Thornthwaite's New System of Climate Classification In Its Application to New Zealand
[Read before the Otago Branch, February 16, 1950; received by Editor, March 9, 1950]
Note. This paper forms part of an investigation into the regional variety of climatic type in New Zealand which is being financed by a grant from the University of New Zealand Research Grants Committee, to whom grateful thanks are tendered for assistance received.
Among modern climatologists there are few who have worked so consistently at the problems of the classification of climates as C. Warren Thornthwaite. His system propounded in 1931 (Thornthwaite, 1931) has gained increasing recognition and has been applied in several areas. The author has always been ready to recognize its limitations, especially its empirical nature, and in an attempt to remedy them somewhat he has recently used his proposals for the computation of potential evapotranspiration as the basis of a new classification of climate (Thornthwaite, 1948).
The 1931 and 1948 systems are, in outward appearance very similar. Thornthwaite has pointed this out himself, but claims that, in actual fact, they are basically different.
The earlier  study adopted Köppen's position that the plant is a meteorological instrument which integrates the various factors of climate and which, with experience, can be “read” like a thermometer or a rain gauge. In the present  study, vegetation is regarded as a physical mechanism by means of which water is transported from the soil to the atmosphere; it is the machinery of evaporation as the cloud is the machinery of precipitation.
Climate boundaries are determined rationally by comparing precipitation and evapotranspiration. The subdivisions of the older classification were justly criticized as being vegetation regions climatically determined. The present climate regions are not open to this criticism, since they come from a study of the climatic data themselves and not from a study of vegetation. (Thornthwaite, 1948, p. 88.)
It is presumably reasonable to suppose that the success of a classification of climate should be judged by its success in portraying the variety of climates over a given area. The results achieved by a classification, therefore, can be regarded of the utmost importance. However rational its basis, it is clearly of little worth unless the climatic pattern it portrays gives expression to the variety of climates known to exist in an area through the human experience of them and the observation of vegetation, soil, and other features closely associated with climate, due allowance being made for the modifications in climatic relationship resulting from cultural interference, natural disasters, past climatic change and, in the case of plants, migrations. This does not, of course, imply the fitting of climatic boundaries to vegetation and other boundaries. It does, however, suggest that the overall pattern of climates should display a general sympathy with the pattern of phenomena closely related to climate.
A feature of the climate of New Zealand is the variety of types which exists within the compass of so small a land. To pass, within the space of a few hours, from the “continental” atmosphere of the deep, interior valleys of Central Otago, to the “oceanic” conditions of Southland or the “ice age” conditions of the neighbouring glacial fields is to experience a rapid and sudden transmutation between different worlds. These contrasts are strong and are not limited to one element of climate, but are expressed in both thermal and moisture
conditions. Thornthwaite's 1931 system has been shown as effectively recognising these differences (Garnier, 1946) and the present study has been undertaken to see if the 1948 proposals provide a better or equally good indication of the climatic variety within the country.
Source Material for the Study1
In any study of climate in New Zealand one is handicapped by the absence of statistics for many places. Not only is the number of stations having both temperature and precipitation normals derived from periods of twenty years or more small, but the location of them, usually in centres of population, is not ideal from the viewpoint of portraying climatic variety. The number of rainfall recording stations is, however, considerable and totals over five hundred. These are widely distributed over the country, but temperature estimates for them must be arrived at by interpolation from those places for which statistics are available. In the present instance temperature figures were obtained by the application of a lapse rate of 2·74°F per 1,000 feet of elevation (Kidson, 1931a), by the use of temperature anomaly maps specially prepared for the work, and by graphs of the variation of average sea-level temperatures at various latitudes in New Zealand, also specially prepared in connexion with the present study.
The classification of climatic types under Thornthwaite's new system is a formidable task. Even with the use of a maximum number of calculating devices such as nomograms, slide rules, and the tables Thornthwaite has prepared it was found that a rate of not more than three stations an hour was the maximum achieved. Where temperature normals were absent, therefore, full monthly calculations were not made, since the amount of work involved was not considered justified in view of the approximate nature of the source statistics. To obtain moisture categories for rainfall stations where no temperature records are kept, therefore, reference was made to a graph (see Figure 1) which was devised from the results of 53 stations for which full computations were made from monthly temperature and precipitation normals.2 Thermal categories for the rainfall stations were obtained on the basis shown in Table I, which was also prepared from the 53 results mentioned above.
