1. 1 Optimal Forest Management Under Risk International Seminars in Life Sciences Universidad Politécnica de Valencia Thursday 2007-02-22 Peter Lohmander Professor of Forest Management and Economic Optimization SLU, Swedish University of Agricultural Sciences Umea, Sweden http://www.Lohmander.com Version 2007-02-18

2. 2 Schedule of Peter Lohmander from Claudio Benavent,International Officer ETSMRE - ETSIA Thursday 2007-02-22 10:30 Welcome at ETSMRE 10:45-13:45 Visits on campus and interviews with UPV Colleagues 14:00 Lunch on campus offered by ETSMRE 16:30 Conference in International Seminars 19:30 Free programme Friday 2007-02-23 10:30 Interview with Prof. Penny McDonald, Coordinator of the course "Preparation for International Study" 11:00-12:15 Institutional Presentation of your home Institution 12:30-13:30 International Office ETSMRE 14:00 Lunch on campus offered by ETSMRE 15:30 Free programme Location: ETSMRE International Office (Avda. Blasco Ibáñez, 19-21).

3. 3 AMBITION Research that leads to economically profitable and practical solutions and at the same time leads to new discoveries of methods and practical general approaches to problems in operations research in general and to forest economics. “Environmentally friendly” solutions are sometimes discovered also to be economically optimal!

4. 4 Stochastic Dynamic Optimization The future state of the world is hard to predict perfectly. Some decisions must be made before the future is perfectly known. Stochastic Dynamic Programming is the relevant approach.

5. 5 The unpredictable world and the decision problems

6. 6 Figure 1. The real (inflation adjusted) stumpage price in Sweden. Source: Swedish Board of Forestry, Yearbook of Forest Statistics, 2000. We may regard the stumpage price as a stochastic processes. There is no method available which can predict future prices without error.

7. 7

8. 8

9. 9

10. 10

11. 11

12. 12

13. 13

14. 14 Stochastic Price

15. 15 Stochastic Price

16. 16 Low Correlation between Energy Prices and Pulp Prices (Source: Statistics Sweden)

17. 17 Low Correlation between Energy Prices and Pulp Prices Price of Electricity (li) Export Price of Kraft Paper and Kraft Board Price of Electricity (li) 10,2450 Export Price of Kraft Paper and Kraft Board 0,24501

18. 18 Low Correlation between Energy Prices and Pulp Prices It has been proved that the expected marginal capacity value of a production plant increases with price variation when different products are produced with the same type of raw material and the correlation between product prices is less than 1. (Lohmander 1989)

19. 19 Low Correlation between Energy Prices and Pulp Prices As a consequence, the most profitable investment level in production capacity, for instance a power plant, is higher with prices that are not perfectly predictable than according to what you find with traditional calculation.

20. 20 Joint probability density function with correlation 0.25 (which corresponds to the prices of electricity and kraft paper)

21. 21 Stochastic dynamic example with heating and pulp plants P1 P2 P1 P2 Time The prices of electricity and kraft paper are not known many years in advance. P1

22. 22 Stochastic dynamic example with heating and pulp plants Time The stock level can be changed over time. The most profitable extraction (harvest) in a particular period is affected by the prices of kraft paper and energy. This is one reason why it has to be sequentially optimized, based on the latest price information from the markets. Stock level

23. 23 Stochastic dynamic example with heating and pulp plants P1 P2 P1 P2 Time Stock level P1

24. 24 The stochastic dynamic optimization problem We maximize the expected present value of all future production. The production of electricity and kraft paper in future periods is affected by the product prices and the stock of resources. The stock of resources is dynamically optimized.

25. 25 The stochastic dynamic optimization problem The optimal expected present value, f, as a function of time, the stock level and the prices electricity and kraft paper.

26. 26 The stochastic dynamic optimization problem The profit in a particular period, t, as a function of the production levels of electricity and kraft paper, time, the stock level and the prices of electricity and kraft paper.

27. 27 The stochastic dynamic optimization problem The cost of the stock in a period as a function of time, the stock level and the production levels of electricity and kraft paper. (The production in period t affects the stock level in period t and in period t+1.)

28. 28 The stochastic dynamic optimization problem The production of electricity and kraft paper in a period, t, is constrained by the production capacities in the kraft paper mill and the energy mill in that period and the entering resource stock level.

29. 29 The stochastic dynamic optimization problem The expected optimal objective function value of period t+1 is discounted to period t. The probabilities of reaching different market state combinations at t+1 in the electricity market and in the kraft paper market are conditional on the prices in these markets at t.

30. 30 The stochastic dynamic optimization problem The total optimization problem is found above. Now, we will illustrate this with a numerical program!

31. 31 X1 X2 Production capacity 2 Production capacity 1 Total wood supply General illustration why the marginal value of production capacity increases with price risk (and connection to heating plants)

32. 32 The economic optimization problem Along the iso profit line we have:

33. 33 X1 X2 Production capacity 2 Production capacity 1 Total wood supply Isoprofit line

34. 34 X1 X2 Production capacity 2 Production capacity 1 Total wood supply Isoprofit line

35. 35 X1 X2 Production capacity 2 Production capacity 1 Total wood supply Isoprofit line

36. 36

37. 37 Link to the software: http://www.lohmander.com/CDP5L.htm

38. 38

39. 39

40. 40

41. 41

42. 42

43. 43

44. 44

45. 45

46. 46 Results: The expected economic value of one more unit of heating plant capacity is 17551 – 16461 = 1090. The economically optimal decision is this: If the investment cost of an extra unit of capacity is less than 1090: Build this extra heating plant capacity! No other investment calculation method would give the correct rule.

