss exceedance curve (LEC) Average annual loss (AAL) Probable maximum loss (PML)

6 Hazard Historic records For example, in northeast Brazil

7 Hazard Historic records For example, in northeast Brazil

8 Hazard Simulated weather time series Simulated time series are generated stochastically from the historical information. The objective is not to forecast future weather conditions, but to generate feasible combinations of drought conditions, such as low precipitation and high temperature. Historical Simulated (not a forecast)

9 Hazard Simulated weather time series Probability distribution of each weather value, for each day Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

10 Hazard Simulated weather time series Probability distribution fitting For example, Precipitation, March 15 th Temperature, October 15 th

11 Hazard Simulated weather time series Correlation There is a certain amount of correlation, in both time and space, that cannot be neglected var var 2 var var 4 Correlation is considered by means of a covariance matrix, between days of the year (for time correlation) and stations in the study area (for space correlation) var 5 var 6 var 7 var 8 var 9 var var var

12 Hazard Simulated weather time series Simulations Correlated random numbers are generated for each day of the year, for as may years as wanted. Each simulated year has different values of the weather parameters in each day, which follow the day-specific probability distribution and the temporal and spatial correlations. Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

13 Hazard Simulated weather time series Simulations This is repeated, for example, one thousand times. Temperature [ C] Precipitation [mm/day] Simulations Max Mean Min

14 Hazard Simulated weather time series Simulations Fitness check of multi-annual averages. Precipitation [mm/day] Daily precipitation mean Historic records 1000 years Temperature mean Temperature [ C]

15 Hazard Climate change impact on hazard modelling Precipitation

16 Hazard Climate change impact on hazard modelling Temperature

17 Hazard Climate change impact on hazard modelling Select the GCM that better describes the area Difference in P [mm/d] 1 ST DEV 2 ST DEV RCP2.6 RCP4.5 RCP6 RCP MIROC5 Run 1 RCP 2.6, 4.5, 6 y 8.5 MIROC5 Run 2 RCP 2.6, 4.5 y 8.5 MIROC5 Run 3 RCP 2.6, 4.5 y 8.5 MIROC5 Mean RCP 2.6, 4.5 y Difference in T mean [ C]

18 Drought Classification Drought indices according to drought type Precipitation deciles SPI Standardized Precipitation EDI Effective drought index Precipitation RDI Reconnaissance Drought Index SPEI Standardized Precipitation Evapotranspiration Index Precipitation Temperature PDSI Palmer Drought Severity Index Precipitation, Temperature, Soil characteristics (porosity, moisture) Surface Water Supply Index Precipitation, snow, streamflow, water storage

19 Drought Indices Time scale variation Drought indicator series for 1000-years simulation

20 Drought Events An index threshold must be defined to identify drought events. Drought Classes Index value Non-drought Index 0 Mild -1 < Index<0 Moderate -1.5 < Index -1 Severe/extreme Index -1.5 Index time series Index threshold Individual drought events can be identified from the index time series

21 Drought Events Each event is unique and has its own distribution in time 2 Index Drought event Time Severity Intensity Duration 1 month Temperature Precipitation

22 Drought Events Drought index series are associated to particular stations.

23 Drought Events A regional event is identified when drought events are simultaneously identified in a substantial number of stations. 3 Scenario 1 Scenario 2 Scenario 3 Scenario 4 RDI Station RDI Station RDI Station RDI Station n RDI Station n Months

24 Drought Hazard Collection of scenarios Hazard is represented as a set of stochastic scenarios. These scenarios (events) are assumed to be: Mutually exclusive Collectively exhaustive They allow probabilistic representation: Occurrence frequency (temporal probability) Gridded statistical moments (spatial probability) Time series, at any location, of weather variables (precipitation and temperature)

25 Exposure Cultivated Land Units Exposure is defined within Cultivated Land Units. Each unit is characterized by the following properties: Geographical location and area Type of crops produced Crops production cost Participation of each crop in the total production

26 Exposure Crop calendar The Cultivated Land Units will be characterized by using the crop calendar to define crop seasonality.

27 Exposure Crop valuation Crop production costs include expenses associated with raw materials (seeds, fertilizers, irrigation), labor and machinery investments. Industrial Agricultural Electrodes Electric energy Chemicals Maintenance Workforce Lubricants Fertilizers Herbicides Pesticides Workforce Administrative Workforce

28 Exposure Soil types The geographical distribution of soil types is required as part of the exposure information.

29 Crop Yield Response to Water As crop seasonality is known, crop calendar is located in the same time-scale for each scenario. Crop development t Drought Index t Drought threshold Sowing Harvest Maturity

30 Vulnerability 5-steps process Soil water balance Crop development Crop transpiration Biomass production Yield (Y) from biomass (B)

31 Crop Vulnerability Crop seasonality P [mm] Sep Oct Nov Dec Jan Feb ETo [mm] T [ C] RDI B, Bx [ton/ha] 20 0 Sep Oct Nov Dec Jan Feb