[Footnote] 1 The statistics on which this study is based were obtained from the Meteorological Office, Wellington, through the courtesy of the Director of Meteorological Services.
[Footnote] 2 Of these 53 stations the record for three was less than 10 years, for 12 was less than 20 years, and for the remaining 38 was over 20 years.
[Footnote] 1 These are mapped in Figure 2. For reference purposes Table 2, showing the main elements of Thornthwaite's classification proposals, has been prepared.
|Thermal Efficiency Group||Northern Areas||Central Areas||Southern Areas|
|Mesothermal (B′2)||0–500 ft. (locally)||Not found||Not found|
|Mesothermal (B′1)||500–3,500 ft.||0–2,500 ft.||0–1,500 ft.|
|Microthermal (C′2)||3,500–6,500 ft.||2,500–5,650 ft.||1,500–4,500 ft.|
|Microthermal (C′1)||6,500–9,500 ft.||5,650–8,700 ft.||4,500–7,700 ft.|
|Tundra (D′)||9,500–12,500 ft.||8,700–11,500 ft.||7,700–10,250 ft.|
|Frost (E′)||above 12,500 ft.||above 11,500 ft.||above 10,250 ft.|
Note. Northern areas refers to approximately north of 39°S; Central areas refers to regions between 39°S and 44°S; and Southern areas are south of 44°S.
|Climatic Type||Moisture Index||Seasonal Variation of Moisture Efficiency|
|A Perhumid||100 & above||r little or no water deficiency||0–16.7|
|B4 Humid||80 to 100||s moderate summer water deficiency||16.7–33.3|
|B3 Humid||60 to 80||w moderate winter water deficiency||16.7–33.3|
|B2 Humid||40 to 60||s2 large summer water deficiency||33.3+|
|B1 Humid||20 to 40||w2 large winter water deficiency||33.3+|
|C2 Moist Subhumid||0 to 20|
|C1 Dry Subhumid||−20 to 0||d little or no water surplus||0–10|
|s moderate winter water surplus||10–20|
|D Semiarid||−40 to −20||w moderate summer water surplus||10–20|
|s2 large winter water surplus||20+|
|E Arid||−60 to −40||w2 large summer water surplus||20+|
|Potential Evapotranspiration||Temperature Efficiency Type||Summer Concentration|
In New Zealand all classes of moisture category from perhumid (A) to semiarid (D) are represented. In both islands it is the perhumid group which has the greatest extent. The mountains and west coast of the South Island fall into this category as do the high-ground areas of the North Island and also the middle portion of the North Auckland peninsula. The remainder of the North Island, except for a small tract between Napier and Hastings, is humid, all four subdivisions of this type being found. North of the plateau the driest subtype of this class is B3, as found in the Waikato-Hauraki lowlands and the Kaipara Harbour area. South of the plateau, moisture categories grade from B4 in the higher areas to B1 in the vicinity of Palmerston North and Wanganui and also in eastern Hawke's Bay. In the South Island humid climates occupy the broad lowland areas of the south and reach the sea in descending order from B4 to B1 in the Nelson area. Elsewhere they form a narrow transitional belt between the perhumid mountains and the subhumid, for the most part moist subhumid (C2) eastern lowlands. Dry subhumid (C1) conditions are found between Timaru and Oamaru and extending up the Waitaki valley to about Kurow and also surrounding the semiarid (D) core of Central Otago, centred upon Alexandra.
