Combining farmers’ and scientists’ tree species and soil fertility assessment for improved cropping decisions in swidden systems of Northern Thailand

Abstract Despite increasing pressure on land, many hill tribe communities in Northern Thailand practice traditional swidden agriculture with long fallow periods that allow the soil to recover from cropping. To determine the most appropriate time for a new cropping cycle, farmers, in a collective decision making process, often use presence or numbers of tree species considered as soil fertility indicators. In Nong Khao, a case study village in Mae Hong Son district, we investigated in how far abundance of indicator tree species found during vegetation surveys corresponded to measured soil fertility parameters and whether an extended list of indicator species could improve prediction of these soil properties. During interviews, farmers listed 11 tree species considered indicators for good upland rice and vegetable soil, unsuitable soil for upland rice, or hard soil. Abundances of indicator species as well as of other species found during comprehensive botanical inventories on 135 plots of different fallow ages were introduced into regression models to predict soil fertility parameters measured on the inventoried plots. Both models – including only indicator species versus all species found in the inventories – were compared regarding their predictive power for soil properties. Soil organic matter (SOM), pH, plant available phosphorus (Pav) – related to farmers’ criteria ‘good soil’ or 'unsuitable for rice cropping’ – as well as bulk density (BD, for ‘hard soil’), changed significantly during the fallow period, from initial maxima towards temporary pessima in years 6 to 7, followed by recovery towards year ten. Most indicator species, like Macaranga denticulata for enhanced Pav or Dalbergia cultrata for high SOM, were clearly related to the soil quality characteristics attributed to them by farmers. Including all inventoried tree species significantly improved predictions of measured soil parameters by ΔAIC > |2|. Ten additional species from the surveys had potential as additional indicator species. Our approach to relate indicator species to measured soil parameters is generic across areas, while soil parameters are site-specific. Applicability of our approach to link environmental parameters to indicator species could be extended if further criteria considered important by farmers such as weed suppression, represented by tree structure parameters, would complement soil fertility.