32 Crop Vulnerability Crop seasonality P [mm] Sep Oct Nov Dec Jan Feb ETo [mm] T [ C] RDI Drought scenario 40 B, Bx [ton/ha] 20 0 Sep Oct Nov Dec Jan Feb

33 Crop Vulnerability Simulations Uncertainty in input parameters is included from the avaliable information and expert criteria (likelihood) Crop characteristics Inputs Soil profile Random variables

34 Crop Vulnerability Simulations Y, Yx [ton/ha] Potential yield (random variable) Actual yield (random variable) Time

35 Crop Vulnerability Simulations Y, Yx [ton/ha] Potential yield (random variable) Actual yield (random variable) Time

36 Probabilistic Risk Assessment Hazard Exposure Vulnerability Loss

37 Probabilistic Risk Assessment Hazard E( l Event 2 ( l Event ) NE E( i l j j 1 NE 2 i ) ( l j j 1 ) ) 2 Loss aggregation NE 1 NE k 1 j 2 k j cov l k, l j Exposure PDF of the loss for each scenario Vulnerability Loss Area of the cultivated unit Number of Crop Types in the unit Participation (%) of type k l i A NCT k 1 C k P Loss for scenario i k Yx k, i Y k, i Production cost of type k Potential yield of type k for scenario i Actual yield of type k for scenario i

38 Probabilistic Risk Assessment Hazard Exposure Vulnerability Risk Loss

39 Probabilistic Risk Assessment Hazard Risk Exposure Loss by scenario Vulnerability Production Economic Loss Loss Exceedance Curve

40 Probabilistic Risk Assessment Aggregation of losses

41 Probabilistic Risk Assessment Aggregation of losses

42 Probabilistic Risk Assessment Aggregation of losses

43 Probabilistic Risk Assessment Aggregation of losses

44 Probabilistic Risk Assessment PDF of the loss for one scenario f(l) p( ) L l Pr L l L f l dl

45 Probabilistic Risk Assessment f(l) f(l) L l L l Loss exceedance rate Loss exceedance probability v l N i 1 Pr ( l L E i ) F i Occurrence frequency of each scenario

46 Probabilistic Risk Assessment f(l) f(l) L l L l Loss exceedance rate v l N i 1 Loss exceedance probability Pr ( l L E i ) F i Occurrence frequency of each scenario Loss exceedance rate (#/year) Loss

47 Probabilistic Risk Assessment Loss Exceedance Curve (LEC) Sets the annual rate of excess of a given loss value v l Exceedance rate [#/year] Loss l

48 Probabilistic Risk Assessment Return period Average time needed for reaching or exceeding a loss value, considering a large enough time window Tr 1 v l It is computed as the inverse of the exceedance rate

49 Probabilistic Risk Assessment Probable Maximum Loss (PML) It is a loss that doesn t occur frequently (related to long return periods) PML Return period

50 Probabilistic Risk Assessment Average Annual Loss (AAL) It represents the amount that has to be paid annually in order to cover future expected losses. 4.50% 4.00% 3.50% 3.00% 2.50% 2.00% 1.50% 1.00% 0.50% Pérdidas por evento Pérdida acumulada 45.0% 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.00% %

51 Probabilistic Risk Assessment Average Annual Loss (AAL) It represents the amount that has to be paid annually in order to cover future expected losses. 4.50% 4.00% 3.50% 3.00% 2.50% 2.00% 1.50% 1.00% 0.50% Pérdidas por evento Pérdida acumulada 45.0% 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% AAL 0.00% Loss and time between events: unknown 0.0%

52 Probabilistic Risk Assessment A holistic approach The LEC provides information on direct physical losses. This metrics can me amplified by incorporating other aspects related to risk. Physical risk 1st Order EFFECTS Potential damage on crops Social, economical and ecological fragilities Lack of resilience or ability to cope and recover 2nd Order EFFECTS Potential Socio-ecological and Economic Impact on Communities- Organizations

53 Probabilistic Risk Assessment A holistic approach Assessment of the influence of each variable Social aspects Xs1, Xs2,., Xsn Economic aspects Xe1, Xe2,., Xen Infrastructural aspects Xi1, Xi2,., Xin Other aspects (governability, security ) Xo1, Xo2,., Xon Weights Ws1, Ws2,., Wsn Weights We1, We2,., Wen Weights Wi1, Wi2,., Win Weights Wo1, Wo2,., Won Aggravating coefficient F N i 1 X i W i

54 Probabilistic Risk Assessment A holistic approach Total risk Total risk Aggravating coefficient R T R 1 F Physical risk metric

55 Drought Risk Assessment Expected contributions of the GRM-A To provide a suite of drought risk metrics that will enable a better characterization of risk in the agriculture sector that allow establishing investment plans and strengthen social and economic resilience. To provide a complete estimate of drought risk and in particular in those countries where agriculture makes a significant contribution to national economies. To estimate correlated risk that could lead to systemic food insecurity (multiple breadbasket failures).