The pattern of moisture areas thus briefly described is very similar to that displayed on the basis of the 1931 system of classification.1 The closeness of this similarity is seen by comparing the moisture indices under the 1948 classification with the P–E indices of the 1931 system.2 The results of 53 observations are grouped in Table III. This
|Ratio of P–E 1948 to P–E 1931||Range of 1931 P–E Indices||Total number of cases|
|Under 60||60 to 80||Over 80|
|1.00 to 1.10||5||13||2||20|
grouping shows a tendency for the range of values under the 1948 system to be extended as it were both ways. Where the 1931 system gave P–E indices of 80 or more the tendency is for the ratio of 1948 P–E index to 1931 P–E index for a given station to be greater than 1·10. Where 1931 P–E indices were less than 60 the tendency was for the ratio to be less than 1·0, and ratios of between 1·0 and 1·10 were generally found for 1931 P–E indices of between 60 and 80. The extreme ratio values between computations under the two systems were 0·95 (Manorburn Dam, 1931 P–E index 50) and 1·27 (Gisborne, 1931 P–E index 73).
[Footnote] 1 Readers who compare the present map with that appearing in B. J. Garnier, op. cit., will notice that the former appears to be more detailed than the latter. This is because the more recent map has been prepared from more data than the earlier one. A revision of the map showing 1931 classifications has recently been published in “New Zealand Weather and Climate,” ed. B. J. Garnier, A Special Publication of the New Zealand Geographical Scoiety, Misc. Series, No. 1, 1950.
[Footnote] 2 The moisture indices of the 1948 classification were converted to P–E indices by the use of the formula P–E = .8I + 48, where I = 1948 moisture index. For the sake of briefness figures derived from this calculation will be referred to as the 1948 P–E Index to distinguish them from the 1931 P–E Index.
The essential variety of moisture areas in New Zealand seems, therefore, to be brought out equally well by either Thornthwaite's 1931 or 1948 system of classification. Such differences as do exist are minor in character or, if anything, favour the earlier proposals. A case in point is in the vicinity of Karioi. This is classed as B4 (1931 P–E Index 114) under the old classification and as A (1948 P–E Index 130) under the new system. The latter arrangement places Karioi within the perhumid area of the central North Island, yet it is at Karioi that an abrupt change takes place in landscape aspects. Northwards is the heavily forested country near Mount Ruapehu with, near Ohakune, gaunt skeletons of trees where burning has taken place in recent times to enable dairying and vegetable growing to take place. South of Karioi one finds open country with the spaciousness of a grassland environment and, near Waiouru and the Desert Road, a tussock setting of treeless, rolling country, which is reminiscent of South Island subhumid conditions. While one cannot discount the importance of historical development, clearing policy, and volcanic ash showers, climatic differences are also there and the vicinity of Karioi seems more allied to the region to the south than to the north. Similarly, the disappearance of B1 in the Wairarapa valley (Masterton, formerly B1 now becomes B2) is an alteration which tends to mask the undoubtedly drier regime of this interior lowland area as compared with the higher land towards the east.
Seasonal Variation of Effective Moisture
The oceanic situation of New Zealand is such that, over the country as a whole, the seasonal variety of climatic conditions is not very marked. In certain localities, however, noteworthy seasonal contrasts appear to exist and some attention has been paid to them. Kidson (Kidson, 1931b), for instance, has divided the country into three rainfall regions: (a) areas with a winter maximum, (b) areas of summer maximum, and (c) areas with February and August minima. Such distributions, however, considered without relation to temperature, are of little use as a guide to effective moisture. Over much of Kidson's summer maximum area, for example, the summer is actually the season of least effective moisture conditions.
Thornthwaite's 1948 system recognises a seasonal variation at one station only in New Zealand. This is Blenheim, where a summer seasonal deficiency is revealed. Other stations are all either r or d for their third letter. No seasonal variety is indicated by the 1931 system either, all stations in this case being also either r or d in their third letter.
There is, however, a tendency in some parts of New Zealand, especially eastern South Island, for summer moisture deficiencies to become apparent. Occasionally these result in very marked changes in farm production levels and it is undoubted farming experience that summer and, to a less extent, autumn are seasons when moisture conservation practices or some application of irrigation water, if available, are necessary if harvests are to be assured. Such contrasts between moisture conditions in summer and winter are mainly confined to the subhumid parts of the country and some indication of them seems desirable in a climate classification for use in New Zealand if it is to bring out one of the significant features of these climates.