47. 47 Conclusions from the numerical model: It is possible to adaptively optimize all decisions over time including production of electricity, kraft paper and resource extraction. The approach makes it possible to determine the expected value of production capacity investments in heating plants and paper mills. The approach can be expanded to cover the complete energy and forest sector.

48. 48 Already in 1981 World Bank Model” to study the Swedish forest sector. (Nilsson, S.) In the model, timber, pulp wood and fuel wood could be produced and harvested in all regions. The energy industry was considered as an option in all regions. It was possible too burn wood, not only fuel wood but also “pulp wood”.

49. 49 Capacity investments The existing capacity in the saw mills, pulp mills and paper mills was investigated and used in the model. It was possible to invest in more capacity of different kinds in the different regions.

50. 50 Structure in 1981 The forest sector of Sweden was modelled as a linear programming problem. The total economic result of all activities in the forest sector of Sweden was maximized. The wood based part of the energy sector was considered as a part of this forest sector.

51. 51 Wood for energy in 1981 Among these results, we found that a large proportion of the “pulpwood” should be used to produce energy. This was particularly the case in the north, at large distances from the coast.

52. 52

53. 53

54. 54

55. 55 Surprise? Not really! The cost of transporting pulpwood large distances is very high. If energy can be produced from pulpwood, far away from the coast and the pulp industry, it is not surprising that this may be the most profitable alternative.

56. 56 Relevant model in 1981? Of course, linear programming models are only models of reality. This is true with all models. Of course, linear programming models do not capture all nonlinear and other “real” properties of the real world such as risk and integer constraints. Better options exist today to handle nonlinearities, risk, integer constraints and all kinds of other properties of the real world.

57. 57 Relevant result from 1981? The general finding that it may be optimal to use some of the wood for energy, still remains!

58. 58 Questions today (#1): Can we combine the forest sector and the energy sector in one modern optimization model for both sectors? The model should include relevant data for the heating and electricity plants and for all types of forest industry mills.

59. 59 President Lars Fritiof, E.ON (March 2006): ”During the next three years, the E.ON Group will invest about SEK 175 billion, of which SEK 155 billion will be invested in plants.”

60. 60 President Lars Fritiof, E.ON (March 2006): ”Secure energy supplies can no longer be taken for granted. Demand is increasing and European reserves of oil and gas are diminishing.The recently published EU Green Book reports that if nothing is done, Europe’s imports of energy will increase from 50 to 70% in the next 20 to 30 years.”

61. 61 Necessary Model Properties: The model should be dynamic and include the options to invest in new production capacity. Such new capacity could, when it comes to investments in energy plants, have different properties with respect to technological choices, possible fuels and degrees of flexibility.

62. 62 Why flexibility? Prices and the availability of different fuels are impossible to predict over horizons of the economic life time of a heating plant. That is why flexibility is valuable. In the old type of optimization models, such things could not be analyzed at all. Now, economic optimization of flexibility is possible.

63. 63 Dynamic options In the model from 1981, one period was analysed. In a new dynamic model, the use of the forest resources can also be optimized over time. In the model from 1981, the capacities of different mills was constant. In the dynamic model, the capacity investments can be optimized over time.

64. 64 The Option A new generation of optimization models is possible to construct. We should not hesitate to develop this generation!

65. 65 Conclusions Optimal forest management under risk contains a large number of topics. The relevant general approach is stochastic dynamic programming.

66. 66 Conclusions Here you may read more about these things: http://www.lohmander.com/Information/Ref.htm

67. 67 Conclusions Optimal forest management under risk contains a large number of topics. Here, you may instantly optimize some decisions! http://www.lohmander.com/Program/Program.htm

68. 68 Conclusions Here, you may study some courses: http://www.lohmander.com/Kurser/Kurser.htm In particular: Optimization in dynamic and stochastic decision problems (Graduate course) Forest Economics (Graduate course) Economic Forest Production (Advanced Undergraduate course) The Forest Sector from an International View (Advanced Undergraduate course)

69. 69

70. 70 INFORMS The Institute for Operations Research and the Management Sciences (INFORMS) is the largest professional society in the world for professionals in the field of operations research (O.R.). It was established in 1995 with the merger of the Operations Research Society of America (ORSA) and The Institute of Management Sciences (TIMS).

71. 71 INFORMS 2007 July 8-11, 2007 INFORMS International Puerto Rico Westin Rio Mar Beach Resort & Spa Rio Grande, Puerto Rico O.R. in the Forest Sector O.R. in the Forest Sector is one of the clusters of INFORMS 2007. The sessions in this cluster will include the most interesting forest sector operations research applications from all countries. Contact: Cluster Chair: Professor Peter Lohmander, SLU, Faculty of Forest Sciences, SE- 901 83 Umea, Sweden. e-mail: peter.lohmander@sekon.slu.sepeter.lohmander@sekon.slu.se

72. 72 Thank you for listening! Here you may reach me in the future: Peter Lohmander Professor of Forest Management and Economic Optimization, SLU, Swedish University of Agricultural Sciences, Faculty of Forest Sciences, Dept. Of Forest Economics, SE-901 83 Umea, Sweden http://www.Lohmander.com peter.lohmander@sekon.slu.se plohmander@hotmail.com Version 2007-02-18