Under the 1948 system this can be achieved very satisfactorily if the idea of water storage is neglected. On this basis a moisture index is arrived at by the formula
where S = the sum of monthly excesses of precipitation over water need, D = the sum of monthly deficits of rainfall compared with water need, and N = water need. The aridity and humidity indices are computed as percentages
of water need in the usual way, but using the values of excess and deficiency of moisture which this proposal results in.1
Figure 3 indicates the distribution of areas of summer “deficiency” when computed on this basis. It shows moisture to be “deficient” in the Napier-Hastings district of the North Island, and in the South Island in the Blenheim area and along the eastern plains from Balmoral to Palmerston South and up into high country basins. Although
|Station||Water Need||Summer Need %||Precipitation||Water Surplus||Water Deficiency||Surplus % of Need||Deficiency % of Need||Moisture Index||Climatic type|
|Walpoua Forest||27.28||38.0||64.31||37.03||0.00||136.00||0.00||136.0||AB′1 ra′|
|Onawe (Akaroa)||26.45||40.2||42.77||16.32||0.00||61.6||0.00||61.6||B3B′1 ra′|
|Alexandra||25.60||45.7||13.11||0.00||12.49||0.00||49.0||−29.4||D B′1 da′|
the pattern is the outcome of a modification of Thornthwaite's proposals such a modification is not without its justification for this part of New Zealand. It is in just these areas that moisture problems are most acutely felt in summer and the parched, dusty landscapes of these parts are familiar summer sights. Moreover, there are theoretical grounds for suspecting that soil moisture storage here is not particularly great and may fall short of the average figure of four inches maximum per month upon which Thornthwaite's postulates depend. The two major grounds for this statement are the light and frequently gravelly nature of the soils and the existence of föehn—like “Nor'-wester”—winds. These are desiccating in their effects and are especially pronounced over the Canterbury Plains, but occur also in the Blenheim area and, to a lesser extent, near Hastings (Kidson, 1932). Moreover, westerly winds are strongest and their evaporative power in eastern districts greatest during spring. The graphs shown in Figure 4 indicate that the moisture deficiencies of summer in the sub-humid climates of New Zealand are not overcome until the late winter and the amount of recharge which takes place from August onwards is important. In view of the strong winds common during this period of recharge it is not unreasonable to suspect that the amount of soil moisture storage is not the full four inches, even where precipitation is in excess of water need by more than this figure.
[Footnote] 1 Sample computations under this proposal are provided in Table 5.
|(a) Allowing for soil moisture storage|
|Water Need||22.95 inches||Water Surplus||2.53 inches||Moisture Index||3.3|
|Summer % Need||45.3%||Water Deficiency||2.95 inches||Classification||C [ unclear: ] B′1 ra′|
|Precipitation||22.53 inches||Surplus % Need||11.0%|
|Deficiency % Need||12.9%|
|(b) Without allowing for soil moisture storage|
|Water Need||22.95 inches||Water Surplus||6.53 inches||Moisture Index||10.22|
|Summer % Need||45.3%||Water Deficiency||6.95 inches||Classification||C [ unclear: ] B′1 sa′|
|Precipitation||22.53 inches||Surplus % Need||28.4%|
|Deficiency % Need||30.3%|
For these reasons, therefore, it might be a useful modification of Thornthwaite's original proposals, as far as their application in New Zealand is concerned, to neglect water storage in computing moisture indices. By so doing, the major classification for much of the country is on the whole unaffected. Such alterations as do occur are confined to the drier half of humid climates and to the subhumid group. The general effect of these alterations is to increase the major moisture index values (see Table VI) which in some cases causes
|Waipoua||136||136||AB′1 ra′||AB′1 ra′|
|Auckland||69||71||B3B′2 ra′||B3B′2 ra′|
|Gisborne||57||61||B2B′2 ra′||B3B′2 ra′|
|Hastings||20||25||C2B′1 ra′||B1B′1 sa′|
|Masterton||44||50||B2B′1 ra′||B2B′1 ra′|
|Balmoral||7||14||C2B′1 ra′||C2B′1 sa′|
|Christchurch||5||11||C2B′1 ra′||C2B′1 sa′|
|Ashburton||18||24||C2B′1 ra′||B1B′1 ra′|
|Fairlie||13||19||C2B′1 ra′||C2B′1 ra′|
|Tekapo||3||10||C2B′1 ra′||C2B′1 sa′|
|Waipiata||−14||−8||C1B′1 da′||C1B′1 sa′|
|Gore||40||42||B1B′1 ra′||B2B′1 ra′|
a change of class but also brings out the seasonal contrasts already discussed.1
Of the stations for which temperature normals are available, two (the Chateau Tongariro and Manorburn Dam) are classed as microthermal (C′2) and the remainder are mesothermal. Of the latter, four stations (Te Paki, Auckland, Gisborne, and Napier) are second category mesothermal (B′2) and the remainder are first category mesothermal (B′1). The warmer group of mesothermal climate occupies, for the most part, favourable pockets of the North Island and only in the far north of the country is the delimitation of a continuous area of any size possible. B′1 is the typical thermal category for New Zealand as a whole and stations fall into this group from the far south almost to the far north. An increase in altitude creates cooler conditions, but it is not until altitudes of above 2,500 feet above sea-level are reached that the evidence warrants the recognition of cooler climates. These are mapped in Figure 5 on the basis of the figures shown in Table I.
As regards the seasonal concentration of thermal efficiency, little need be said. All recording stations in the country indicate a result typical of oceanic situations and are, accordingly, classified as a′ in the fourth letter. This is probably not true of the colder, mountain areas, but no attempt at mapping was undertaken since, in the absence
[Footnote] 1 A similar result showing seasonal moisture variety can be achieved under the 1931 system. P–E ratios for the three summer months are summed, and if they total less than 12, i.e. ¼ of 48, the summer climate is classed as “subhumid”. A map of “subhumid” summers follows an almost identical pattern to that shown in Figure 3.
of any statistical evidence, the map would have merely followed the pattern of that showing general thermal categories.
This delimitation of thermal efficiency areas seems rather unsatisfactory. It certainly does not help to differentiate one part of the country from another on the basis of temperature to the same extent as does the 1931 system. This portrays a major thermal division between the interior and coastal regions of the North Island and between the North Island and the northern South Island, on the one
hand, and the rest of the country, on the other (see Fig. 6). Further, if a 1931 T–E index of 72 is used as a sub-division, the recognition of northern North Island as a separate thermal area can be achieved.1
[Footnote] 1 A corresponding sub-division under the 1948 system would appear to be a water need of 24.74 inches. This achieves little. Reference to Table 7 shows that the north-south contrasts achieved under the 1931 system are not obtained by the new system even with the addition of this suggested refinement.
The altitudinal limits of thermal boundaries devised under the 1931 system are also encouraging, especially in the distinction between taiga and microthermal climates. This limit lies at about 3,000 feet of elevation in the south and from 4,500 to 5,000 feet above sea-level in the north.1
The general conformity of these elevations with the altitudinal zoning of vegetation in New Zealand has been referred to elsewhere (Garnier, 1946). Further investigation has simply served to support these earlier statements. In the mountain area near lakes Pukaki, Ohau, and Tekapo, for example, where taiga climate appears in theory at about 3,500 feet above sea-level, recent observation by the writer has shown that the altitudinal limits of tall trees lies between 3,500 and 4,000 feet. Moreover, observations on the occurrence of snow grass (Danthonia flavescens) in the same general locality show a distinct tendency for it to become dominant at altitudes which, except under special local circumstance, are from 3,200 to 3,500 feet above sea-level.
|1948 Classification||1931 Classification|
|Station||Latitude South||Altitude in feet||Water Need (inches)||Thermal Category||Seasonal Concentration||T—E Index||Thermal Category||Seasonal Concentration|
Reference to Table VII shows that the 1931 system recognises a differentiation between stations on the basis of the seasonal concentration of thermal efficiency which its successor does not achieve. Under the earlier classification several east coast stations, especially the cooler ones, have b as the fourth letter of their classification in contrast to a which is characteristic of the majority of lowland stations and is uniformly found in the north, the west, and the south. In addition, one should note the contrast between interior and coastal stations revealed by the fourth letter of the 1931 system but masked under the new classification.
It seems reasonably clear, therefore, that Thornthwaite's new system is not so successful as his earlier one for the delimitation of thermal contrasts in New Zealand. To follow it would show no difference between the temperature conditions of such contrasting stations as Invercargill, Timaru, the Hermitage at Mount Cook, Hastings,
[Footnote] 1 It varies from 4,500 feet in the south to 6,000 feet in the north under the 1948 system.
Palmerston North and Waipoua forest. These differences contribute markedly to the climatic contrasts within the country and should be revealed by a system for use here.
This survey of Thornthwaite's new proposals for climate classification seems to indicate that they yield a system which, like its predecessor, usefully differentiates the major moisture types of the country but fails to indicate tendencies towards seasonal contrasts in moisture efficiency which observation and experience suggest are there. A modification makes such contrasts appear, but it can be reasonably argued that the modification proposed is against the whole philosophy behind the original proposals. As regards temperature divisions, the 1948 scheme does not appear so satisfactory as the earlier one. Little differentiation between north and south is achieved, the only major contrasts being those between mesothermal climates below 2,500 feet of elevation, and the microthermal and colder climates above.
It can be objected that criticisms of this type are not valid, since they not only assume that patterns of vegetation, soil, and other earth features should fit those of climate, whereas the system proposed is based upon a different viewpoint, but also make the a priori assumption that differences exist in thermal conditions between different parts and the moisture effectiveness of different seasons which are sufficiently pronounced for their recognition to be possible in general, as opposed to detailed, studies.
The grounds on which these contrasts are postulated are largely those of vegetation and soil characteristics and also general observation and experience, especially of farming practice and problems. These grounds, admittedly subjective as they in large part are, are the best we have to go on in the absence of controlled experiment. Furthermore, provided they are the product of carefully reflective judgment of observed situations, they constitute a guide to the climatic pattern of a country which, if not objectively precise, at least indicates the general regional framework. It seems not unreasonable, therefore, to expect that a classification of climate should reveal, in general terms, those diversities in climatic characteristics which the considerations referred to suggest are significant elements in regional climates. These elements include, in New Zealand, not only differences in moisture effectiveness in different parts, but also thermal contrasts, both between highland and lowland and between north and south, and seasonal moisture contrasts as a feature of the drier portions of the country.
If these statements are accepted, one must conclude that the most successful part of Thornthwaite's new proposals lies in their delimitation of moisture regions. But the difference between the new and the old in this respect is very little and seems hardly worth the tremendous effort of calculation involved in the 1948 system. This is in no way to imply that one must discard a classification simply because the computations involved are tedious. However complicated the mathematics of a system, it must surely be the chosen one if the results obtained are even slightly in advance of a simpler one. In the present instance, however, the application of the system to New Zealand does
not seem to provide an improvement in the portrayal of the country's climatic variety and it seems unreasonable to give up a procedure which is relatively simple and quick in its computations in favour of one involving very lengthy calculations, however empirical the former or rational the latter.
One cannot fail to be impressed, however, by the idea behind the new system of classification. Its explicit statement of the concept of potential evapotranspiration and its use of the relationship of this to the supply of moisture actually available is a much clearer exposition of the really significant features of climate than was produced by former ideas which centred mainly upon the distribution of precipitation effectiveness. It is hoped to report in more detail on the application of this concept to moisture problems in New Zealand in a subsequent paper.
Garnier, B. J., 1946. The Climates of New Zealand: According to Thornthwaite's
Classification. Ann. Assn. Amer. Geographers, vol. 36, no. 3, pp. 151–177.
Kidson, E., 1931a. Mean Temperature in New Zealand. N.Z. Journ. Sci. and Tech.,
vol. 13, pp. 140–153.
——, E., 1931b. The Annual Variation of Rainfall in New Zealand. N.Z. Journ.
Sci. and Tech., vol. 12, pp. 268–271.
——, E., 1932. The Canterbury “North-wester.” N.Z. Journ. Sci. and Tech., vol. 14,
Thornthwaite, C. W., 1931. The Climates of North America according to a New
Classification. Geogr. Rev., vol. 21, no. 4, pp. 633–655.
——, C. W., 1948. An Approach toward a Rational Classification of Climate. Geogr.
Rev., vol. 38, no. 1, pp. 55–94.